قضية بنك Raiffeisen: كيفية معالجة الاحتيال في شبكات CPA

جدول المحتويات

• تحد
• المحلول
  • الخطوة 1. اجمع البيانات الأولية
  • الخطوة 2. معالجة البيانات
  • الخطوة 3. بناء التقارير
• نتائج

تحد

المحلول

الخطوة 1. اجمع البيانات الأولية

    SELECT cp.promo.date AS Date, cp.promo.clientId AS ClientId, cp.promo.time AS Promo_time, cp.cpc.time AS CPC_time, send.time AS SEND_time FROM ( SELECT promo.date, promo.clientId, promo.time, cpc.time FROM ( SELECT date, clientId, MIN(time) AS time FROM TABLE_DATE_RANGE({dataSetName}.{tableName},TIMESTAMP('{startDate}'), TIMESTAMP('{endDate}')) WHERE page.pagePath CONTAINS '/promo/' AND type = 'pageview' AND traffic.medium != 'cpc' GROUP BY date, clientId, ORDER BY clientId ASC) AS promo LEFT JOIN ( SELECT date, clientId, traffic.medium, time FROM TABLE_DATE_RANGE({dataSetName}.{tableName},TIMESTAMP('{startDate}'), TIMESTAMP('{endDate}')) WHERE traffic.medium = 'cpc' AND type = 'pageview' GROUP BY date, clientId, traffic.medium, time ORDER BY clientId ASC) AS cpc ON promo.clientId=cpc.clientId WHERE promo.time < cpc.time) AS cp LEFT JOIN ( SELECT date, clientId, time FROM TABLE_DATE_RANGE({dataSetName}.{tableName},TIMESTAMP('{startDate}'), TIMESTAMP('{endDate}')) WHERE eventInfo.eventCategory = 'send_ok' AND type = 'event' GROUP BY date, clientId, time ORDER BY clientId ASC) AS send ON cp.promo.clientId = send.clientId WHERE cp.cpc.time < send.time
SELECT cp.promo.date AS Date, cp.promo.clientId AS ClientId, cp.promo.time AS Promo_time, cp.cpc.time AS CPC_time, send.time AS SEND_time FROM ( SELECT promo.date, promo.clientId, promo.time, cpc.time FROM ( SELECT date, clientId, MIN(time) AS time FROM TABLE_DATE_RANGE({dataSetName}.{tableName},TIMESTAMP('{startDate}'), TIMESTAMP('{endDate}')) WHERE page.pagePath CONTAINS '/promo/' AND type = 'pageview' AND traffic.medium != 'cpc' GROUP BY date, clientId, ORDER BY clientId ASC) AS promo LEFT JOIN ( SELECT date, clientId, traffic.medium, time FROM TABLE_DATE_RANGE({dataSetName}.{tableName},TIMESTAMP('{startDate}'), TIMESTAMP('{endDate}')) WHERE traffic.medium = 'cpc' AND type = 'pageview' GROUP BY date, clientId, traffic.medium, time ORDER BY clientId ASC) AS cpc ON promo.clientId=cpc.clientId WHERE promo.time < cpc.time) AS cp LEFT JOIN ( SELECT date, clientId, time FROM TABLE_DATE_RANGE({dataSetName}.{tableName},TIMESTAMP('{startDate}'), TIMESTAMP('{endDate}')) WHERE eventInfo.eventCategory = 'send_ok' AND type = 'event' GROUP BY date, clientId, time ORDER BY clientId ASC) AS send ON cp.promo.clientId = send.clientId WHERE cp.cpc.time < send.time

الخطوة 2. معالجة البيانات

تم إنشاء قائمة القيم الضرورية لتحديد قيم مصدر الزيارات التي تم تغييرها مع الشركة التابعة:

• معرف المستخدم.
• مصدر ووسيط الجلسات الأولى والتالية.
• الوقت بين الجلسات.
• عناوين URL الأولى والنهائية لكل جلسة.
• الأحداث في كل جلسة.
• حدث الصفقة في الجلسة النهائية.

بعد ذلك ، للتأكد من إعادة كتابة قيم مصدر حركة المرور بالتأكيد ، اختار محللو OWOX BI تصفية البيانات مع مراعاة هذه الشروط:

• يجب أن تصل الفترة الزمنية بين الجلستين إلى 60 ثانية.
• يجب أن تظل صفحة موقع الويب كما هي عند تغيير المصدر.
• يجب أن تكون هناك صفقة في الجلسة النهائية.
• يجب أن يكون وسيط المرور في الجلسة الأخيرة تابعًا.

إليك استعلام SQL الذي تم استخدامه للحصول على البيانات الضرورية:

    SELECT * FROM ( SELECT traff.clientId clientId, traff.page.pagePath pagePath, traff.traffic.source startSource, traff.traffic.medium startMedium, traff.time startTime, aff.evCategory eventCategory, aff.evlabel eventLabel, aff.evSource finishSource, aff.evMedium fifnishMedium, aff.time finishTime, aff.isTransaction isTransaction, aff.pagePath link, traff.time - aff.time AS diff FROM ( SELECT clientId, page.pagePath, traffic.source, traffic.medium, date, INTEGER(time) time FROM TABLE_DATE_RANGE({dataSetName}.{tableName},TIMESTAMP('{startDate}'), TIMESTAMP('{endDate}')) WHERE traffic.medium != 'affiliate')AS traff JOIN ( SELECT total.date date, total.time time, total.clientId clientId, total.eventInfo.eventCategory evCategory, total.eventInfo.eventLabel evlabel, total.traffic.source evSource, total.traffic.medium evMedium, tr.eventInfo.eventCategory isTransaction, total.page.pagePath pagePath FROM ( SELECT clientId, page.pagePath, eventInfo.eventCategory, eventInfo.eventLabel, traffic.source, traffic.medium, date, INTEGER(time) time FROM TABLE_DATE_RANGE({dataSetName}.{tableName},TIMESTAMP('{startDate}'), TIMESTAMP('{endDate}')) WHERE traffic.medium = 'affiliate') AS total LEFT JOIN ( SELECT clientId, date, eventInfo.eventCategory, INTEGER(time) time FROM TABLE_DATE_RANGE({dataSetName}.{tableName},TIMESTAMP('{startDate}'), TIMESTAMP('{endDate}')) WHERE eventInfo.eventCategory = 'send_ok' GROUP BY 1, 2, 3, 4) AS tr ON total.clientId = tr.clientId AND total.date = tr.date WHERE tr.eventInfo.eventCategory = 'send_ok' AND tr.time>total.time)AS aff ON traff.clientId = aff.clientId) WHERE diff >-60 AND diff<0 GROUP BY 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13 ORDER BY clientId, finishTime
SELECT * FROM ( SELECT traff.clientId clientId, traff.page.pagePath pagePath, traff.traffic.source startSource, traff.traffic.medium startMedium, traff.time startTime, aff.evCategory eventCategory, aff.evlabel eventLabel, aff.evSource finishSource, aff.evMedium fifnishMedium, aff.time finishTime, aff.isTransaction isTransaction, aff.pagePath link, traff.time - aff.time AS diff FROM ( SELECT clientId, page.pagePath, traffic.source, traffic.medium, date, INTEGER(time) time FROM TABLE_DATE_RANGE({dataSetName}.{tableName},TIMESTAMP('{startDate}'), TIMESTAMP('{endDate}')) WHERE traffic.medium != 'affiliate')AS traff JOIN ( SELECT total.date date, total.time time, total.clientId clientId, total.eventInfo.eventCategory evCategory, total.eventInfo.eventLabel evlabel, total.traffic.source evSource, total.traffic.medium evMedium, tr.eventInfo.eventCategory isTransaction, total.page.pagePath pagePath FROM ( SELECT clientId, page.pagePath, eventInfo.eventCategory, eventInfo.eventLabel, traffic.source, traffic.medium, date, INTEGER(time) time FROM TABLE_DATE_RANGE({dataSetName}.{tableName},TIMESTAMP('{startDate}'), TIMESTAMP('{endDate}')) WHERE traffic.medium = 'affiliate') AS total LEFT JOIN ( SELECT clientId, date, eventInfo.eventCategory, INTEGER(time) time FROM TABLE_DATE_RANGE({dataSetName}.{tableName},TIMESTAMP('{startDate}'), TIMESTAMP('{endDate}')) WHERE eventInfo.eventCategory = 'send_ok' GROUP BY 1, 2, 3, 4) AS tr ON total.clientId = tr.clientId AND total.date = tr.date WHERE tr.eventInfo.eventCategory = 'send_ok' AND tr.time>total.time)AS aff ON traff.clientId = aff.clientId) WHERE diff >-60 AND diff<0 GROUP BY 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13 ORDER BY clientId, finishTime

الخطوة 3. بناء التقارير

نتائج