How to Read What Your Agency Sends You: Four Questions for Any Report

Once a month a document arrives. It has charts. The numbers are larger than last month’s. Somebody has clearly worked on it.

And you have no idea whether you made money.

That’s the normal condition of a contractor with an agency, and it isn’t usually because anyone is lying. It’s because a report can be entirely accurate and still be constructed so that the question you actually care about never comes up.

The one thing a report must do

Put two numbers on the same page: what a customer cost, and what a customer is worth.

That’s it. Everything else is detail.

One cleaning company had a monthly report for a year. It said how many leads came in. The number was rising. Nobody was lying. What the report never did was put the cost of a lead next to the value of a customer — so nobody ever had to notice that one was bigger than the other, and the business lost $15.75 on every one-time job it bought.

A report that counts volume without pricing it is not a report. It’s a receipt for activity. It tells you that work was done, which you already assumed, and nothing about whether the work paid.

So: the minimum acceptable report contains cost per lead, cost per job, and your maximum for both — the last of which you brought, because they can’t compute it without your margin. Our maximums calculator produces those three ceilings from seven inputs.

If those three things are on one page, everything after this is easy.

Then: broken out, not lumped together

A single averaged cost per lead across all channels hides more than it shows.

By service line. Kitchens and decks have different maximums. A single averaged figure that looks fine can be a healthy deck campaign carrying a kitchen campaign that’s losing money.

By channel. Your conversion from lead to job differs by channel — a search lead and a shared marketplace lead are not the same object. Blend them and you’ll fund the wrong one.

By job type where it matters. The recurring cleaning customer and the one-time move-out justify completely different acquisition costs.

Net of credits. On pay-per-lead channels, refunded leads must be out of both the cost and the count. Most of that screening happens automatically, but the credits only arrive if your dispatcher has been keeping lead statuses current. If nobody has, the reported figure is wrong in a direction that makes the channel look worse than it is.

You don’t need all of this in month one. You need it before anyone proposes increasing the budget.

Whose number is whose

This is the part that turns a monthly argument into a monthly decision.

The calculation had inputs from two sides. The market’s price for attention belongs to whoever runs your advertising. Your website conversion and your close rate belong to you.

So when the economics don’t work, there are only three possibilities, and the report should make it obvious which one you’re in.

The market’s number came in above estimate. Cost per click or cost per lead is higher than planned. That’s the contractor’s side. Ask what’s being done: search terms pruned, bids adjusted, budget moved between channels, disputes filed. This is ordinary work and there should be evidence of it weekly.

Your number came in below estimate. The site converts worse than assumed, or the close rate does. That’s yours. No amount of campaign work fixes it, and a contractor who spends three months trying is wasting your money politely. The order of repair is where to start.

Both were roughly right and the economics still don’t close. Then the plan was wrong, not the execution, and you go back to the two-point test with real numbers. That’s a good outcome for month one and a bad one for month six.

The value of naming this in advance is that it removes the reflex. The default response to a disappointing report is to suspect the agency; the default response to a defensive agency is to suspect the client. Neither is a diagnosis. The four numbers are.

A bad month or a bad decision?

Some variance is weather. Some is signal. Confusing them costs money in both directions — firing a working partner after a rough three weeks, or tolerating decline for two quarters.

Three tests, in order of how much they tell you.

Is the sample big enough to mean anything? Three leads on a Tuesday and none on Wednesday is noise. A month with a handful of conversions supports almost no conclusions. Before you act on a number, ask how many events it rests on — and if nobody can answer that, that’s your answer. The significance calculator settles it in about two minutes.

Is it a wobble or a trend? Week-to-week variation averages out. A metric that gets worse every week for a month is not variance. That’s a signal about your business, your market, or something that changed, and it deserves an investigation rather than a bid adjustment.

Did anything change outside the account? A competitor started spending. A season turned. Your best salesperson left. Your phone system broke and nobody mentioned it. More reports are wrongly blamed on advertising than on any other cause, and the fix is usually one honest question asked internally before the meeting.

Now the uncomfortable part: incentives

You should know how the person running your advertising gets paid, because it tells you which advice will come easily to them and which won’t.

Percentage of ad spend is the most common model in this industry. It’s simple and it aligns effort with account size. It also means your agency earns more when you spend more — which is fine right up until the honest recommendation is to spend less.

Think about the two moments where that matters. An appliance business that shouldn’t scale until its dispatcher improves. A storage company that fixed its sales process and no longer needed advertising at all. Both of those recommendations cost the agency money.

Percentage-based agencies can give that advice — ours is paid that way, and we’ve given it. But you should know the shape of the incentive, and you should notice whether the person you’re paying ever recommends anything that reduces their own invoice.

That’s the test, and it’s a good one. Not “do they seem honest.”

Has this person, at any point, told you to spend less, pause a channel, fix something internally first, or hold volume flat?

If the answer over a year is never, one of two things is true: your business is in unusually good shape, or nobody is looking.

What to do when it doesn’t converge

Not fire anyone, first of all. That reflex — the client pays and therefore blames — produces a new agency every eighteen months and the same results.

The sequence that works:

1. Establish which number missed. Yours or theirs. One conversation, four figures.
2. If it’s theirs, ask for the weekly work. What search terms were removed, what was reallocated, what disputes were filed, what changed and when. A good answer is specific and boring. A bad answer is about impressions, brand awareness, or the algorithm.
3. If it’s yours, stop the campaigns while you fix it. Genuinely stop them. Buying leads for a process that loses them is the most expensive patience there is.
4. If the plan was wrong, redo the two-point test with the real numbers. Sometimes it still says go, at a different volume or a different channel. Sometimes it says wait, and you’ve bought that answer for one month instead of finding it out over a year.

Fire someone when the pattern repeats: no weekly evidence, no answer about hours, reports that still count volume without prices after you’ve asked, and a consistent absence of any recommendation that costs them money. Those four together are a pattern, not a bad month.

Four questions for any report

Print these. They take five minutes and they work on any document, from any agency, in any trade.

1. What did a job cost me, and what is a job worth to me? If the report can’t answer both, it isn’t finished.
2. Which number moved away from plan — mine or the market’s?
3. How many events is this conclusion based on? If the answer is a handful, there is no conclusion — only a feeling with a chart attached.
4. What did you change last week, and why?

The fourth is the one that separates a managed account from a parked one. Advertising isn’t a setup, it’s a weekly habit — the difference between a lead at ninety-six dollars and a lead at twenty-two was three weeks of it. If nothing changed last week, and nothing changed the week before, you are not paying for management. You are paying rent on a login.

This chapter is drawn from Win in the Spreadsheet First. The full method, including what to ask before you hire, is in the book.