A marketing lead opens three reports. Paid search claims four hundred conversions, email claims three hundred and the chat tool claims two hundred and eighty.
The business had six hundred conversions that month. Everybody is reporting correctly under their own rules, and the sum of the truth is larger than what happened.
The chat numbers get questioned first, because they are newest. Within two quarters nobody quotes them at all, and a tool that may genuinely be working has lost its case on a reporting technicality.
1. Classify chat correctly from the start
It is an interaction inside a journey, not a source of traffic. Almost nobody arrives at a website because of chat, and treating it as a channel guarantees double counting.
2. Decide the attribution window deliberately
Same session, twenty four hours or seven days. Longer windows credit chat for conversions it had little to do with, and the default is usually generous.
3. Decide whether an engagement means an exchange
A widget opened and closed is not an engagement. Requiring at least one meaningful exchange removes a large amount of noise from the numbers.
4. Separate assisted from originated conversions
An assisted conversion is honest and useful. Reporting assists as though they were originations is the single most common way chat reporting loses credibility.
5. Report leads captured separately
A lead that only exists because somebody engaged is a genuinely new object, and it is the cleanest attribution available. Count those on their own.
6. Isolate the after hours conversions
Where nobody was available, the alternative was no response. These are the most defensible incremental conversions in the whole dataset.
7. Reconcile totals across channels monthly
Somebody should check that the sum of claimed conversions resembles the actual number. This exercise usually finds the problem within an hour.
8. Agree who owns the definition
One person, one document, one set of rules that every report follows. Without an owner, each tool reports under its own defaults and the argument never ends.
9. Tag the source into the lead record
Which page, which campaign, which hour. A lead arriving in a sales system with its origin attached is worth far more than one arriving anonymous.
10. Track through to revenue rather than to a form
An enquiry is not a customer. The useful measure is what proportion of chat originated leads become revenue, which frequently differs from other channels.
11. Watch for cannibalisation
A visitor who would have used the enquiry form and used chat instead has not created anything new. Comparing total conversions rather than channel totals catches this.
12. Keep the measurement honest about effect
The counterfactual, not the total, is the question, which is the ground in conversion rate questions marketers ask about chat.
13. Track the intermediate step that matters
Question asked to booked call is a cleaner measure than any conversion rate, and it is what question to booked call describes.
14. Set the expectation before the first report
Agreeing what will be claimed, and what will not, is what keeps a channel credible, which is the same discipline as media coverage expectation questions in first conversations. The underlying friction it is meant to remove is described in sizing questions are killing your conversion rate. See features, start at try, and the sector view sits under marketing agencies.