Under the Hood
When a campaign goes live, you stop building features and start watching data. Except when you realize the data is wrong. Then you're building features again, just faster, because now someone is paying to look at the numbers.
The dashboard
Nick's campaign was running. Impressions were flowing in. What we didn't have was a way to show him any of it.
So we turned his proposal page into a live dashboard. Same URL. The "Ready to launch in Nashville?" section became "Need anything?" The measurement baseline shifted tone. "This is where we started." An interactive map where you can click through days and see which venues lit up. Progress bars showing delivered versus goal. Same page, completely different job.
We shipped it. Deployed it. Opened it up to admire our work.
The data was wrong.
The pixel mystery
In programmatic digital out-of-home, you track impressions with a pixel URL. You provide a tracking URL when you set up the campaign, and when your creative wins an auction and gets served on a screen, the platform pings that URL. Each ping carries a venue ID so you can map every impression to a specific location.
Our venue IDs weren't matching.
The database had IDs from inventory files. The pixel was sending different IDs for the same venues. Same locations. Different numbers. After a full session of detective work, comparing pixel data against CSV exports and matching on everything except the ID itself, we figured it out. There are three separate ID formats for every venue. A venue ID, a targeting ID, and an industry ID. The pixel sends the industry ID.
This is not documented anywhere.
Once we mapped all three formats together, every single impression resolved to a venue. 8,857 impressions backfilled. The map went from mostly blank to fully populated. There's a meaningful difference between a tracking system that works and one that looks like it works, especially when someone is paying you $749 a month to look at it.
Billboard is a marketing term
While debugging, we noticed "billboard" existed as both a venue category and a venue type. The industry terms are bulletin, poster, junior poster. I know this. I spent twelve years at Clear Channel. The category should have been "roadside" from the start, which is what I always thought we should call it there too.
But "billboard" is what normal people say. It's the marketing language. Everyone knows what a billboard is. Nobody knows what a bulletin is. So we kept "billboard" as the customer-facing venue type and renamed the category to "roadside" underneath. The right answer was always the simple one.
Two products
Same week, we rebuilt the homepage around a two-product model. The original pitch was one thing: subscribe to a city, your show appears on screens. The new version introduces two strategies. Presence goes deep in one metro (what Nick is doing). Advance goes wide across a tour schedule (a different kind of buyer).
The rebuild touched everything. New page structure. "Markets" renamed to "cities" in all customer-facing copy. New nav. Product stubs. 301 redirects. FAQ rewrite. And we deleted 2,700 lines of an old plan-builder feature that never shipped. That felt good.
It's on a branch. We want the product pages done before it goes live. But the bones are solid.
Creative rotation
Nick publishes about six episodes a month. Each one should become a new ad. We built a rotation system: newest episode gets double weight, older ones split the rest, oldest drops off when a fourth arrives. By the end of the week, 96 of 165 creative submissions were approved by venue operators across Nashville. Zero rejections.
The campaign refreshes itself whenever he publishes. Nobody has to touch it.
What's next
Nick's campaign has a week of real data. We're building a reporting view so we can actually see the patterns.
The homepage rebuild is waiting on product pages. And Brian still hasn't sent the CSV.