Email and customers · Lesson 5.4
Measuring AI Lead Response: The Metrics That Matter
It is easy to measure the wrong thing in AI lead response. Replies sent, messages drafted and hours saved all look good on a screen and tell you little about whether leads are turning into customers. This post names a handful of numbers that do tell you something, explains how to collect them with nothing fancy, and shows how to read them in a short weekly review. All sample figures are made up to show the method, not results from anyone.
Lesson 5.4 sets up the reply itself: Reply to every new lead in five minutes.
Track five numbers weekly: time to first reply, reply rate, booking rate, edit rate and lead to customer rate, and judge them over weeks, not days.
Activity versus outcome
Activity metrics count what the system did: replies sent, drafts created. Outcome metrics count what happened to the lead: answered you, booked, bought. Activity is easy to inflate and is mostly useful for checking the setup works. Outcomes are what you care about.
Speed sits between them. A 2011 Harvard Business Review study of 1.25 million leads found that firms contacting a lead within an hour were nearly seven times as likely to have a real conversation with a decision maker as those who waited an hour longer. It was observational, US based and mostly about larger firms, so it is a reason to watch your own speed, not a rule that will hold for you.
The five numbers
| Metric | How to calculate | What it tells you |
|---|---|---|
| Time to first reply | Reply time minus lead arrival time | Whether the speed promise is kept |
| Reply rate | Leads who answered / leads you replied to | Whether the message earns a response |
| Booking rate | Leads who took the next step / leads replied to | Whether the step is clear and small |
| Edit rate | Drafts you changed / drafts reviewed | Whether the instructions are good enough to trust |
| Lead to customer rate | Customers / leads, over a longer period | The outcome that pays |
Time to first reply
Use the middle value, not the average, because a single overnight lead can distort an average. Also record the share of leads answered within an hour. Check after hours separately, since that is where automation should help most.
Reply rate
This tells you if the message works. If few leads answer, look at the opening line and the ask. A low rate with fast replies suggests the content is the problem, not the speed.
Booking rate
Count the share who took the one step you offered. If reply rate is fine but booking is low, the step may be too big or unclear. Try a smaller step or a clearer link.
Edit rate
This is the number most people skip. It shows how much you still fix by hand. Track it during review. When it is low for several weeks, you can consider letting routine replies send on their own. When it rises, something changed: new offers, new questions, out of date facts.
Lead to customer rate
The slow number. It needs weeks or months, and it is affected by price, season and word of mouth, not only your reply. Use it to look at trends, not to judge one week.
How to collect them
You do not need analytics software. A simple sheet with one row per lead works.
- Date and time the lead arrived.
- Source: form, email, other.
- Time of first reply.
- Hot or routine.
- Did they answer?
- Did they book?
- Did you edit the draft?
- Outcome: customer, not now, no answer.
Fill the last columns during your weekly review. The CRM lesson shows a neat way to hold this so it does not become a chore.
An illustrative weekly review
Say you had 20 leads last week. These figures are invented.
- 18 got a reply within an hour, 2 arrived overnight and waited until morning.
- 9 answered. Reply rate is 9 out of 20, or 45%.
- 4 booked. Booking rate 20%.
- You reviewed 20 drafts and edited 6. Edit rate 30%.
Reading it: speed is fine. Replies are decent. Bookings are the weak point, so test a clearer next step. Edits are fairly high, so keep reviewing and look at what you changed. Do one change this week, not four.
Rules for reading numbers honestly
- Wait for enough data. With 20 leads, one booking changes the rate by 5 points. Look at four weeks together before concluding.
- Compare like with like. A holiday week is not a normal week.
- Change one thing at a time. Write the date of each change.
- Do not claim causes you cannot see. If sales rose, was it the reply, the season or an ad? You may not know.
- Keep a before figure. Measure a couple of weeks before automation if you can, from your inbox history.
What not to track
- Number of automated replies, which only shows the machine ran.
- Hours saved estimated from nothing. If you cannot time it, do not claim it.
- Open rates alone. They can be unreliable, and a reply from a person is a stronger signal.
Measuring consent and opt outs
If your follow ups are advertising, add one more line: unsubscribes handled. The FTC says the CAN-SPAM Act covers all commercial email, including business to business, and that opt outs must be honored within 10 business days, with accurate sender details and your valid postal address included. Check the weekly list of unsubscribes against what you sent. Messages that only facilitate an agreed transaction are treated differently. This is general information, not legal advice. For the follow up sequence itself, see the welcome email lesson.
Questions people ask
What is a good reply rate?
There is no universal figure. Compare your own rate over time and change one thing at a time.
How long before I can trust the numbers?
Look at several weeks together. Small samples swing a lot.
Do I need software to track this?
No. A sheet with one row per lead is enough to begin.
What if the numbers are flat?
Check the first line and the next step first. They are the quickest to change, as in How to brief AI like a senior hire.
Start with one sheet
Set up the sheet, write your starting figures, and review it on the same day each week. Lesson 5.4, Reply to every new lead in five minutes, gives you the reply to measure, and the five email journey shows what to track after it.