{"templateId":"markdown","sharedDataIds":{"sidebar":"sidebar-@l10n/ja/sidebars.yaml"},"props":{"metadata":{"markdoc":{"tagList":["admonition"]},"type":"markdown"},"seo":{"title":"マルチタッチアトリビューション(MTA)","description":"Treasure AIのAI Signals「マルチタッチアトリビューション(MTA)」で、ファーストタッチやラストタッチのルールではなく、MarkovモデルとShapleyモデルによるデータドリブンな分析で、コンバージョンに貢献しているマーケティングチャネルを特定します。","siteUrl":"https://docs.treasure.ai","image":"/assets/treasure-ai-social-preview.e6fefb35676f6da91c2d1bc783bd39e618e7af2b1ad3aa52a96ae69d72e700fe.b1bff462.png","lang":"en-US","jsonLd":{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://www.treasure.ai/","name":"Treasure AI","url":"https://www.treasure.ai/","logo":"https://www.treasure.ai/hubfs/assets/images/logos/primary-logo.svg"},{"@type":"WebSite","@id":"https://docs.treasure.ai/#website","name":"Treasure AI Documentation","url":"https://docs.treasure.ai/","inLanguage":["en","ja"],"publisher":{"@id":"https://www.treasure.ai/"}}]}},"dynamicMarkdocComponents":[],"compilationErrors":[],"ast":{"$$mdtype":"Tag","name":"article","attributes":{},"children":[{"$$mdtype":"Tag","name":"Heading","attributes":{"level":1,"id":"マルチタッチアトリビューションmta","__idx":0},"children":["マルチタッチアトリビューション(MTA)"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["最初や最後にたまたま接触したチャネルではなく、実際にコンバージョンを生み出しているマーケティングチャネルを見極めます。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":2,"id":"概要とユースケース","__idx":1},"children":["概要とユースケース"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["マルチタッチアトリビューション(MTA)AI Signalsは、顧客が実際にたどるジャーニーに基づいて、各マーケティングチャネルにコンバージョンへの貢献度を割り当てます。ファーストタッチ、ラストタッチ、リニア、U字型といったルールベースのアトリビューションモデルは、ジャーニー内の位置に応じて貢献度を割り当てます。説明しやすい反面、チャネル同士がどのように連携して働くかは考慮されません。MTA AI Signalsは、代わりにお客様のデータから学習します。"]},{"$$mdtype":"Tag","name":"ul","attributes":{},"children":[{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["Markov"]}," は、ジャーニーをチャネルからチャネルへの遷移の連鎖としてモデル化します。連鎖から特定のチャネルを取り除いたときにコンバージョンがどう変化するかを調べることで、各チャネルの貢献度を測定します。また、顧客があるチャネルから次のチャネルへどのように移動するかを示す ",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["遷移行列"]}," も出力します。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["Shapley"]}," は、ジャーニー内のチャネルをチームとして扱います。そのチャネルが含まれるあらゆるチャネルの組み合わせにわたって、コンバージョン率に対する各チャネルの平均的な貢献を測定するため、同時に現れることの多いチャネル同士で貢献度が公平に分配されます。"]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["主なユースケース"]}]},{"$$mdtype":"Tag","name":"ul","attributes":{},"children":[{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["予算配分"]},": 測定された貢献度に基づいて、マーケティング予算をチャネル間で配分する"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["チャネルのWhat-if分析"]},": チャネルを削減または拡大する前に、そのチャネルの重要度を見積もる"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["ジャーニーフロー分析"]},": Markovの遷移行列からSankeyなどのフロー図を作成し、次にどのチャネルが来ることが多いかを把握する"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["ラストタッチの影響度"]},": Shapleyのルックバックサイズを使って、コンバージョン直前の数回の接触で特に重要なチャネルを把握する"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["モデルの比較"]},": データドリブンな貢献度を、ルールベースのモデルと比較する。両者に大きな差がある場合は調査する価値がある"]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["特にメリットが大きい方:"]}," タッチポイントデータ(Web、メール、広告、その他のチャネル)とコンバージョンイベントをすでにTreasure AIに保有している、マーケティングアナリスト、データエンジニア、データサイエンティスト。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"他のai-signalsとの使い分け","__idx":2},"children":["他のAI Signalsとの使い分け"]},{"$$mdtype":"Tag","name":"div","attributes":{"className":"md-table-wrapper"},"children":[{"$$mdtype":"Tag","name":"table","attributes":{"className":"md"},"children":[{"$$mdtype":"Tag","name":"thead","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"th","attributes":{"width":"45%","data-label":"知りたいこと"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["知りたいこと"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"55%","data-label":"使用するソリューション"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["使用するソリューション"]}," "]}]}]},{"$$mdtype":"Tag","name":"tbody","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["過去の購買履歴に基づいて、重要な顧客は誰か?"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"/ja/products/customer-data-platform/machine-learning/ai-signals/rfm-ai-signals"},"children":["RFM AI Signals"]},": モデリング不要の記述的セグメント"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["この顧客がXを実行する可能性はどのくらいか?"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"/ja/products/customer-data-platform/machine-learning/ai-signals/propensity-ai-signals"},"children":["Propensity Scoring AI Signals"]},": 定義したイベントごとの確率"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["この顧客は将来どれだけの価値をもたらすか?"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"/ja/products/customer-data-platform/machine-learning/ai-signals/cltv-ai-signals"},"children":["CLTV AI Signals"]},": 金額の予測とパーセンタイルランク"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["次にどの商品をレコメンドすべきか?"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"/ja/products/customer-data-platform/machine-learning/ai-signals/nbp-ai-signals"},"children":["NBP AI Signals"]},": 商品・サービス・コンテンツのランク付きリスト"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["この顧客にどのアクションを取るべきか?"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"/ja/products/customer-data-platform/machine-learning/ai-signals/nba-ai-signals"},"children":["NBA AI Signals"]},": 候補アクションから学習したユーザーごとのポリシー"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["どのマーケティングチャネルがコンバージョンに貢献しているか?"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["MTA AI Signals(現在のページ)"]},": データドリブンなチャネルごとの貢献度"]}]}]}]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["MTAと顧客単位のシグナルのどちらを使うべきか?"]}," 他のAI Signalsは、顧客一人ひとりに対してスコアやレコメンドを出力します。MTAはチャネル単位で動作します。1人の顧客が次に何をするかを予測するのではなく、全顧客にわたる過去のコンバージョンを説明します。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["MarkovとShapleyは、少し異なる問いに答えます。Markovは、どのチャネルが代替しにくいか、顧客がチャネル間をどう移動するかを示します。Shapleyは、同時に現れることの多いチャネル間で貢献度を公平に分配する方法を示します。両方を実行することをおすすめします。詳細は",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"#%E3%83%A2%E3%83%87%E3%83%AB%E3%81%AE%E9%81%B8%E3%81%B3%E6%96%B9"},"children":["モデルの選び方"]},"を参照してください。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":2,"id":"クイックスタート","__idx":3},"children":["クイックスタート"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["この例では、1つのタッチポイントテーブルから2つのモデルを並行して実行します。テーブルの各行は1回の接触を表し、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["user_id"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["touch_time"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["channel"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["converted"]}," の列を持ちます。このワークフローでは、元の接触テーブルの ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["converted"]}," を整数の ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["0"]}," または ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["1"]}," として保存してください。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["開始する前に、次の点を確認してください。"]},{"$$mdtype":"Tag","name":"ul","attributes":{},"children":[{"$$mdtype":"Tag","name":"li","attributes":{},"children":["お客様のアカウントでMTAが有効になっている必要があります。有効になっていない場合は、Treasure AIのアカウントチームにお問い合わせください。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":["使用するAPIキーに、入力データベースの読み取り権限と出力データベースの書き込み権限が必要です。"]}]},{"$$mdtype":"Tag","name":"Admonition","attributes":{"type":"warning","name":"セキュリティ"},"children":[{"$$mdtype":"Tag","name":"p","attributes":{},"children":["APIキーは、ワークフロー定義にリテラル値として記述せず、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["td.apikey"]}," という名前のWorkflowシークレットとして保存してください。これによりキーがバージョン管理に含まれなくなります。詳細は ",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"/products/customer-data-platform/data-workbench/workflows/setting-workflow-secrets-from-td-console"},"children":["Treasure コンソール からのワークフローシークレットの設定"]}," を参照してください。"]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["Markovは1ジャーニーを1行とするテーブルを必要とするため、ワークフローはまず接触テーブルからジャーニーテーブルを作成します。Shapleyは接触テーブルを直接読み込みます。"]},{"$$mdtype":"Tag","name":"CodeBlock","attributes":{"data-language":"yaml","header":{"controls":{"copy":{}}},"source":"# mta_ai_signals.dig\ntimezone: UTC\n\n_export:\n  ml_batch_endpoint: \"https://ml-batch-api.treasuredata.com\"\n  db: \"my_mta_db\"\n  touch_table: \"mta_touches\"              # 1接触につき1行\n  markov_input_table: \"mta_markov_input\"  # 1ジャーニーにつき1行(以下で作成)\n\n+build_markov_input:\n  td>: queries/build_markov_input.sql     # 「mta_markov > データの準備」を参照\n  database: ${db}\n  create_table: ${markov_input_table}\n\n+attribution:\n  _parallel: true\n\n  +markov:\n    +submit:\n      http>: ${ml_batch_endpoint}/v1/runs/\n      method: POST\n      headers:\n        - authorization: \"TD1 ${secret:td.apikey}\"\n        - X-TD-ML-SESSION-ID: ${session_id}\n        - X-TD-ML-ATTEMPT-ID: ${attempt_id}\n      content_format: json\n      content:\n        solution_name: \"mta_markov\"\n        input_table: \"${db}.${markov_input_table}\"\n        output_table: \"${db}.mta_markov_attribution\"\n        solution_arguments:\n          user_id_col: \"journey_id\"\n          time_col: \"touch_time\"\n          channel_colname: \"journey\"\n          conversion_col_name: \"converted\"\n          transition_output_table: \"${db}.mta_markov_transition\"\n      store_content: true\n    +print:\n      echo>: ${http.last_content}\n    +poll:\n      http>: ${ml_batch_endpoint}/v1/runs/${JSON.parse(http.last_content)['id']}/status\n      method: GET\n      headers:\n        - authorization: \"TD1 ${secret:td.apikey}\"\n\n  +shapley:\n    +submit:\n      http>: ${ml_batch_endpoint}/v1/runs/\n      method: POST\n      headers:\n        - authorization: \"TD1 ${secret:td.apikey}\"\n        - X-TD-ML-SESSION-ID: ${session_id}\n        - X-TD-ML-ATTEMPT-ID: ${attempt_id}\n      content_format: json\n      content:\n        solution_name: \"mta_shapley\"\n        input_table: \"${db}.${touch_table}\"\n        output_table: \"${db}.mta_shapley_attribution\"\n        solution_arguments:\n          user_id_col: \"user_id\"\n          time_col: \"touch_time\"\n          channel_colname: \"channel\"\n          conversion_col_name: \"converted\"\n          shapley_lookback_window: 4\n          shapley_order: true\n      store_content: true\n    +print:\n      echo>: ${http.last_content}\n    +poll:\n      http>: ${ml_batch_endpoint}/v1/runs/${JSON.parse(http.last_content)['id']}/status\n      method: GET\n      headers:\n        - authorization: \"TD1 ${secret:td.apikey}\"\n","lang":"yaml"},"children":[]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["各ブランチは、送信(submit)とポーリング(poll)のステップを1つのグループにまとめて保持しているため、ブランチが並行して実行されても、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["http.last_content"]}," は常にそのブランチ自身の実行を指します。ステータスエンドポイントは、実行中はHTTP ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["408"]}," を返します。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["http>"]}," オペレーターは実行が完了するまでリトライを続け、完了すると最終的な ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["status"]}," と ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["error_messages"]}," を返します。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["想定される出力:"]}," ワークフローが完了すると、3つのテーブルが作成されます。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_markov_attribution"]},": チャネルごとの貢献度"]},{"$$mdtype":"Tag","name":"CodeBlock","attributes":{"header":{"controls":{"copy":{}}},"source":"channels  attribution_markov_algorithmic  removal_effect\nemail     1.358                           0.750\nsocial    1.132                           0.625\norganic   1.509                           0.833\ndisplay   0.000                           0.000\n"},"children":[]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_markov_transition"]},": あるチャネルから次のチャネルへ移動する確率"]},{"$$mdtype":"Tag","name":"CodeBlock","attributes":{"header":{"controls":{"copy":{}}},"source":"channels  email   social  organic  _conversion_\nemail     0.000   0.400   0.200    0.400\nsocial    0.250   0.000   0.500    0.250\norganic   0.000   0.000   0.000    1.000\n"},"children":[]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_shapley_attribution"]},": ルックバックサイズごと、チャネルごとの貢献度"]},{"$$mdtype":"Tag","name":"CodeBlock","attributes":{"header":{"controls":{"copy":{}}},"source":"channels  variable                                              value  size\nemail     attribution_shapley_size0_conv_rate_order_algorithmic  1.42   0\norganic   attribution_shapley_size0_conv_rate_order_algorithmic  1.61   0\nemail     attribution_shapley_size2_conv_rate_order_algorithmic  1.18   2\norganic   attribution_shapley_size2_conv_rate_order_algorithmic  1.87   2\n"},"children":[]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["上記の値は例です。実際の結果はお客様のデータによって異なります。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["このワークフローの作成とスケジュール設定については、",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"/products/customer-data-platform/data-workbench/workflows/getting-started-with-treasure-workflow"},"children":["Treasure Workflowを使ってみる"]},"を参照してください。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":2,"id":"モデルの設定","__idx":4},"children":["モデルの設定"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["2つのソリューションはどちらも、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["user_id_col"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["time_col"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["channel_colname"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["conversion_col_name"]}," という同じ4つの列パラメータを使用します。列名はハードコードされていないため、自由に指定できます。ただし、2つのソリューションは ",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["異なる形のテーブル"]}," を読み込むため、入力を入れ替えることはできません。"]},{"$$mdtype":"Tag","name":"ul","attributes":{},"children":[{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_markov"]}," は ",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["1ジャーニーにつき1行"]}," を読み込み、チャネルの経路全体を1つの列に格納します。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_shapley"]}," は ",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["1接触につき1行"]}," を読み込みます。"]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["どちらの入力も、同じタッチポイントデータから作成できます。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["次のフィールドは、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["solution_arguments"]}," の内部ではなく、両方のソリューションで実行リクエストの最上位に設定します。"]},{"$$mdtype":"Tag","name":"div","attributes":{"className":"md-table-wrapper"},"children":[{"$$mdtype":"Tag","name":"table","attributes":{"className":"md"},"children":[{"$$mdtype":"Tag","name":"thead","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"th","attributes":{"width":"18%","data-label":"パラメータ"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["パラメータ"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"8%","data-label":"型"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["型"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"10%","data-label":"デフォルト"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["デフォルト"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"9%","data-label":"必須"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["必須"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"55%","data-label":"説明"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["説明"]}," "]}]}]},{"$$mdtype":"Tag","name":"tbody","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["solution_name"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_markov"]}," または ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_shapley"]},"。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["input_table"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["入力テーブル。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["database.table"]}," 形式。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["output_table"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["アトリビューション結果の出力先テーブル。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["database.table"]}," 形式。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["solution_arguments"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["object"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ソリューション固有の設定。以下のセクションを参照してください。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["version"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["stable"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["いいえ"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ソリューションのバージョン。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["stable"]}," を使用してください。"]}]}]}]}]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"mta_markov","__idx":5},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_markov"]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["Markovチェーンによるアトリビューションでは、各ジャーニーを、開始状態から一連のチャネルを経て、コンバージョンまたはジャーニーの終了に至る歩みとして扱います。コンバージョンに至ったすべてのジャーニーから、顧客が各チャネルから次の各チャネルへ移動する確率を推定します。次に、各チャネルの ",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["除去効果(removal effect)"]}," を測定します。これは、コンバージョンに至る経路がそのチャネルにどれだけ依存しているかを表します。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":4,"id":"データの準備","__idx":6},"children":["データの準備"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["入力テーブルは ",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["1ジャーニーにつき1行"]}," です。各ジャーニーは、コンバージョンの時点で分割済みであり、コンバージョンとなった接触でジャーニーが終わる必要があります。これにより、コンバージョン後に発生した接触が、そのコンバージョンに至った経路に含まれることを防ぎます。"]},{"$$mdtype":"Tag","name":"div","attributes":{"className":"md-table-wrapper"},"children":[{"$$mdtype":"Tag","name":"table","attributes":{"className":"md"},"children":[{"$$mdtype":"Tag","name":"thead","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"th","attributes":{"width":"28%","data-label":"フィールド"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["フィールド"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"14%","data-label":"型"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["型"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"10%","data-label":"必須"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["必須"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"48%","data-label":"説明"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["説明"]}," "]}]}]},{"$$mdtype":"Tag","name":"tbody","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ジャーニーID(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["user_id_col"]},")"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["varchar"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ジャーニーごとの一意のID。1人の顧客が複数のジャーニーを持つことがあります。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["経路(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["channel_colname"]},")"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["varchar"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ジャーニー内のチャネルを順に、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["' > '"]},"(スペース、大なり記号、スペース)で連結したもの。例: ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["email > social > organic"]},"。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["コンバージョン(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["conversion_col_name"]},")"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["int"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ジャーニーがコンバージョンで終わる場合は ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["1"]},"、それ以外は ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["0"]},"。以下の準備用SQLは整数のフラグを出力します。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["時刻(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["time_col"]},")"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["timestamp / varchar"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ジャーニーの最後の接触の時刻。Markovはアトリビューションの計算にこの列を使用しません。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["max_load_rows"]}," を設定した際に、どのジャーニーを先に読み込むかを決めるためにのみ使用されます。"]}]}]}]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["行の例:"]},{"$$mdtype":"Tag","name":"CodeBlock","attributes":{"header":{"controls":{"copy":{}}},"source":"journey_id  journey                           converted  touch_time\ncust_001-0  email > paid_search > organic     1          2026-08-02 10:15:00\ncust_001-1  social > email                    0          2026-08-20 18:40:00\ncust_002-0  display                           0          2026-08-11 09:02:00\ncust_003-0  paid_search > paid_search > email 1          2026-08-14 21:30:00\n"},"children":[]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["ワークフローの ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["touch_table"]}," から、ウィンドウ関数を使ってこのテーブルを作成できます。次のクエリは、コンバージョンのたびに新しいジャーニーを開始し、チャネルを時系列順に連結します。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["クエリでは ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["SUM(converted)"]}," を使用するため、元の ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["converted"]}," 列には整数の ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["0"]}," または ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["1"]}," を格納する必要があります。テキストのフラグは、クエリの実行前にこれらの値に変換してください。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["'true'"]}," や ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["'false'"]}," を直接整数にキャストすることはできません。認識できない値はNULLに変換せず、データ準備の段階でエラーとして扱ってください。"]},{"$$mdtype":"Tag","name":"CodeBlock","attributes":{"data-language":"sql","header":{"controls":{"copy":{}}},"source":"-- queries/build_markov_input.sql\nWITH segmented AS (\n  SELECT\n    user_id,\n    touch_time,\n    channel,\n    converted,\n    COALESCE(SUM(converted) OVER (\n      PARTITION BY user_id ORDER BY touch_time, channel\n      ROWS BETWEEN UNBOUNDED PRECEDING AND 1 PRECEDING), 0) AS journey_index\n  FROM ${touch_table}\n  WHERE user_id IS NOT NULL AND channel IS NOT NULL\n)\nSELECT\n  CONCAT(user_id, '-', CAST(journey_index AS VARCHAR))                 AS journey_id,\n  MAX(touch_time)                                                      AS touch_time,\n  ARRAY_JOIN(ARRAY_AGG(channel ORDER BY touch_time, channel), ' > ')   AS journey,\n  CAST(MAX(converted) AS BIGINT)                                       AS converted\nFROM segmented\nGROUP BY user_id, journey_index\n","lang":"sql"},"children":[]},{"$$mdtype":"Tag","name":"Admonition","attributes":{"type":"warning","name":"データ要件"},"children":[{"$$mdtype":"Tag","name":"ul","attributes":{},"children":[{"$$mdtype":"Tag","name":"li","attributes":{},"children":["Markovに接触単位の行を渡さないでください。テーブル内のどの行にも ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["' > '"]}," 区切り文字が含まれていない場合、各行がジャーニーではなく単一の接触であるとみなされ、実行は拒否されます。複数回の接触を含むジャーニーが一部にあれば、単一の接触のみのジャーニーが含まれていても問題ありません。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":["入力には、少なくとも1件のコンバージョンが含まれている必要があります。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":["ジャーニーIDまたは経路がNULLの行はスキップされます。"]}]}]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":4,"id":"ワークフローパラメータ","__idx":7},"children":["ワークフローパラメータ"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["次のパラメータは ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["solution_arguments"]}," の内部に設定します。"]},{"$$mdtype":"Tag","name":"div","attributes":{"className":"md-table-wrapper"},"children":[{"$$mdtype":"Tag","name":"table","attributes":{"className":"md"},"children":[{"$$mdtype":"Tag","name":"thead","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"th","attributes":{"width":"22%","data-label":"パラメータ"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["パラメータ"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"8%","data-label":"型"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["型"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"12%","data-label":"デフォルト"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["デフォルト"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"9%","data-label":"必須"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["必須"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"49%","data-label":"説明"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["説明"]}," "]}]}]},{"$$mdtype":"Tag","name":"tbody","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["user_id_col"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ジャーニーIDの列。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["time_col"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ジャーニーのタイムスタンプ列(最後の接触)。件数を制限して読み込む際の並び替えにのみ使用されます。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["channel_colname"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["' > '"]}," で連結されたチャネル経路を保持する列。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["conversion_col_name"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["コンバージョンフラグの列。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["transition_output_table"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["いいえ"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["遷移行列の出力先テーブル。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["database.table"]}," 形式。省略した場合、遷移行列は書き込まれません。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["max_load_rows"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["integer"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["0"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["いいえ"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["読み込むジャーニーの最大数。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["0"]}," は無制限を意味します。コンバージョンしたジャーニーが先に読み込まれ、その後、新しいものから読み込まれます。",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"#%E5%85%A5%E5%8A%9B%E3%82%B5%E3%82%A4%E3%82%BA%E3%81%AE%E5%88%B6%E5%BE%A1"},"children":["入力サイズの制御"]},"を参照してください。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["output_mode"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["overwrite"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["いいえ"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["overwrite"]}," または ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["append"]},"。",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"#%E5%87%BA%E5%8A%9B%E3%83%A2%E3%83%BC%E3%83%89"},"children":["出力モード"]},"を参照してください。"]}]}]}]}]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":4,"id":"出力","__idx":8},"children":["出力"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["アトリビューションテーブル(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["output_table"]},")"]}]},{"$$mdtype":"Tag","name":"div","attributes":{"className":"md-table-wrapper"},"children":[{"$$mdtype":"Tag","name":"table","attributes":{"className":"md"},"children":[{"$$mdtype":"Tag","name":"thead","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"th","attributes":{"width":"30%","data-label":"フィールド"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["フィールド"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"10%","data-label":"型"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["型"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"60%","data-label":"説明"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["説明"]}," "]}]}]},{"$$mdtype":"Tag","name":"tbody","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["channels"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["varchar"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["チャネルのラベル。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["attribution_markov_algorithmic"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["double"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["チャネルに割り当てられたコンバージョン数。この列の合計は100%ではなく、コンバージョンしたジャーニーの総数になります。小数第3位に丸められます。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["removal_effect"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["double"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["コンバージョンに至る経路がそのチャネルにどれだけ依存しているか。0から1の値。小数第3位に丸められます。"]}]}]}]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["各チャネルの貢献度の割合を求めるには、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["attribution_markov_algorithmic"]}," を全チャネルの合計で割ります。コンバージョンしなかったジャーニーにのみ現れるチャネルも、貢献度0、除去効果0として一覧に表示されます。これは想定された動作です。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["遷移行列(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["transition_output_table"]},")"]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["横に広いテーブルです。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["channels"]}," 列は、顧客の移動 ",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["元"]}," のチャネルです。移動 ",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["先"]}," となりうるチャネルごとに1列があり、さらにコンバージョン状態を表す列(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["_conversion_"]},")があります。各値は確率で、小数第3位に丸められます。列はお客様のチャネルラベルから生成されるため、列の構成はデータによって異なります。ラベルに関するヒントは、",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"#%E3%83%81%E3%83%A3%E3%83%8D%E3%83%AB%E3%83%A9%E3%83%99%E3%83%AB%E3%81%AE%E7%AE%A1%E7%90%86"},"children":["チャネルラベルの管理"]},"を参照してください。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["channels"]}," の行には、コンバージョンに至る経路の終わりを表す合成の吸収状態 ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["(conversion)"]}," が含まれることもあります。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["遷移行列は、Sankey図やフロー図の作成や、あるチャネルの次に最も来やすいチャネルを調べるために使用します。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"mta_shapley","__idx":9},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_shapley"]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["Shapley値によるアトリビューションは、協力ゲーム理論に由来します。ジャーニー内のチャネルの集合を ",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["提携(coalition)"]}," とし、各提携は、それを含むジャーニーの ",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["コンバージョン率"]}," でスコアリングされます。チャネルのShapley値は、そのチャネルが加わりうるすべての提携にわたる、コンバージョン率への平均的な貢献です。ほぼすべてのジャーニーに現れるチャネルは、存在しているだけでは貢献度を得られません。上乗せしたコンバージョンに対してのみ貢献度が与えられます。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":4,"id":"データの準備-1","__idx":10},"children":["データの準備"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["入力テーブルは ",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["1接触につき1行"]}," です。このソリューションは、接触をユーザーごとにジャーニーにまとめ、時刻順に並べ、コンバージョンのたびに新しいジャーニーを開始します。そのため、1人の顧客が、コンバージョンしたジャーニーと、それに続くコンバージョンしていないジャーニーの両方を持つことがあります。"]},{"$$mdtype":"Tag","name":"div","attributes":{"className":"md-table-wrapper"},"children":[{"$$mdtype":"Tag","name":"table","attributes":{"className":"md"},"children":[{"$$mdtype":"Tag","name":"thead","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"th","attributes":{"width":"28%","data-label":"フィールド"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["フィールド"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"14%","data-label":"型"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["型"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"10%","data-label":"必須"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["必須"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"48%","data-label":"説明"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["説明"]}," "]}]}]},{"$$mdtype":"Tag","name":"tbody","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ユーザー(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["user_id_col"]},")"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["varchar"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["顧客またはユーザーのID。ジャーニーIDを渡すこともできます。ジャーニーがすでにコンバージョンの時点で分割されている場合も、結果は同じです。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["時刻(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["time_col"]},")"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["timestamp / varchar"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["接触のタイムスタンプ。timestamp型、または ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["YYYY-MM-DD HH:MM:SS"]}," 形式のテキスト。",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["Unixエポックの整数は使用しないでください。"]}," 誤った日付として読み込まれます。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["チャネル(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["channel_colname"]},")"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["varchar"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["1行につき1つのチャネルラベル。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["コンバージョン(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["conversion_col_name"]},")"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["int / boolean"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["コンバージョンとなった接触は ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["1"]},"、それ以外は ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["0"]},"。"]}]}]}]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["行の例:"]},{"$$mdtype":"Tag","name":"CodeBlock","attributes":{"header":{"controls":{"copy":{}}},"source":"user_id   touch_time           channel      converted\ncust_001  2026-08-01 09:00:00  email        0\ncust_001  2026-08-01 12:30:00  paid_search  0\ncust_001  2026-08-02 10:15:00  organic      1\ncust_001  2026-08-19 08:00:00  social       0\ncust_001  2026-08-20 18:40:00  email        0\ncust_002  2026-08-11 09:02:00  display      0\n"},"children":[]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["この例では、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["cust_001"]}," は2つのジャーニーを持ちます。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["organic"]}," で終わるコンバージョンしたジャーニーと、コンバージョンしていないジャーニー(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["social > email"]},")です。"]},{"$$mdtype":"Tag","name":"Admonition","attributes":{"type":"warning","name":"データ要件"},"children":[{"$$mdtype":"Tag","name":"ul","attributes":{},"children":[{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["コンバージョンしていないジャーニーを含めてください。"]}," Shapleyは各提携をコンバージョン率でスコアリングするため、その率を計算するにはコンバージョンしなかったジャーニーが必要です。コンバージョンしたジャーニーのみを渡すと、すべての提携のコンバージョン率が100%になり、結果が意味をなしません。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":["入力には、少なくとも1件のコンバージョンが含まれている必要があります。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":["ユーザーIDまたはチャネルがNULLの行はスキップされます。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":["2つの接触のタイムスタンプが同じ場合は、チャネル名の順で並べられるため、実行ごとに結果は同じになります。"]}]}]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":4,"id":"ワークフローパラメータ-1","__idx":11},"children":["ワークフローパラメータ"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["次のパラメータは ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["solution_arguments"]}," の内部に設定します。"]},{"$$mdtype":"Tag","name":"div","attributes":{"className":"md-table-wrapper"},"children":[{"$$mdtype":"Tag","name":"table","attributes":{"className":"md"},"children":[{"$$mdtype":"Tag","name":"thead","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"th","attributes":{"width":"22%","data-label":"パラメータ"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["パラメータ"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"8%","data-label":"型"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["型"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"10%","data-label":"デフォルト"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["デフォルト"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"9%","data-label":"必須"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["必須"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"51%","data-label":"説明"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["説明"]}," "]}]}]},{"$$mdtype":"Tag","name":"tbody","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["user_id_col"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ユーザー(またはジャーニー)IDの列。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["time_col"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["接触のタイムスタンプ列。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["channel_colname"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["チャネルの列。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["conversion_col_name"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["—"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["コンバージョンフラグの列。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["shapley_lookback_window"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["integer"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["4"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["いいえ"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["スコアリングする最大のルックバックサイズ。1から10。サイズ0からこの値までの各サイズについて、1つずつモデルが実行されます。",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"#shapley%E3%81%AE%E3%83%AB%E3%83%83%E3%82%AF%E3%83%90%E3%83%83%E3%82%AF%E3%82%A6%E3%82%A3%E3%83%B3%E3%83%89%E3%82%A6%E3%81%A8%E9%A0%86%E5%BA%8F"},"children":["Shapleyのルックバックウィンドウと順序"]},"を参照してください。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["shapley_order"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["boolean"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["true"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["いいえ"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["提携内でチャネルの順序を考慮するかどうか。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["true"]}," の場合、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["email > social"]}," と ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["social > email"]}," は異なる提携として扱われます。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["max_load_rows"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["integer"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["0"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["いいえ"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["読み込む行の最大数。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["0"]}," は無制限を意味します。ジャーニー単位で、コンバージョンしたジャーニーから先に読み込まれるため、上限は最大でも1ジャーニー分だけ超過することがあります。",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"#%E5%85%A5%E5%8A%9B%E3%82%B5%E3%82%A4%E3%82%BA%E3%81%AE%E5%88%B6%E5%BE%A1"},"children":["入力サイズの制御"]},"を参照してください。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["output_mode"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["string"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["overwrite"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["いいえ"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["overwrite"]}," または ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["append"]},"。",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"#%E5%87%BA%E5%8A%9B%E3%83%A2%E3%83%BC%E3%83%89"},"children":["出力モード"]},"を参照してください。"]}]}]}]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["${...}"]}," のようなワークフロー変数は、テキストとして渡されます。どちらのソリューションも、数値とブール値を文字列として受け付けるため、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["\"4\""]}," や ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["\"false\""]}," も使用できます。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":4,"id":"出力-1","__idx":12},"children":["出力"]},{"$$mdtype":"Tag","name":"div","attributes":{"className":"md-table-wrapper"},"children":[{"$$mdtype":"Tag","name":"table","attributes":{"className":"md"},"children":[{"$$mdtype":"Tag","name":"thead","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"th","attributes":{"width":"15%","data-label":"フィールド"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["フィールド"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"10%","data-label":"型"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["型"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"75%","data-label":"説明"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["説明"]}," "]}]}]},{"$$mdtype":"Tag","name":"tbody","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["channels"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["varchar"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["チャネルのラベル。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["variable"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["varchar"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["その行を生成したモデル。例: ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["attribution_shapley_size2_conv_rate_order_algorithmic"]},"。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["shapley_order"]}," が ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["false"]}," の場合、名前に ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["_order"]}," は含まれません。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["value"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["double"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["このサイズにおける、チャネルのShapleyアトリビューション値。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["size"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["integer"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ルックバックサイズ。0から ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["shapley_lookback_window"]}," まで。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["0"]}," はジャーニー全体を表します。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["N"]}," は、各ジャーニーの直近N個のユニークなチャネルのみをスコアリングすることを意味します。"]}]}]}]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["テーブルはロング形式で、チャネルとサイズの組み合わせごとに1行です。チャネルを比較する前に、必ず ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["size"]}," で絞り込むかグループ化してください。値をパーセンテージに変換するには、",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"#shapley%E3%81%AE%E5%87%BA%E5%8A%9B%E3%82%92%E3%83%91%E3%83%BC%E3%82%BB%E3%83%B3%E3%83%86%E3%83%BC%E3%82%B8%E3%81%AB%E5%A4%89%E6%8F%9B%E3%81%99%E3%82%8B"},"children":["Shapleyの出力をパーセンテージに変換する"]},"を参照してください。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":2,"id":"モデルの選び方","__idx":13},"children":["モデルの選び方"]},{"$$mdtype":"Tag","name":"div","attributes":{"className":"md-table-wrapper"},"children":[{"$$mdtype":"Tag","name":"table","attributes":{"className":"md"},"children":[{"$$mdtype":"Tag","name":"thead","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"th","attributes":{"width":"22%","data-label":"観点"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["観点"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"39%","data-label":"Markov( mta_markov )"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["Markov(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_markov"]},")"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"39%","data-label":"Shapley( mta_shapley )"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["Shapley(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_shapley"]},")"]}," "]}]}]},{"$$mdtype":"Tag","name":"tbody","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["入力"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["ジャーニー"]}," ごとに1行。経路は ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["' > '"]}," で連結"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["接触"]}," ごとに1行"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["貢献度を決める要素"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["チャネル間の遷移と除去効果"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["チャネルの組み合わせごとのコンバージョン率"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["コンバージョンしていないジャーニーを使用するか"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["いいえ。結果に影響しません。"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["はい。コンバージョン率の計算に必要です。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["順序の影響を受けるか"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["常に受ける"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["設定可能(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["shapley_order"]},")"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["直近の接触の分析"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["なし(経路全体)"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["あり(サイズ1〜N = 直近Nチャネル)"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["追加の出力"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["遷移行列(オプション)"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ルックバックサイズごとの結果セット"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["主なコスト要因"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ジャーニー数とチャネル数"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ジャーニーあたりのユニークなチャネル数 × ルックバックウィンドウ"]}]}]}]}]},{"$$mdtype":"Tag","name":"div","attributes":{"className":"md-table-wrapper"},"children":[{"$$mdtype":"Tag","name":"table","attributes":{"className":"md"},"children":[{"$$mdtype":"Tag","name":"thead","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"th","attributes":{"width":"38%","data-label":"知りたいこと"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["知りたいこと"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"22%","data-label":"使用するモデル"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["使用するモデル"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"40%","data-label":"理由"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["理由"]}," "]}]}]},{"$$mdtype":"Tag","name":"tbody","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["どのチャネルが最も代替しにくいか。チャネルXを停止したらどうなるか。"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_markov"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["除去効果は、チャネルのグラフに対する直接的なWhat-if分析になります。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["顧客はチャネル間をどのように移動しているか。"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["transition_output_table"]}," を指定した ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_markov"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["遷移行列を出力するのはMarkovだけです。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["一緒に現れることが多いチャネル間で、予算をどう配分すべきか。"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_shapley"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["Shapleyは、一緒に現れるチャネル間で貢献度を公平に分配します。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["コンバージョン直前に最も重要なチャネルはどれか。"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_shapley"]},"(サイズ1〜3)"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["小さいサイズでは、直近の数チャネルのみを対象にします。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":["経営層にはどのような結果を示すべきか。"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["両方、およびルールベースのモデル"]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["複数のモデルを比較することで、一致する点と異なる点がわかります。"]}]}]}]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["推奨:"]}," ",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"#%E3%82%AF%E3%82%A4%E3%83%83%E3%82%AF%E3%82%B9%E3%82%BF%E3%83%BC%E3%83%88"},"children":["クイックスタート"]},"のように、両方のモデルを並行して実行してください。これらは独立した実行であり、状態を共有しません。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":2,"id":"関連トピック","__idx":14},"children":["関連トピック"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"markovの結果の解釈","__idx":15},"children":["Markovの結果の解釈"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["Markovは、コンバージョンしたジャーニーのみから遷移を推定します。コンバージョンしなかったジャーニーは、入力に含めても結果に影響しません。このため、次の点に注意してください。"]},{"$$mdtype":"Tag","name":"ul","attributes":{},"children":[{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["removal_effect"]}," は、チャネルがコンバージョンに至る経路上にどれだけ位置しているかを示します。おおまかには、それらの経路をたどる際にそのチャネルを通過する確率です。除去効果の高いチャネルは代替しにくいチャネルです。1つのジャーニーが複数のチャネルを通過しうるため、除去効果の合計は1にはなりません。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["attribution_markov_algorithmic"]}," は、除去効果を合計がコンバージョン総数になるようにスケールしたものです。「チャネルXには約N件のコンバージョンが割り当てられている」と説明する際に使用します。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":["Markovは、あるチャネルを通るジャーニーがどのくらいの頻度で失敗するかは示しません。チャネルがコンバージョン率に与える影響を確認するには、Shapleyを使用してください。"]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["遷移レポートやSankey図を作成する際は、除去効果が非常に小さいチャネル(たとえば ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["0.005"]}," 未満)を除外すると、見やすくなります。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"shapleyのルックバックウィンドウと順序","__idx":16},"children":["Shapleyのルックバックウィンドウと順序"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["shapley_lookback_window"]}," は、Shapleyが実行するモデルの数を制御します。デフォルトの ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["4"]}," では、Shapleyはサイズ0、1、2、3、4の5つのモデルを実行します。"]},{"$$mdtype":"Tag","name":"ul","attributes":{},"children":[{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["サイズ0"]}," はジャーニー全体をスコアリングします。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["サイズN"]}," は、各ジャーニーの ",{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["直近N個のユニークなチャネル"]}," のみを対象にします。単位は日数ではなくチャネル数です。"]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["ほとんどのジャーニーのユニークなチャネル数が3未満の場合や、実行に時間がかかりすぎる場合は、ウィンドウを2または3に下げてください。1段階下げるごとに、モデルが1つ減ります。ジャーニーに4つを超えるユニークなチャネルが含まれることが多く、より長いウィンドウが実際のビジネス上の問いに答える場合にのみ、(最大10まで)引き上げてください。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["次の場合は、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["shapley_order"]}," を ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["false"]}," に設定します。"]},{"$$mdtype":"Tag","name":"ul","attributes":{},"children":[{"$$mdtype":"Tag","name":"li","attributes":{},"children":["チャネル数が多くデータが少ないため、順序付きの提携のデータが疎になる場合"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":["多くの接触が同じタイムスタンプを持つなど、接触の順序が信頼できない場合"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":["順序に関係なく、チャネルの集合単位で貢献度を見たい場合"]}]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["順序を考慮しない提携は、より多くのジャーニーをまとめるため、推定値がより安定します。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"shapleyの出力をパーセンテージに変換する","__idx":17},"children":["Shapleyの出力をパーセンテージに変換する"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["サイズごとに各チャネルの貢献度の割合を求めるには、その値を同じサイズの合計で割ります。履歴を残すには、実行時刻を追加します。"]},{"$$mdtype":"Tag","name":"CodeBlock","attributes":{"data-language":"sql","header":{"controls":{"copy":{}}},"source":"SELECT\n  ${session_unixtime}                          AS time,\n  '${session_local_time}'                      AS run_time,\n  size,\n  channels,\n  ROUND(value, 4)                              AS value,\n  value / SUM(value) OVER (PARTITION BY size)  AS percentage\nFROM my_mta_db.mta_shapley_attribution\n","lang":"sql"},"children":[]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"入力サイズの制御","__idx":18},"children":["入力サイズの制御"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["デフォルト(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["max_load_rows = 0"]},")では、すべての行が読み込まれます。デフォルトのままにして、代わりに準備用SQLで、たとえば直近180日間のようなルックバック期間を設定したり、まれなチャネルを ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["other"]}," にまとめたりして、データ量を制限することをおすすめします。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["max_load_rows"]}," を安全のための上限として設定する場合は、実行後に実行統計を確認してください。実行結果には ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["input_total_records"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["loaded_rows"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["output_rows"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["truncated"]}," が報告されます。",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["truncated"]}," が ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["true"]}," の場合、上限によって入力の一部が切り捨てられています。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["2つのモデルでは、上限の扱いが異なります。"]},{"$$mdtype":"Tag","name":"ul","attributes":{},"children":[{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["Markov"]}," は、コンバージョンしたジャーニーを先に読み込みます。コンバージョンしなかったジャーニーはMarkovに影響しないため、コンバージョンしたすべてのジャーニーが上限内に収まる限り、結果は上限なしの実行と同じになります。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["Shapley"]}," もコンバージョンしたジャーニーを先に読み込み、常にジャーニー全体を読み込みます。コンバージョンしなかったジャーニーの読み込みが減るため、コンバージョン率が上がり、アトリビューションに偏りが生じます。Shapleyでは、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["max_load_rows"]}," ではなく、準備用SQLで意図的なサンプリング方法を採用してください。"]}]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"出力モード","__idx":19},"children":["出力モード"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["MTAは実行のたびに入力全体からアトリビューションを再計算するため、デフォルトの ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["output_mode: overwrite"]}," のままにしてください。出力テーブルには実行IDの列がないため、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["append"]}," を使うと異なる実行の結果が混ざってしまいます。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["履歴を残すには、",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"#shapley%E3%81%AE%E5%87%BA%E5%8A%9B%E3%82%92%E3%83%91%E3%83%BC%E3%82%BB%E3%83%B3%E3%83%86%E3%83%BC%E3%82%B8%E3%81%AB%E5%A4%89%E6%8F%9B%E3%81%99%E3%82%8B"},"children":["Shapleyの出力をパーセンテージに変換する"]},"のように、実行ごとの出力を別のテーブルにコピーし、実行時刻を追加します。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"チャネルラベルの管理","__idx":20},"children":["チャネルラベルの管理"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["チャネルラベルはレポートに表示され、Markovの遷移行列では列名になります。結果を見やすく保つために、次の点に注意してください。"]},{"$$mdtype":"Tag","name":"ul","attributes":{},"children":[{"$$mdtype":"Tag","name":"li","attributes":{},"children":["句読点を含まない、短い小文字のラベルを使用します。たとえば、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["Paid Search (Google)"]}," ではなく ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["paid_search"]}," とします。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":["正規化後に、2つのラベルが同じ列名にならないようにしてください。たとえば、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["Paid-Search"]}," と ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["paid_search"]}," は衝突する可能性があります。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":["まれな流入元を ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["other"]}," にまとめ、ユニークなチャネル数を15程度以下に保ちます。これにより、Shapleyの提携に十分なデータが集まり、遷移行列も読みやすくなります。"]},{"$$mdtype":"Tag","name":"li","attributes":{},"children":["NULLのチャネルは、両方の入力で同じ方法で扱ってください。除外するか、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["direct"]}," などのラベルに置き換えます。"]}]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":2,"id":"faq","__idx":21},"children":["FAQ"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"はじめに","__idx":22},"children":["はじめに"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["どのようなデータが必要ですか?"]}," ユーザーID、タイムスタンプ、チャネルラベル、コンバージョンフラグを含むタッチポイントデータが必要です。Webイベント、メールのエンゲージメント、広告クリック、注文を1つの接触テーブルにまとめられます。Markovでは、そのデータを先にジャーニー単位にまとめる必要があります。Shapleyはそのまま読み込めます。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["両方のモデルを実行する必要がありますか?"]}," いいえ。各ソリューションは単独で実行できます。ただし、2つのモデルは異なる問いに答え、比較することで結果への信頼性が高まるため、両方の実行をおすすめします。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["ソリューションが有効になっていないために実行が拒否されました。どうすればよいですか?"]}," MTAはお客様のアカウントで有効化されている必要があります。Treasure AIのアカウントチームにお問い合わせください。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"データと品質","__idx":23},"children":["データと品質"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["コンバージョンフラグにはどのような値を指定できますか?"]}," ML Batch APIは、ブール値、任意の数値(0以外はコンバージョンとみなされます)、またはテキスト(",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["true"]},"/",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["false"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["t"]},"/",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["f"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["yes"]},"/",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["no"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["y"]},"/",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["n"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["1"]},"/",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["0"]},")を受け付けます。テキストは大文字と小文字を区別せず、前後のスペースは削除されます。NULLはコンバージョンなしとして扱われます。それ以外の値を指定すると実行は失敗し、エラーメッセージに読み取れなかった値の一覧が表示されます。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":["クイックスタートのワークフローには、より厳しい要件があります。Markovの準備用SQLが ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["SUM(converted)"]}," を使用するため、元の接触テーブルの ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["converted"]}," は整数の ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["0"]}," または ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["1"]}," として保存する必要があります。",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"#%E3%83%87%E3%83%BC%E3%82%BF%E3%81%AE%E6%BA%96%E5%82%99"},"children":["データの準備"]},"に従い、このステップの前にテキストのフラグを変換してください。Shapleyは、このSQLステップを介さずに接触単位の入力を直接読み込みます。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["「no conversions」で実行が失敗したのはなぜですか?"]}," 入力には、少なくとも1件のコンバージョンが含まれている必要があります。コンバージョンのない実行は、以前の結果が空のテーブルで上書きされないよう、意図的に失敗します。コンバージョンフラグの列と、準備用SQLの日付範囲を確認してください。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["テーブルにデータがあるのにMarkovの実行が失敗したのはなぜですか?"]}," Markovには、チャネルを ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["' > '"]}," で連結した、1ジャーニーにつき1行のテーブルが必要です。どの行にも区切り文字が含まれていない場合、入力は接触単位のデータとみなされ、実行は拒否されます。",{"$$mdtype":"Tag","name":"MarkdownLink","attributes":{"href":"#%E3%83%87%E3%83%BC%E3%82%BF%E3%81%AE%E6%BA%96%E5%82%99"},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["mta_markov"]}," のデータの準備"]},"に示すとおり、先にジャーニーテーブルを作成してください。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["Shapleyの結果がおかしく、タイムスタンプが数値になっています。なぜですか?"]}," Unixエポックの整数は誤った日付として読み込まれ、接触の順序が崩れます。代わりに、timestamp型、または ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["YYYY-MM-DD HH:MM:SS"]}," 形式のテキストを渡してください。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["Markovで、あるチャネルの貢献度が0になるのはなぜですか?"]}," そのチャネルがコンバージョンしなかったジャーニーにのみ現れるためです。Markovはコンバージョンしたジャーニーのみを使用するため、そのチャネルには貢献度が割り当てられません。これは想定された動作であり、データ品質の問題ではありません。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"モデルの選び方-1","__idx":24},"children":["モデルの選び方"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["MarkovとShapleyで数値が異なります。どちらが正しいのですか?"]}," どちらも妥当です。測定しているものが異なります。Markovは、コンバージョンに至る経路がどのチャネルに依存しているかを示します。Shapleyは、コンバージョン率に対する各チャネルの貢献を示します。両者が一致する点に注目してください。一致しない点は、ルールベースのモデルと比較して調査してください。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["どのShapleyサイズを報告すればよいですか?"]}," 全体像にはサイズ0(ジャーニー全体)を使用します。コンバージョンに最も近い接触を重視する場合は、サイズ1〜3を使用します。各サイズは別々のモデルであるため、サイズをまたいで値を合計しないでください。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":3,"id":"スケールと制限","__idx":25},"children":["スケールと制限"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["MTAはどのくらいのデータ量を処理できますか?"]}," アトリビューションはすべてのジャーニーに依存し、分割できないため、各ソリューションは単一ノードで実行されます。MTAは、最大で約2,000万行のモデル入力を想定して設計されています。それより大きな入力でもテストで完了した実行はありますが、入力サイズに応じて実行時間とメモリ使用量が増加します。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["Shapleyが遅くなるのはなぜですか?"]}," チャネルの組み合わせの数は、ジャーニーあたりのユニークなチャネル数に応じて急速に増加し、さらにShapleyはルックバックサイズごとに1つのモデルを実行します。高速化するには、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["shapley_lookback_window"]}," を下げる、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["shapley_order"]}," を ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["false"]}," に設定する、またはまれなチャネルを ",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["other"]}," にまとめてください。"]},{"$$mdtype":"Tag","name":"p","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["MTAはリアルタイムですか?"]}," いいえ。MTAはバッチソリューションです。ワークフローをスケジュール(たとえば週次や月次)して、アトリビューションを更新してください。"]},{"$$mdtype":"Tag","name":"Heading","attributes":{"level":2,"id":"用語集","__idx":26},"children":["用語集"]},{"$$mdtype":"Tag","name":"div","attributes":{"className":"md-table-wrapper"},"children":[{"$$mdtype":"Tag","name":"table","attributes":{"className":"md"},"children":[{"$$mdtype":"Tag","name":"thead","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"th","attributes":{"width":"22%","data-label":"用語"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["用語"]}," "]},{"$$mdtype":"Tag","name":"th","attributes":{"width":"78%","data-label":"定義"},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["定義"]}," "]}]}]},{"$$mdtype":"Tag","name":"tbody","attributes":{},"children":[{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["タッチポイント(接触)"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["メールのクリックやサイト訪問など、顧客とチャネルとの1回の接点。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["ジャーニー"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["1人の顧客の接触を順に並べたもの。コンバージョンまで(コンバージョンを含む)、またはコンバージョンしなかった場合はデータの終わりまで。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["コンバージョンしたジャーニー"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["コンバージョンで終わるジャーニー。コンバージョンのたびに新しいジャーニーが始まります。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["チャネル"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"code","attributes":{},"children":["email"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["paid_search"]},"、",{"$$mdtype":"Tag","name":"code","attributes":{},"children":["organic"]}," など、接触のマーケティング上の流入元。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["Markovチェーン"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["次のステップが現在のステップのみに依存するモデル。MTAでは、チャネル間の移動をモデル化するために使用します。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["遷移行列"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["顧客があるチャネルから別のチャネル、またはコンバージョンへ移動する確率の表。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["除去効果(Removal effect)"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["コンバージョンに至る経路がチャネルにどれだけ依存しているか。0から1の値。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["提携(Coalition)"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ジャーニー内で一緒に現れるチャネルの集合(または順序付きの並び)。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["Shapley値"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["そのチャネルが加わりうるすべての提携にわたる、コンバージョン率への平均的な貢献。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["ルックバックサイズ"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ジャーニーの終わりから遡ってShapleyがスコアリングするユニークなチャネルの数。サイズ0はジャーニー全体を意味します。"]}]},{"$$mdtype":"Tag","name":"tr","attributes":{},"children":[{"$$mdtype":"Tag","name":"td","attributes":{},"children":[{"$$mdtype":"Tag","name":"strong","attributes":{},"children":["ルールベースのアトリビューション"]}]},{"$$mdtype":"Tag","name":"td","attributes":{},"children":["ファーストタッチ、ラストタッチ、リニア、U字型など、位置に応じて貢献度を割り当てるモデル。"]}]}]}]}]}]},"headings":[{"value":"マルチタッチアトリビューション(MTA)","id":"マルチタッチアトリビューションmta","depth":1},{"value":"概要とユースケース","id":"概要とユースケース","depth":2},{"value":"他のAI Signalsとの使い分け","id":"他のai-signalsとの使い分け","depth":3},{"value":"クイックスタート","id":"クイックスタート","depth":2},{"value":"モデルの設定","id":"モデルの設定","depth":2},{"value":"mta_markov","id":"mta_markov","depth":3},{"value":"データの準備","id":"データの準備","depth":4},{"value":"ワークフローパラメータ","id":"ワークフローパラメータ","depth":4},{"value":"出力","id":"出力","depth":4},{"value":"mta_shapley","id":"mta_shapley","depth":3},{"value":"データの準備","id":"データの準備-1","depth":4},{"value":"ワークフローパラメータ","id":"ワークフローパラメータ-1","depth":4},{"value":"出力","id":"出力-1","depth":4},{"value":"モデルの選び方","id":"モデルの選び方","depth":2},{"value":"関連トピック","id":"関連トピック","depth":2},{"value":"Markovの結果の解釈","id":"markovの結果の解釈","depth":3},{"value":"Shapleyのルックバックウィンドウと順序","id":"shapleyのルックバックウィンドウと順序","depth":3},{"value":"Shapleyの出力をパーセンテージに変換する","id":"shapleyの出力をパーセンテージに変換する","depth":3},{"value":"入力サイズの制御","id":"入力サイズの制御","depth":3},{"value":"出力モード","id":"出力モード","depth":3},{"value":"チャネルラベルの管理","id":"チャネルラベルの管理","depth":3},{"value":"FAQ","id":"faq","depth":2},{"value":"はじめに","id":"はじめに","depth":3},{"value":"データと品質","id":"データと品質","depth":3},{"value":"モデルの選び方","id":"モデルの選び方-1","depth":3},{"value":"スケールと制限","id":"スケールと制限","depth":3},{"value":"用語集","id":"用語集","depth":2}],"frontmatter":{"seo":{"title":"マルチタッチアトリビューション(MTA)","description":"Treasure AIのAI Signals「マルチタッチアトリビューション(MTA)」で、ファーストタッチやラストタッチのルールではなく、MarkovモデルとShapleyモデルによるデータドリブンな分析で、コンバージョンに貢献しているマーケティングチャネルを特定します。"}},"lastModified":"2026-11-18T15:12:31.000Z","pagePropGetterError":{"message":"","name":""}},"slug":"/ja/products/customer-data-platform/machine-learning/ai-signals/mta-ai-signals","userData":{"isAuthenticated":false,"teams":["anonymous"]},"isPublic":true}