ANALYTICS AND ATTRIBUTION AUDITThe tracking was never broken.
Client:
An independent nonprofit publisher and media cooperative with a subscription base, a books list, and an ethical mail-order business
Context:
An independent nonprofit publisher was losing attribution on three quarters of its revenue. The problem was not the tracking. It was that the organization was measuring a retention business as though it were an acquisition business.
-
The organization could see its revenue. It could not see where the revenue came from.
Roughly three quarters of it was arriving in analytics with no source attached at all. For a publisher running paid campaigns, an email programme, and an active social presence, that is close to unusable: every marketing decision was being made against a quarter of the picture, and the working assumption inside the organization was that something in the tracking had failed.
The stakes were not academic. The organization had recently run a public fundraising campaign it described in its own words as make or break. Marketing spend that could not be justified was spend that would be cut.
-
Establish what the data can and cannot support before analyzing it. The first phase was not analysis. It was working out what each system was actually recording, where the same order could be counted twice, and which questions the current setup was structurally incapable of answering.
Reconcile the analytics against the commerce records. Analytics platforms and commerce platforms disagree for reasons that are usually mundane and occasionally load-bearing. Reading them against each other order by order is what separated the real defects from the imagined one.
Trace every gap to a cause, not a symptom. An unattributed order and a misattributed order are different failures with different fixes, and treating them as one problem is how organizations end up rebuilding tracking that was working.
Report against decisions. The deliverable was structured so that leadership could tell which findings changed what they should do, and which were true but inert.
-
The missing attribution was not a defect. Every unattributed order was an automatic subscription renewal. A renewal has no acquisition source to record, because nobody acquired it: the subscription was sold once and has been renewing ever since. The tracking had been behaving correctly the entire time. What looked like a broken system was a correctly functioning system being asked a question it could not answer.
Email was being undercounted by about 2.7 times. The commerce platform credited the email programme with roughly two and a half times the orders and revenue that analytics did. The cause was a single misconfigured parameter carrying the wrong value, and most of the orders it was hiding were early renewals, which made the email programme look far less effective at retention than it actually was.
Self-referral was the largest genuine attribution fault. The organization's own domain was overwriting the true source on nearly half of all attributed new orders, which means the channels that were genuinely working were being credited to the site itself.
No custom dimensions were registered against roughly a dozen firing custom events. The events were collecting data that was, in practice, unreadable.
Fewer than six percent of orders came from new customers. This was the finding underneath all the others. The organization was structurally a retention business, and every measurement decision, report, and campaign target had been built as though it were an acquisition business. New subscribers also behaved differently from the existing base in ways the reporting could not see, skewing more digital, more international, and far less likely to pay by direct debit.
Login friction was surfacing as a top-three page. The password reset page was among the most-viewed pages on the site, which is a retention problem wearing a technical costume.
-
A phased audit covering the tag container, event configuration, attribution model, and revenue reconciliation
An executive report written for leadership rather than for an analytics team, presented live
A prioritized remediation plan itemized so that the organization could see exactly which items its own team could do, which either party could do, and which needed me
Fixed pricing on every item, in three scoped options, so that the decision was about depth rather than about whether
-
Leadership went into the engagement believing it had a tracking failure and a marketing attribution problem. It came out with a documented account of what its data could and could not support, the four defects that were real, the one that was not, and a costed order of operations for fixing them.
The most valuable finding was the one that removed work rather than adding it. Rebuilding tracking that was already functioning correctly would have consumed months and produced nothing.
-
Attribution audits usually get sold as a way to find the channel that is secretly working. This one found something more useful and less comfortable: the organization's entire measurement frame was pointed at the wrong business. A retention business measured for acquisition will always look like it is failing, because it is being scored on a game it is not playing.
That is not a tracking finding. It is a decision finding, and it is the difference between fixing the instrument and fixing what you are asking it.