Where the thread breaks

How the setup got this way

How did it get this way? One sensible call at a time.

The ad platforms asked for a pixel, so the pixel went on the site. Analytics went on next, because the team needed to see the traffic. Each campaign got its own landing page and form, because the tests said dedicated pages convert better. The CRM got a connector that turns every form into a lead. Somebody added a chat widget. Sales started keying in the leads that called. Tracking tags went on the ads that someone remembered to tag.

Each of those calls made sense when it was made. Together they leave five or six systems that each know part of the story, with no thread running through them.

Pick the best account the business signed last quarter. Which ad brought it in? Which form did it fill in, and what did it say about itself on the way? In most setups, the answer is a source field that says “Facebook,” “Direct” or nothing at all. An app stripped the tags, or the lead came in on a different form, or two systems each created the lead and the copy that survived lost its tags. The click is in the ad platform, the lead is in the CRM, the decision is in underwriting and the revenue is in billing, and nothing joins them.

So the platform’s report and the finance report never meet. The platform counts conversions. Finance counts revenue. The only thing that connects them is an ID that every system kept.

The platforms learn from what you count

Automated bidding on the major algorithmic advertising platforms learns from the conversions it’s handed. Google says so in its own help pages: “Google uses the data included in your conversions reporting to optimize your bids when you use an automated bidding strategy.” It works the other way too: “Choosing not to include a conversion action means your automated bid strategies won’t bid for these particular conversions.”

Hand the bidding every form fill as a conversion, and it goes and finds more form fills, wherever they cost the least. The algorithm treats every conversion as worth the same, as if the leads sat on a bell curve around an average. In a power law business the leads aren’t alike, and the cheapest ones turn into the smallest accounts. It’s a snake eating its tail. Every time the bidding optimizes to the average, the average gets a little worse.

Count only the leads that carry a signal of value, and the bidding goes looking for more of those. That’s the whole idea. The rest of this guide is what it takes to do it: a thread that lets you see which leads turned into value, a signal that shows up early enough to act on, and a conversion setup that tells the platforms about the right leads and nothing else.