Most AI projects can’t show a return because nobody wrote down where they started. “It feels faster” isn’t a number. So before we build anything for a client, we spend the first stretch of every engagement measuring. You can do a rough version yourself this week, no software required.
Why speed is the first thing to check
The best-known study on this is from Harvard Business Review in 2011 (“The Short Life of Online Sales Leads”). Companies that contacted a web lead within an hour were nearly seven times as likely to qualify it as companies that waited even one hour longer. The exact multiple will differ for your business. The direction won’t: people who ask for a quote are usually asking several businesses, and the first one to answer has a real advantage.
The one-week count
Pick a normal week. Keep a simple sheet with these columns:
| What to count | Where it comes from |
|---|---|
| Inbound calls, and how many went unanswered | Your phone provider’s call log (most export to a spreadsheet) |
| Unanswered calls that got a callback, and how long it took | Call log plus whoever returns calls |
| Web form and email inquiries | Your website form notifications, shared inbox |
| Time from inquiry to first reply | Timestamp of the inquiry vs. timestamp of your first reply |
| Quotes requested and quotes sent, with dates | Your quoting tool or sent folder |
| Jobs won from those quotes | Your CRM or job list |
Don’t try to be perfect. A week of honest, slightly messy counts beats a year of guesses.
Turning it into money
Once you have the counts, the math is simple:
- Missed opportunities per month = unanswered calls with no same-day callback + inquiries that waited more than a few hours, × 4.3.
- Of those, how many were real jobs? Use your normal close rate. If you win 1 in 4 quotes, assume roughly 1 in 4 of the missed ones would have become work. That’s conservative, since some of them never got a quote at all.
- Value = recovered jobs per month × your average job value × your gross margin.
Write down each input, so anyone can check the logic and replace a guess with a real figure later.
What to do with the number
- If it’s small, great. You’ve learned that speed isn’t your leak, and you shouldn’t buy a system to fix it. (We’ll tell clients this too. It happens.)
- If it’s large, you now have a target: the after-number a system has to beat, measured the same way, at day 30 and day 90.
- Either way, you have a baseline that belongs to you, not to a vendor’s sales deck.
That’s the whole method behind how we work: count first, build second, and report the difference the same way you counted it.