Every small business we look at has the same shape of leak. Enquiries arrive — a form, a chat, a missed call — and a good share of them go quiet. Not because the person lost interest, but because they were comparing three options on a Tuesday evening and yours was the one that did not follow up.
The acquisition cost for that person is already spent. Ads, search, the referral, the years of being good at the job. What is missing is the cheap bit at the end.
Same-session conversion is the wrong target
Website assistants are usually sold on instant answers: someone asks about opening hours at 11pm and gets a reply. That is genuinely useful, and it is where most implementations stop — the assistant answers, captures a name and number, fires a notification, and considers the job done.
But the notification lands in an inbox alongside everything else. If nobody acts on it that evening, the lead ages. By the time someone follows up on Thursday the job has gone elsewhere. The assistant did its part; the process around it did not.
The question worth asking is not "did we answer?" but "what happened to the ones who asked and never came back?"
Make the gap explicit
The approach we build is simple and mechanical. Every captured lead carries its own state:
- When it was captured, and whether it has converted
- If it converted, whether that happened in the same session or later
- Whether a follow-up has been scheduled, and whether it has fired
If a lead has not converted after a configurable delay, the system re-notifies the team — unless the person has since booked, in which case the follow-up is silently cancelled. No chasing customers who already said yes.
The delay is per-client, because the right interval for an emergency plumber and a wedding-hair booking are not remotely the same.
The number that makes the case
The reason to track conversions as same-session versus after-follow-up is that the second category is the entire argument. It is the set of jobs that would not have happened otherwise.
Multiply that count by average job value and you have a figure, in pounds, for what the follow-up recovered. Not a projection or an industry benchmark — a count of this business's own jobs that came back after being nudged.
That is the difference between selling "AI for your website" and showing an owner a dashboard that says these specific enquiries were going cold, and here is what came back. One is a feature. The other is an invoice they are happy to pay.
Worth being honest about
Two caveats we would rather state up front than have discovered later.
The recovered-revenue figure depends on an average job value that someone has to supply honestly. Put an optimistic number in and you get an optimistic number out; the dashboard is only as good as that input.
And follow-up is a tool with a short half-life. One well-timed nudge to someone who genuinely enquired is a service. Four is harassment, and it will cost you more goodwill than the job was worth. The point is to close a gap, not to install a pursuit.