The least glamorous arc — and maybe the one I'm proudest of. Marketing runs on numbers people assume are right. This is the work of making them actually right: connecting systems that didn't talk, reconciling data against reality, and automating the checks so trust doesn't depend on someone remembering to look.
① The mess — numbers nobody could fully trust
The data lived in disconnected systems — media calendars in Smartsheet, delivery data in the ad platforms, budgets in TapClicks — each with its own naming, none reconciled against the others. Pacing was manual, discrepancies went unnoticed until they were expensive, and "how are we doing?" had no single trustworthy answer. Before anything could be optimized, the data itself had to be made honest.
② Reconciliation — auditing the numbers against reality
The first move was proving what was actually true. I treated each client's CRM as the source of record and audited every platform-reported conversion against it — splitting leads by type and source, segmenting by platform and geography, and tracing each discrepancy to its origin. "The numbers don't match" became a precise, explainable account of exactly where and why — so reporting could rest on reconciled reality rather than platform optimism.
③ The plumbing — making the systems talk
Then I built the connective tissue. For pacing, I wired Smartsheet media calendars, TapClicks, and platform delivery data into Tableau — with an intermediate mapping layer to reconcile the naming mismatches that kept the systems from lining up — so budget-versus-actual and over/under-pacing surfaced automatically instead of by manual audit.
The deeper fix was a standardized naming taxonomy — wired into a tool that made errors impossible. I built a controlled vocabulary coding every campaign along consistent dimensions (client, channel, campaign type, goal, measurement, funnel stage, program or audience), then turned it into a guided builder: every taxonomy field is a dropdown, so nothing can be misspelled or invented. Enter how many ad groups and ads you need and the system generates exactly those rows — no manual construction, no drift — assembles the correct UTMs from the filled-in tabs, and runs Apps Script validation across the whole output before it ships. The media team just says what they need and gets perfectly structured, benchmark-ready naming every time.
And to bring the existing book of business along, I built a parser that mapped clients' already-in-market campaigns onto the new taxonomy — so ongoing work could join the benchmarks without re-tagging everything by hand. Slower than the greenfield builder, but a fraction of the fully manual alternative.
That taxonomy — and the guardrails around it — is the quiet foundation under everything else: the reason data from any channel, client, or program can be joined and compared, and the reason benchmarking across the whole book of business became possible at all.
④ The payoff — trustworthy, automated, benchmarkable
The result was infrastructure that did the watching itself: automated pacing that flagged budget problems as they emerged, and a clean, consistent data layer that finally made internal benchmarking possible — comparing performance across clients and channels on numbers everyone could trust. The unglamorous foundation that lets every dashboard, report, and recommendation built on top of it actually mean something.
This is the plumbing beneath the practice — and the clearest expression of what Text is Data is really about: not more numbers, but numbers you can believe.