The five-minute rule, measured on your own leads

The claim, and why we do not make it

You have read the statistic. Respond within five minutes and you are some large multiple more likely to win the job. It is repeated in every sales deck in the category, usually without a date, a sample, or an industry.

We do not use it. Not because it is wrong, but because it is not yours. A benchmark drawn from software leads in 2007 tells you nothing about whether the plumbing company you actually run converts faster callbacks better than slower ones. It might. It might not. The only way to know is to measure your own records.

So the detector does exactly that, and it will tell you there is nothing here rather than borrow someone else's number.

Two buckets, from your own leads

Every lead in your export is sorted by the gap between when it arrived and when someone first made contact. Reached within five minutes goes in one bucket. Reached after more than sixty minutes goes in the other. Everything between the two is left out on purpose — a lead answered at 40 minutes is neither fast nor neglected, and including it would blur the very comparison being made.

Then we compute the booking rate in each bucket and subtract. That difference is the whole finding. If your fast leads book at 45% and your slow leads book at 24%, the gap is 21 points, and it was measured on your business, in the period your export covers.

Both buckets need at least thirty leads before we will say anything. Below that the difference between them is noise wearing a percentage sign.

When the answer is no

A gap of zero or less means fast callbacks are not converting better for you. That is a real result, and the detector reports nothing rather than reaching for a benchmark to rescue the story.

A gap that rounds to less than one point is also refused. At 35.2% against 34.8% the summary would have read as a zero-point gap sitting beside a dollar figure, which is incoherent on its face and exactly the kind of line that makes a reader stop trusting everything else on the page.

The export that cannot answer this question

Plenty of CRMs export bare dates with no time of day. Parsed, every one of those timestamps lands on midnight, so every response gap computes as a whole number of days and nearly every contacted lead classifies as slow — on an artifact of the file format, not on anything your team did.

When almost every timestamp is midnight-exact, the detector makes no claim at all. This is the single most important behaviour in it: a minutes question has no answer in day-precision data, and answering anyway would produce a large, confident, completely fictional number.

Turning a gap into dollars, carefully

The gap alone is a percentage. To reach money it is multiplied by how many slow leads you get in a month, by what a completed job is worth, and by the share of won leads that actually finish.

That last multiplier is where this kind of estimate usually goes wrong, and it is worth being specific about. Jobs and leads arrive as separate exports, and nothing forces them to cover the same months. A jobs file going back three years divided by a leads file covering one quarter produces a completion rate above 100%, which then gets quietly clamped to 100% — and the formula ends up asserting that every won lead completes, which is the most expensive assumption in it.

So the jobs are windowed to the same period the leads cover before the division happens. If too few jobs fall inside that window, the wider set is used and the finding says so, because a number computed from a different period is a weaker number and you are entitled to know which one you are looking at.

What to do with it

The action is not a training programme. It is a routing rule: every new lead goes to whoever is free, with a five-minute clock, and anything unanswered escalates. Most shops that close this gap do it with a phone tree change, not a hire.

And the number moves next month, measured the same way, on the same records. That is the point of computing it from your data rather than a benchmark — you can check whether the fix worked.