Before you buy anything to fix lead follow-up, measure three things for two weeks: time from enquiry to first human reply (the median and the worst 10%, never the average), leads contacted per rep per day, and the share that got no reply at all. Those three numbers cost nothing, take a fortnight, and usually change what you would have built.
Most teams guess the first number is a couple of hours. Measured properly it is often more than a day.
The leak nobody logs
An enquiry arrives. Someone means to call. Something more urgent happens — a customer escalation, a deadline, a meeting that overruns. Two days pass.
Nothing broke. No alarm went off, no ticket was missed, nobody filed a complaint. The lead just went cold, and the only record is a row in the CRM that never changed status. This is why the problem survives for years in otherwise well-run companies: there is no failure event to notice.
The cost is real even though it is invisible. Harvard Business Review’s study of 2,241 companies found that firms responding to a new lead within an hour were seven times more likely to qualify that lead than those responding an hour later, and sixty times more likely than those who waited a day. A reply in minutes catches somebody mid-interest. A reply tomorrow catches somebody who has forgotten they enquired, and has probably contacted two of your competitors since.
The three measurements
1. Time from enquiry to first human reply
Not to an autoresponder. To a person, or to something that gives the enquirer information they actually needed.
Take the median, and take the worst 10% separately. This is the single most important instruction here. Averages are actively misleading for response time, because one enquiry that sat over a bank holiday weekend drags the mean somewhere unrecognisable, while a comfortable-looking mean can hide the fact that a fifth of your enquiries wait two days.
The worst 10% is where the lost revenue lives.
2. Leads contacted per rep, per day
Throughput. If it is low, the question is whether reps are short of leads or short of time — and the answer decides everything downstream. Automating follow-up when the real constraint is lead volume just gets you to the end of the list faster.
3. The share that got no reply at all
This is the one nobody wants to look at, and the one that most often produces a visible reaction in the room. Not “replied late” — never replied. In most teams that number is not zero, and almost nobody knows what it is before they measure.
What the numbers usually reveal
Three patterns come up repeatedly.
The tail, not the middle. Response time is fine for most enquiries and catastrophic for a minority — typically those arriving on Friday afternoon, out of hours, or while the usual owner is away. The fix is routing and cover, not speed.
Round-robin against reality. Leads are distributed evenly to reps who are not evenly available. Somebody on holiday keeps receiving enquiries. The fix is assignment based on who is actually free.
Nobody owns the inbox. Everyone assumes someone else has picked it up. This one needs no software at all.
Only one of those three is solved by buying a tool.
What to automate, in order of usefulness
Once you know which pattern you have, the options stop being generic.
An instant acknowledgement carrying a real time commitment. Not “thanks for your enquiry” — something that answers the question they actually have, and states when a person will reply. This alone moves the median dramatically, because the clock a customer feels is the one to the first useful response.
Routing by actual availability, rather than round-robin: who is working, who is not at capacity, who handles this kind of enquiry.
A drafted first reply, assembled from what the enquirer wrote and your past responses, so the rep edits rather than composes. This is where AI earns its place — not deciding anything, just removing the blank page.
Escalation when nothing has happened. If an enquiry has had no human contact in four hours, somebody senior should know without having to check.
It might also be nothing. Sometimes the measurement shows the process is fine and the constraint is elsewhere. That is a good outcome: you found out for the price of two weeks’ attention.
Doing it in the right order
The measuring costs two weeks and no money. The automation costs real money. Doing them in that order is most of the trick, for three reasons:
- You find out what to build. The three patterns above lead to three different builds. Guessing gets you the wrong one two-thirds of the time.
- You can prove it later. On day 90 you re-measure the same three numbers the same way, and the difference is not a matter of opinion.
- You might not need to build. The cheapest engagement is the one that ends with “one person owns the inbox from Monday.”
How to start on Monday
You do not need a project for this. You need a spreadsheet and two weeks.
- Export every enquiry from the last two weeks with its arrival timestamp.
- For each one, find the timestamp of the first human reply. Yes, by hand — the tedium is the point, because it is where you notice what the CRM is not recording.
- Calculate the median and sort to find the worst 10%.
- Count the ones with no reply.
- Write the four figures down with today’s date and show them to whoever owns the process.
That last step is what turns a measurement into a decision.
Lead follow-up is one of the workflows we most often start with — see what we actually build for the others, and how it works for what happens after the baseline is signed. If you would rather have someone take the measurement with you, that is the AI Impact Diagnostic, and its fee comes off the build if you go ahead.