REAL BUILDS,
TOLD IN FULL.

Two dashboards gave two different answers to the same question, and nobody could say which one was right. Reporting had been built on top of reporting, one layer patched onto the last, until even the people who built it were not fully sure what a given number meant anymore. Decisions kept getting made anyway, on numbers nobody was willing to defend out loud in the room.

We rebuilt the tracking underneath first. Server-side tracking now runs across all four of the platforms this business advertises on, with browser and server events reconciled against each other so the same visit is never counted twice. Every one of those events lands in a single table, and every chart on the dashboard reads from that same table, so a number never means one thing on one page and something else on another.

The dashboard now runs about 210 charts off a single event table with 332,659 recorded events, and every one of them traces back to the same source. Leadership stopped asking which number to trust and started asking what to do about what the numbers said.

The ad platform’s own reporting said one campaign was mediocre, crediting the sale to whichever channel happened to touch it last. Budget kept drifting away from that campaign every quarter because the report said so, even though the team suspected the platform was not telling the whole story.

We rebuilt attribution across the entire path a customer actually took, not just the last click before a purchase, and set one canonical rule for what counts as a paid lead instead of letting each report re-derive its own definition. The same rebuild caught a separate report that had been counting unpaid, organic visitors as if they were paid conversions.

The campaign the platform had flagged as mediocre had actually paid for itself 3.6 times over once the full path was counted. And the conversion rate a prior report had shown as 5.0 percent turned out, under the corrected and now-standard method, to be 2.2 percent, a number the team could finally defend in the room where the budget decision got made.

The clinic wanted to cut its ad budget, convinced the leads coming in were garbage. Its own spam filter and front-desk triage were marking a large share of inbound inquiries as junk, and on paper that looked like a straightforward case for spending less on ads.

We went back through the leads the filter had discarded and read what people had actually written, then traced each one against the call log to see what really happened after it arrived, rather than trusting a label a busy front desk had applied in the moment.

41% of the leads marked as junk turned out to be real patients describing real conditions in their own words, people who had been called once, gotten a voicemail, and never been reached again. The problem was never the leads. It was follow-up, so the fix was not a smaller budget, it was a faster and more persistent response.

Every lead from the site was supposed to land in the same automated pipeline the whole team relied on. Nobody had a way to know when a piece of that pipeline quietly failed. The tools involved retried a few times on their own, then gave up without telling anyone, so a broken step looked, from the outside, exactly like a quiet day.

We audited the pipeline end to end and found it was silently dropping close to 18% of the leads passing through one of its central steps, with no alert and nothing surfacing the failure to a person. We built a monitoring layer underneath the whole system: daily checks against what a healthy day looks like, watching for the exact fingerprint the kind of failure that had been running invisibly leaves behind.

The failure that used to run silently for days now gets caught by a system built specifically to watch for it, before it turns into a lead nobody ever called back.