How data integrity works
Analytics does not observe your business. It observes whatever a set of tags managed to send from a browser, subject to consent, network conditions, ad blockers and whatever a developer changed last Tuesday. Data integrity is the distance between those recordings and reality, and it rests on four qualities: completeness, accuracy, consistency and timeliness.
Completeness asks whether every real event was captured. Accuracy asks whether the values attached to it — revenue, currency, form type, page — are correct. Consistency asks whether the same action is named and counted the same way across platforms and across months. Timeliness asks whether the data arrived soon enough to act on. A dataset can fail any one of these while looking perfectly healthy on a chart, which is the whole problem.
Why data integrity matters
Every decision downstream inherits the fault. Budget moves to the channel that appears to convert, bidding algorithms train on the conversions they are fed, and a report defends work that may or may not have happened. Corrupted data does not produce obviously wrong decisions; it produces confident wrong decisions, which are far more expensive.
Automated bidding raises the stakes. When a platform is optimising towards a conversion signal, a duplicated or mistagged event teaches it to buy the wrong traffic and it does so at speed. Fixing the tag afterwards does not undo the learning or refund the spend, so integrity is cheaper to protect than to repair.
Where data integrity goes wrong
Most damage comes from ordinary events rather than exotic ones. A website release removes a thank-you page, and conversions stop. A theme update strips the container, and a whole site goes dark. A form is duplicated for a campaign and fires the event twice. Staff, developers and the agency browse the site all day and inflate sessions because internal traffic was never filtered. Bots crawl a landing page. Test orders sit in the revenue total.
Then there are the quieter mismatches: a property left in the wrong time zone, so days are cut in the wrong place; a currency set to the platform default rather than the one the customer paid in; payment gateways and booking systems on another domain breaking the session; and definitions drifting so that this year’s leads include newsletter sign-ups that last year’s did not. None of these announce themselves. They simply make the chart wrong.
How to act on it
Decide first what a correct number would look like. Take a period, count the enquiries that genuinely reached the inbox, the phone and the messaging apps, and compare that with what analytics reports. A gap in either direction is the finding; matching totals are the only real proof that tracking works.
After that, make integrity a routine rather than a rescue. Re-test the critical events after every website release, filter internal traffic, exclude test transactions, keep one written definition of each conversion, and put a note on the chart whenever tracking changes. A periodic tracking audit catches what routine checks miss, and the wider analytics and tracking setup is where the definitions should be documented once and reused.