Analytics and attribution systems contain inaccuracies or gaps.
Broken tracking means every decision downstream is built on corrupted data. Unreliable attribution, misfiring events, and inconsistencies across tools make real optimization structurally impossible.
Hopkins said advertising reached "the status of a science" because of measurement: "keyed advertising, by traced returns, largely by the use of coupons." Without trustworthy tracking you're back to guessing. Ogilvy: "Advertising people who ignore research are as dangerous as generals who ignore decodes of enemy signals."
• Missing conversions
• Inconsistent attribution
Owned — Forms not firing events.
Earned — Referral sources misattributed.
Paid — Conversion tracking unreliable.
Analytics — Data inconsistencies across tools.
AI — AI trained on flawed data.
• Audit tracking quarterly
An overall number can point the opposite way from every segment it is built from whenever those segments were sampled at different rates. This happens to perfectly clean data, because an aggregate is a weighted average and the weights can decide the result. Thomas Crook, Brian Frasca, Ron Kohavi and Roger Longbotham describe it in "Seven Pitfalls to Avoid when Running Controlled Experiments on the Web" (KDD 2009). In their example a site ramps a test from a 1% split on Friday to 50% on Saturday. The treatment wins on Friday and wins again on Saturday, yet loses when the days are combined, because Saturday was a weaker day overall and the treatment was far more exposed to it. The same trap appears across geographies sampled at different rates. The best-known illustration is the 1975 Berkeley admissions study by Bickel, Hammel and O'Connell in Science, where an aggregate gap between men and women reversed direction once applications were broken out by department.
What it means for this rule: Reliable tracking is necessary but not sufficient: a correctly measured aggregate over a changing mix can be a true number that means the opposite of what it seems to say. Marketing mixes shift constantly through budget ramps, geographic expansion, new audiences and platform delivery changes. Before quoting a blended figure, check what share each segment held in each period; if the shares moved, report the segments or hold the weights constant. A tracking audit should cover the weighting as well as the wiring.
Sources: Crook, Frasca, Kohavi & Longbotham (2009), Seven Pitfalls to Avoid when Running Controlled Experiments on the Web · Bickel, Hammel & O'Connell (1975), Sex Bias in Graduate Admissions, Science
