Measuring which touch actually wins the deposit

The Challenge

The deposit date was stored as a day with no time, and for thousands of contacts it had been filled by a one-off import — so the property history showed the moment of the import, not the moment of the deposit. Any «what came before» question asked against that field was answering about the wrong day.

The fields the portal already had made it worse: they counted calls regardless of when the call happened, including calls placed after the deposit had landed. The number being reported was several times higher than reality.

The Solution

• Derived the real moment of each deposit from property history, accurate to the second, and used that as the reference point instead of the date field
• Separated the contacts whose history can be trusted from those whose value came from the migration, and scoped every calculation to the former
• Built a dedicated property group with counters for calls, email replies, email opens and logged WhatsApp inside the window before the deposit
• Classified how each deposit was won into three tiers, using call disposition rather than call duration alone — hundreds of unanswered calls run past two minutes of ringing and would otherwise count as conversations
• Wrote the backfill scripts to be idempotent, with the previous value kept in the export so any run can be reversed
• Created every workflow switched off, for review in the interface before anything touched live records
• Exported the text of every note written before a deposit, so the manual comments could be read rather than counted

The Result

Attribution that respects time order, and a documented reason to distrust the previous figures. The reporting now distinguishes deposits the call centre won, deposits it assisted, and deposits that arrived on their own.