The Day the Data Lied – Fixing Attribution Before It Killed a Campaign

It was a rainy Tuesday when I got the call.

A senior marketing leader from a SaaS client was convinced their direct mail (DM) campaigns were outperforming digital efforts by a factor of ten. They were ready to triple the DM budget.

Something felt off.

We pulled the numbers — 734 conversions in a year. Sounded amazing. Until we looked under the hood.

The attribution logic relied on a regex that matched just 10 characters of the mailing address. That meant “1100 Market St, Suite 302” and “1100 Market St, Suite 480” were being counted as the same.

The more we audited, the more inflated the data became.
The real number? Just 82.

Correcting it was messy. We had to:

• Push back on third-party vendors
• Rewrite regex logic
• Rethink lookback windows
• Rebuild internal trust in the data

But once we surfaced the truth, the impact was massive. The CMO redirected the budget—away from direct mail and into podcast advertising and a high-visibility billboard in Palo Alto.

Lessons That Day Taught Me:

• Attribution lies when it’s not tested. “Set it and forget it” is a dangerous mindset.
• Regex is powerful—and risky. Always validate match logic with real-world examples.
• Marketing needs a RevOps counterbalance. Someone who says: “Why does this look so good?”
• Saving money is as valuable as earning it. In this case, a regex audit saved them six figures.