Nine Seconds vs. Four Minutes

A home service company was spending $25,000 a month on ads. Their agency reported twelve booked jobs from Google and eight from Meta. The numbers looked roughly even.

We put the phone data next to the ad data. Google calls averaged four minutes. Meta calls averaged nine seconds.

Nine seconds is a hang-up. It is a wrong number. It is not a customer.

The Meta numbers had been inflated for months. Not because anyone was lying, but because the ad platform counts a connected call as a conversion, and the booking software counts a booked job as a booking, and nobody had ever put those two lists in the same room.

The four-system problem

Every home service business we talk to has the same shape: an ad platform reporting clicks and calls, a website with contact forms, a phone system logging every inbound, and a booking or CRM tool recording jobs. Four systems. None of them speaking to each other.

So the owner asks what the ads produced last month, and gets four different answers from four different places. The budget decision gets made on a gut call. Every month.

This is the problem Prism is built to solve.

What Prism connects

We built Prism as a portal that sits over all four systems and traces each dollar to its outcome. An ad click becomes a session, the session becomes a phone call, the call gets matched to a booked job. When the trail holds, you see the full line. When it breaks, we say so rather than assigning the credit somewhere plausible.

The dashboard shows spend by channel next to booked revenue, with the previous period beside each number. Every call is logged: how long it ran, which ad it came from, what was actually said. Short calls filter out in about four seconds of looking.

One screen we spent real time on: lifetime customer value, attributed back to the original acquisition. When a customer books their second job two years later, the credit stays with the channel that first brought them in. The ad that looks expensive on its first sale is often not expensive at all once you account for what that customer does next.

What it will not do

Prism does not run your ads. It does not guess when the attribution trail goes cold. When someone clicks an ad from one number and books under another, the job sits in an unknown column instead of being padded into whichever channel looks best that week.

The findings screen shows a short list: what to change, in order, with the numbers it was measured against. Each item carries the evidence so the owner can argue with it if the finding is wrong.

How we built it

Setup takes two weeks. The first couple of days are mapping: where does your data live, what does your phone system export, how does your booking tool label things. Days three through seven are the connections and the cross-check, verifying every number against its source until the totals match. The second week is calibration: what counts as a qualified call here, what a bad week looks like for this business specifically.

Prism launched on the Twenty1 site at /prism on September 19. If you want to see it running on your own numbers, the call is fifteen minutes and you bring your last ninety days.