What to measure before you spend on ads

Most advertising problems we are asked to fix are not advertising problems. The campaigns are competent. What is missing is any agreed definition of what the campaigns are supposed to produce — so nobody can tell whether they worked, and the argument gets settled by whoever is most senior in the meeting.

The fix takes about a fortnight and has to happen before the spending starts. Here is what it involves.

1. What counts as a lead

Write it down. Not "an enquiry" — the specific, checkable condition that makes someone a lead rather than a form submission.

A newsletter signup and a request for pricing are not the same event, but a default tracking setup counts them identically. So does a recruiter, a competitor doing research, and a supplier trying to sell you something. In B2B, a meaningful share of inbound form fills are none of them a prospect.

The definition needs to be specific enough that two people in your business would sort the same twenty enquiries into the same two piles. If it isn't, you don't have a definition — you have a word everyone is interpreting differently.

2. What a lead is worth

You need three numbers, and they can be rough:

• What proportion of qualified leads become customers
• What a customer is worth, over the whole relationship rather than the first invoice
• What proportion of that is margin rather than revenue

Multiply them and you have the most useful number in the business: what you can afford to pay for a lead and still make money. Without it, every conversation about cost per lead is aesthetic — £40 sounds better than £120, but one of them might be profitable and the other ruinous, and nothing in the ad platform will tell you which.

In long-cycle B2B this number is harder to pin down and more valuable when you do. If the sales cycle is nine months, the first campaign report lands long before the revenue does, and the temptation is to judge the campaign on the only number available — volume. That is how businesses end up optimising towards cheap leads that never close.

3. Where leads come from now

Before adding a channel, establish what the existing ones produce. Not what the analytics dashboard says — what the sales team can actually confirm.

The usual method is unglamorous: take the last fifty closed deals and trace each one back to where it started. It takes an afternoon and it routinely contradicts the dashboard, because most analytics setups quietly credit the last click before purchase rather than the thing that created the interest.

This gives you a baseline. It also tells you whether you have a traffic problem or a conversion problem, and those need completely different budgets. Adding advertising to a site that does not convert its existing traffic buys you more people to fail to convert.

4. How long the cycle really takes

Ask sales for the spread, not the average — shortest, typical, longest. The average hides the shape.

That spread decides when you are allowed to judge the campaign. If deals typically take four months, a review at week six is measuring noise, and acting on it means killing campaigns before they could possibly have produced revenue. We have watched a company cancel the campaign that eventually produced its largest contract of the year, because the review happened at the wrong time.

Agree the review date before launch, at a point that makes sense for your cycle, and agree what you would do at that review if the number is bad.

What this looks like when it is in place

You get a sentence that sounds obvious and is unusually hard to earn:

We can pay up to £X per qualified lead. We currently get Y a month, mostly from Z. We will review at month N, and if we are above £X per lead we will do the following.

Every subsequent decision follows from it. Whether to increase budget, which channel to cut, whether the agency is working — all of it becomes arithmetic instead of opinion.

The cost of doing it afterwards

Skipping this stage does not save time; it moves the work to a worse moment. Six months in, you have spend, some leads, no agreed definition of a good one, and a disagreement between marketing and finance that nobody can settle with evidence.

Reconstructing the baseline retrospectively is also harder than building it up front, because the period you would compare against is now contaminated by the advertising.

Where to start

If you do only one of the four, do the second — what a lead is worth. It is the number that makes every other decision tractable, and most businesses can get a workable version of it from data they already hold.

If you would like a hand establishing it, that is the first fortnight of most of our analytics engagements, and you keep the model whether or not you continue with us.