Chapter one
The four words that book a job
Here is a real exchange from our book. Watch the last thing the contractor says. Nothing about it is specific to a trade.
She talked herself out of the appointment. The reply is four words and a new day. No apology. No "no problem, just let me know what works for you." No re-pitch. A slot.
The same shape turns up again and again once you start looking for it:
- 1Get them talking about timing. An open question works. So does naming a couple of times, which is what happens above. Either one is fishing for the same thing: their real constraint.
- 2Let them name the obstacle. Afternoon. Next week. My mom's appointment. This is the useful part, and most people talk over it.
- 3Put a concrete time back on the table. A day, or a day and an hour. Anything except "let me know what works for you", which hands the work back to the person who has just finished telling you they are busy.
Chapter two
The follow-up you were about to skip
Everyone knows the first follow-up works best. That is true, and it is the wrong thing to act on.
Each chase pulls a smaller share than the last. But the leads you recover keep stacking, and the cost of getting one keeps falling:
| If you stop after… | Leads recovered | Cost per response |
|---|---|---|
| 1 follow-up | 21.1% | $237 |
| 3 follow-ups | 35.3% | $142 |
| 6 follow-ups | 41.8% | $119 |
| 8 follow-ups | 45.5% | $110 |
The part that surprised us
Of the 543 leads still being chased at the eighth message, 3.7% answered it — the same share the fifth message pulled from its own pool. That is a different measure from the table above, which counts against every lead you started with. Read this one as: the well does not run dry. It flattens out and stays there.
A text costs about a cent. So the eighth chase bought another 20 responses for roughly five dollars of sending, against leads you had already paid thousands to acquire.
Chapter three
What to take out of the text
We split every message by one feature at a time. Nearly every instinct that feels like marketing makes the number worse.
| Do this | Rate | Not this | Rate |
|---|---|---|---|
| No emoji | 18.4% | Any emoji | 3.5% |
| No offer words | 23.8% | "free / discount / $ / expiring" | 5.7% |
| 61 to 100 characters | 23.6% | Over 260 characters | 9.1% |
| One or two lines | 18.2% | Multi-line block | 11.3% |
| Asks a question | 18.4% | No question at all | 13.2% |
Openers that actually earned a conversation
Every one of these is a plain question. Not one leads with an offer. They are quoted as sent, so swap the trade word for yours: furnace, panel, water heater, windows.
And the ones that got ignored
1.7% and 2.5%. Both sent hundreds of times. Rewrite these for your own trade and they still fail, because the emoji and the offer are what killed them.
Chapter four
The clock
Reply rate by day and time, in the homeowner's local time. Pin this up.
| Mon | Tue | Wed | Thu | Fri | Sat | Sun | |
|---|---|---|---|---|---|---|---|
| 6–9am | 15% | 14% | 11% | 18% | 18% | 14% | 26%° |
| 9am–1pm | 25% | 22% | 16% | 15% | 46% | 23% | 22% |
| 1–6pm | 14% | 15% | 11% | 15% | 36% | 14% | 22% |
| 6–9pm | 26%° | 31%° | 38%° | 29% | 37% | 15% | 39%° |
Fewer replies More · Cells marked ° had fewer than 200 texts, treat them as a hint rather than a number.
The short version
- Friday late morning is your best hour of the week. By a distance, and on a big enough sample to trust.
- Weekday afternoons are the weak spot and they are where most home service contractors send the most.
- Sunday is not dead. It is one of the strongest days we measured and gets the least attention.
- You are sending in the wrong block. Early afternoon takes 6,373 of our texts and returns 17.2%. Evening takes 1,464 and returns 29.2%, level with late morning. The evening sample is thin enough that we would test it before moving the whole schedule, but it is the clearest misallocation on the grid.
- When they reply, they reply fast. Median 3.5 minutes, and 65% inside a quarter of an hour. If you are not there to answer, the window closes.
Before you use any of this
What this is, and what it isn't
This is our own book, not an industry study. 1,500 roofing leads, 17,350 text messages, 3,877 replies. One lead source, one trade, weighted to Florida. Your numbers will differ. The shapes probably won't.
It says home services on the cover and roofing in the sample. Here is why we think that is fair. Every lever we tested is a property of the message and the clock, not of the job: whether it carried an emoji, whether it led with an offer, how long it ran, whether it ended in a question, what hour it went out, and how many times we chased. A homeowner deciding whether to answer an unknown number is doing the same thing whether the text is about a roof or a furnace. That is a reasonable expectation, not a measured result. We have not run this book in HVAC or plumbing. Treat every number here as the hypothesis to test against your own CRM, and the direction as the thing worth acting on.
There are two reply rates in here and they are not the same measure. The clock counts replies landing in an hour against texts sent in that hour. It is a proxy, because a reply can arrive in a later hour than the text that caused it, and it needs a timezone to place each text in the homeowner's own day, so it covers the 1,484 conversations of 1,500 whose number we could resolve (17,293 of the 17,350 texts). The message table in chapter three is stricter: a text counts as answered only if the homeowner replied before we sent again, across all 17,350 texts. Never read a number from one as if it came from the other.
Where the sample was thin we marked it rather than dressing it up. Every cell under 200 texts carries a °.
Phone calls are excluded. Those same leads also took 12,182 calls. Nothing in here says anything about what happens on the phone.
Do this week
Five things worth changing
- Ask about timing, wait for the obstacle, then put a concrete time back on the table.
- Keep following up past the third. It is where your cost per response halves.
- Delete the emoji and the word "free" from your templates.
- Aim for 60 to 100 characters and end it with a question. Shorter than 60 does worse, not better.
- Move volume to Friday morning and off Wednesday afternoon.
- Try a week of evening sends. It is our joint-best block and our smallest.