Close the loop

With the thread in place, every lead can be judged by what it became. That’s what turns a question on a form into a signal you can bid on, and it’s how you find out which channels, forms and doors are carrying the business.

Tier the book. Line up every account by what it’s worth over time, its net revenue, and read it on contribution too where you have it (the numbers, defined). Cut the book at the average, then cut each half again at its own average. The top two tiers, A and B, are the above-average accounts. C and D are below. In a power law book, A is a small group carrying most of the revenue.

Check the signal before you bid on it. For each answer band, compare how often its leads became above-average accounts with how often the book’s leads did. That comparison is the band’s lift. Whether a band has earned its place in the bidding is checked on your own book, cohort after cohort. Use only fields known at the time of the decision they’ll inform. A field filled in after the sale predicts perfectly and is useless as an early signal.

When one question stops being enough, score the first days. What a lead does in its first days with you can sort leads too, and it’s known soon enough to stay inside the platforms’ windows. Keep it to what the lead did with you. Nothing about the person’s finances, no trait the law protects, nothing bought from a data broker.

Read the funnel by cohort. Compare the share of each group’s leads that reaches each stage, from the lead through closed won. Web against Non-Web is where the calls, referrals and rep-entered leads finally get their own rates through the funnel instead of hiding inside “unknown.” When one cohort reaches the SQO at twice the rate of another at the same cost, the budget should move toward it.

Read by cohort month, at one, three and twelve months. A lead from January and a lead from June have had different amounts of time to become revenue. Group leads by the month they arrived and compare them at the same age, so a new channel isn’t judged against an old one that’s had a year to pay off.

ABPRO_ runs each of these on your own book.

What breaks

The thread breaks in the same handful of places, over and over. These are patterns from real builds:

• A tracking plugin overwrites the tags. Something on the site rewrites the tracking parameters, and every lead after the install arrives with the plugin’s values.
• Two tag manager containers, or one nobody owns. Legacy tags keep firing, and the pixel fires more than once on every page.
• An account-level goal overrides the conversion you set. The campaign bids toward whatever the account’s default goal says, including conversions nobody meant it to.
• A variable change sets every code to 0000. One renamed field and the qualified conversion can never fire again, and nothing errors.
• Forms die when an old system is switched off. Pages that posted to the old CRM keep taking traffic and send the leads nowhere.
• Several versions of the same form at once. Copies of the form live as separate pages, and each one drifts from the rest.
• The integration runs under one employee’s login. It stops the day that person leaves.
• Personal details in the step-two address. Names, emails and phone numbers end up in browser history, server logs and every tag on the next page.
• Reps re-create leads by hand. The new record has no tags and no lead ID, and the chain breaks at the CRM.
• Leads go only to the CRM. When the connection fails, the leads are gone, and nobody can say how many.
• Calls are never logged as leads. Inbound calls look weaker than they are, and their sources never get the credit.
• Customer records get deleted. The history goes with them, and so does the seed list for lookalike audiences.
• The rules change in a meeting the people who build the tracking weren’t in.

Two weekly habits catch them. The first is the tie-outs: each system’s count against the next one’s, so a failed connection, a dropped tag or a run of unlogged calls shows up the week it happens. The second is a check of the coverage of every field, by channel and by form. When a field drops on one channel, a template broke, an app is stripping the address, or a page went live without the script. The blueprint’s QA checklist covers the test clicks, the test submissions, the platform test tools and the weekly field check, and the basis columns show the same problem from the other side. When more of a channel starts resolving from the form or the browser instead of its own tags, something is stripping the tags.

And hand the raw data over, so the people who rely on the reports can check them.