We replaced the marketing agency in October. The handoff was clean, and I'll take some credit for that: reassignments planned, checklists written, a month of overlap so nothing dropped. The new agency cost about a third less, and within a quarter lead volume was up by close to half.
Eight months later we ended the program.
None of that is a criticism of the second agency. They improved every number they controlled. The problem was that the numbers they controlled were never the ones deciding whether marketing made money, and I couldn't see that clearly until I'd rebuilt the funnel from the bottom.
You can't hold a vendor to a standard you can't measure
Before any of this there was a tracking problem. Roughly a third of leads arrived with no recorded source.
That gets filed as a data-hygiene issue, which undersells it. If a third of your leads are unattributed, you don't know cost per lead for any channel, which means you don't know whether a channel works, which means every conversation with the agency is two people reading an incomplete report in whichever direction suits them.
Fixing it wasn't clever. I audited where the missing leads had actually come from, built the capture into the process so it didn't depend on anyone remembering, trained the team on it, and tied it to how their performance was assessed. By the end, unattributed leads were near zero.
Only then was there anything worth arguing about.
Read the funnel from the bottom
Most people read a funnel top-down: leads, bookings, consults, clients. It's the order the customer moves through, so it feels like the order to analyze.
It's the wrong order. Top-down makes each stage a percentage of the one above it, and you end up managing volume, which is the easiest thing to buy and the least likely thing to be broken. From the bottom, each stage raises the question the next one answers.
Clients. Ours were low-value and high-friction, which showed up as complaints and chargebacks rather than in any marketing report. That's a marketing finding before it's a service finding: the campaigns were attracting people shopping on price.
Completed consults. Sales conversion held at 25 to 30 percent of completed consults. That sounds respectable, and at a low price barrier it isn't. Cheap things should close easier than that.
Booked consults. We booked 25 to 35 percent of leads, which I first read as fine. It wasn't quite real. Intake was cancelling some booked consults after a qualification call, so the booking rate counted appointments that were never going to happen and the show-up rate absorbed the difference. Two metrics describing the same disqualification from opposite ends.
Qualified leads. Not tracked at all. I wrote criteria for what qualified meant, added the field, and it came back at 55 to 65 percent.
Total leads and cost per lead. Which is where it resolved.
The arithmetic that ended the program
Multiply the stages and combined lead-to-client conversion was 1 to 2 percent. Cost per lead was around $150, sometimes over $200.
At 2 percent you buy fifty leads to get one client. At $150 each that's $7,500 to acquire a client, and the average new-client sale was under $3,000. For the spend to break even, a client had to be worth more than $10,000. They were worth less than a third of that.
No vendor closes a gap that size. A cheaper one just slows the loss.
So two things happened at once. The agency switch brought cost per lead under $100 and raised volume. Separately, I restructured pricing, and the average new-client sale went from about $3,000 to about $5,000.
Combined conversion didn't move. Two percent before, two to three after, across two agencies and eight months of incremental work on targeting and messaging. But run the same calculation again: at $100 a lead and 2 percent conversion, a client costs about $5,000 to acquire, and a client was now worth about $5,000. Acquisition cost had gone from $7,500 against a sub-$3,000 client to $5,000 against a $5,000 one.
The math closed. It closed because of the pricing, with the agency's improvements making the result survivable rather than creating it.
Why we stopped anyway
Break-even isn't a reason to keep spending.
Everything that could improve had. Attribution was clean, cost per lead was down by a third, volume was up, prices were up. What hadn't moved in eight months was conversion, and that was the number that would have turned marketing from a cost recovery into a profit center.
At that point continuing is a decision to buy revenue at cost. Sometimes that's right, if the volume is strategically valuable or you think conversion is about to break. I didn't think so, and I'd run out of evidence to argue otherwise.
The decision was made jointly, by the owner and me, and the marketing arithmetic was one input among several. The others were strategic and aren't mine to discuss. But the arithmetic on its own would have got us there, and it's the part that transfers to anyone else's business.
What to do before you switch agencies
A marketing function has four levers. The agency controls two: cost per lead and volume. You control the other two: conversion and price. Most agency changes are an attempt to fix all four by replacing the people who control half.
So, in order.
Fix attribution. Not to a comfortable level — to near zero unattributed. Nothing downstream is real until that's done, and it's usually weeks of work, not months.
Compute combined conversion, lead to paying client, as one number. Multiply every stage rather than carrying them separately in your head. Most owners have never done this and are startled by the result. One to two percent is common.
Then divide cost per lead by combined conversion. That's what a client costs you. Compare it to what a client is actually worth, not to what your best client was worth.
If the gap is small, you have an execution problem and a better agency might close it. If the gap is a multiple, you have a pricing or offer problem wearing a marketing costume, and switching vendors buys you the same outcome at a lower monthly fee.
Establish what's true, set the goal the truth can carry, then decide the actions. That's the TAG method — See the Truth. Implement Actions. Achieve your Goals. Here the truth was that we'd been asking a vendor to solve an arithmetic problem only we could solve, and we'd have kept asking indefinitely if nobody had multiplied the stages together.