Reconciling Marketing Data Against the System of Record

Challenge

A multi-market client needed to know which sources produced revenue, not which ones produced leads. Every platform in a stack reports its own version of success, each counts a different event, and none of them knows whether the job was ever sold.

Solution

We stopped reporting from the marketing platforms and started reporting from the client's CRM, then joined the marketing data to it. That meant working the CRM's API directly, joining it against the call tracking platform, and following one funnel end to end, inquiry to appointment to sold job to revenue, split by source and by market, across 2,511 inquiries.

The API had its own personality, as they do. A silent 5,000-row cap that returns success rather than an error. Sort parameters that fail on some entities and not others. Deep expansions that hang. Every one of those is written into the workflow document, so the next person does not rediscover them, and so a number that looks wrong can be checked against a known quirk instead of being explained away.

The same discipline drives the smaller decisions. When we set the click lookback window on that client's offline conversion imports, we measured the actual lag from click to sold job, found a 95th percentile of 77 days, and set the window at 90. The platform default would have discarded most of their revenue as out of range.

One more thing, because it is the part clients do not usually get to see. Our first read of that dataset attributed part of a 65% drop in paid inquiries to source tagging. That was wrong, and the real cause was an ad-schedule outage we had separately identified. We corrected it in writing to the client rather than letting the softer version stand. A measurement practice that never publishes a correction is not being careful. It is not checking.

Result

Paid search produced 417 inquiries, 17% of the total, which converted to appointments at 75% against a 49% all-source average, became 59 sold jobs, and produced $445,354 of the $1.85M sold in the window. Twenty-four percent of the revenue from seventeen percent of the leads. In the same analysis, the single largest lead source in the account turned out to be the worst converter in it, 742 inquiries at a 4% inquiry-to-job rate, which is invisible from the top of the funnel and expensive to keep buying.