How to write SMS outreach that converts

Not every business text is trying to get a reply. A promotional broadcast can pay for itself in clicks without a single response, and a delivery notification needs no conversation at all. A follow-up text should be personal, and start or continue a conversation with the person who enquired. Here are the six ways to do it.

1. Be specific. Name the actual job, the address, the quote. In a study of over 1,000 real customer conversations, more concrete wording raised satisfaction by about 9% and actual spending by at least 13% per standard deviation of concreteness, because customers took it as proof the employee listened.^[1] A first name on its own does not do this. In a randomized SMS trial, adding the recipient's name moved response rates only marginally, and pooled with earlier trials the effect disappeared.^[2]
2. Ask one easy question. End the message with one thing the reader can answer from the lock screen. In studies of live conversations, people who asked more questions, especially follow-up questions about what the other person just said, were better liked and rated more responsive.^[3] No study proves a one-question limit for sales SMS. The reason to hold to it is practical, as two questions ask for a longer reply, and longer replies tend to get put off.
3. Cut what the first reply does not need. Company history, links and offers all have a place, and it is after the lead has replied. A link in an opener reads as marketing and gives the reader something to do other than reply. Carriers do not filter links as a class. The known trigger is public link shorteners, which US carriers treat as a spam signal, while links on your own branded domain are normal in business SMS.^[4]
4. Send it while the enquiry is fresh. A 2007 study of web-generated leads found the odds of qualifying a lead were about 21 times higher when the first attempt came at five minutes rather than 30.^[5] It measured phone calls, so it supports replying while the lead is still thinking about the problem rather than proving a five-minute text multiplies anything by 21. The average business is nowhere near either mark. Harvard Business Review's audit of 2,241 companies clocked the average first response to a web lead at 42 hours.^[6]
5. Let the answer decide the next message. A sequence that sends step three no matter what the lead said at step two is still a broadcast. The question-asking research singles out follow-up questions, the ones built on what the person just said, as the kind that made askers most liked.^[3]
6. Keep compliance out of the copywriting. Consent, sender identification, opt-out handling and carrier registration are legal and carrier requirements, covered at the end of this page. They apply regardless of what the template says.

For contrast, the typical blast:

Hi! Great news from {company}! For a LIMITED TIME we are offering 20% off all services. Click here to claim your discount today!

Promotions typically aim to drive clicks. Conversational SMS is more effective when selling high-ticket services, where the purchasing decision requires a tailored approach.

About the examples below. Each template shows the conversational copy only. Depending on the destination, production messages also need sender identification, contact details and opt-out wording; Twilio's messaging policy, for example, requires the first message to a recipient to include "Reply STOP to unsubscribe" or an equivalent standard phrase.^[8] That line does not have to sound like a terms page. Vendita adds it as its own line under whichever opener fits the lead source:

Also, if you'd prefer not to hear from me, reply STOP to opt out.

It appears on the first message only, and Vendita attaches it automatically.

The reply is written before the enquiry arrives, which is what lets it go out in seconds instead of whenever someone gets to it. What those seconds are worth is measured on the speed-to-lead statistics page. One template per lead source, each naming the exact action the lead just took:

Web form:

Hey {first name}, it's Mia at {company}. Your request about the {job} just came in. What's happening with it at the moment?

Ad lead:

Hey {first name}, Mia at {company} here. You asked about {job} through our ad. Are you hoping to get it taken care of this month, or planning ahead?

Portal inquiry (real estate):

Hi {first name}, it's Mia at {company} about {address}. Did you want to walk through in person, or are you just starting to look?

Missed call text-back:

Hey, it's {company}. Sorry we missed your call just now. What can we do for you?

Each message proves a person saw the specific thing the lead did, then asks a question that sorts the lead, urgent job or early research, and the answer reveals the next question to ask. A lead who reads "the ceiling stain" knows a person saw their form; "your enquiry" could have been sent to anyone. In the field data that difference carried through to satisfaction and actual spending, because customers take concrete wording as proof they were listened to.^[1] An HVAC lead texting during a heatwave wants an honest arrival window more than a discount, and the opener's only job is to start that dialogue while attention is still on the problem.

A quote that didn't purchase is rarely a no. In roofing it is usually sticker shock or a hope that the job can wait a couple of years; in solar it is payback math or waiting on a rebate decision; in HVAC it is a comparison between three quotes that all recommend different things. "Just following up on the quote" only tells the lead you want something from them. Naming the likely worry shows them you understood why they stopped replying.

The first pattern is the label, from Chris Voss's Never Split the Difference: open with seems, sounds, or looks like, name the worry, and drop the word "I":

Hey {first name}, it's Mia at {company}. Seems like the quote for the {job} might have landed higher than expected?

The second is the no-oriented easy-no question. It is less committal to say no than to say yes, which naturally gives a no-based ask less pressure, and "no, still interested" is a common reply:

{first name}, have you given up on the {job} for this year?

Vertical variants, each built around the worry that actually stalls that sale:

Roofing variant:

Hey {first name}, Mia at {company}. No push on the reroof quote. Would it help if we split out a repair-now option so the full job can wait?

Solar variant:

Hey {first name}, it's Mia at {company}. No rush on the solar quote. Would a payback estimate built from your actual quarterly bill make the decision easier?

HVAC variant:

Hey {first name}, Mia at {company}. Totally understand if the quote is parked. Is the back bedroom still running hotter than the rest of the house?

One caution: the message after their reply must hand them something they can use, an option, an estimate, a straight answer, never a pivot straight to a pitch. A conversational opener followed by a sales pitch reads as bait-and-switch, and the lead stops replying for good.

A phone call gives a salesperson tone, pace and the ability to keep talking when the customer hesitates. SMS gives none of that. Every message stands alone, the lead can ignore any of them for free, and anything over a few lines reads as pressure. That rules out rebuttals, stacked questions and paragraphs of reassurance. What works inside those constraints is a two-move framework, both moves from Chris Voss's negotiation work.

Label, then ask. Open with seems, sounds or looks like, name the worry in their words, drop the word "I", and end with one question. The label proves you read what they said; the question keeps the conversation moving without pushing.

If, then ask. When they push back, do not argue the point. Raise a hypothetical that changes the situation, then ask a question that is easy to answer: if the scope were smaller, if you could look at it together on an evening, would you be opposed to revisiting it? "Would you be opposed" invites a no, and a no to that question is the lead agreeing to look again.

One objection, worked through. The lead says: "I'd need to speak to my husband first."

Sounds like it's definitely a joint effort for you guys to make a decision, is it something you've ever looked into properly together?

The reply branches, and each branch gets its own label and an easy question:

"No, it's just our schedule."

OK great, well sounds like you just need a time to look at it properly, is it a terrible idea to look at it on an evening when you're both home?

"It's not been top of his to-do list."

OK sounds like a few other things have got in the way, would you be opposed to seeing the full scope so you can see whether it's worth doing sooner rather than later?

"Yes, but the last salesman was pushy."

Sounds like you had quite an uncomfortable experience, would it be worth a quick comparison where we can help answer any questions you have, no obligations?

These are negotiation techniques, not SMS research; treat them as strong defaults and test them against your own reply and booking rates.

The leads who inquired six months ago and never bought are usually the largest list a business owns, and the easiest to burn. Working through that list is called database reactivation, and it starts with the opener:

Hi {first name}, it's Mia from {company}. You spoke to my colleague {colleague} about the {job} a while back. Have they been in touch about {x} yet?

The named sender and company make the message identifiable, which the compliance rules below require. The colleague reference is personal and verifiable, because it really happened. And the question puts the follow-up on your side, so the customer never has to explain why they went quiet. Never ask whether they went ahead with the job or got it sorted, as it hands them an easy way to close the conversation. "Has he been in touch?" makes "no" the easy answer, and a no here simply means it is your turn to follow up.

HVAC variant:

Hi {first name}, it's Mia from {company}. You spoke to my colleague Sean about the upstairs AC last year. Has he been in touch about getting it looked at before summer?

Solar variant:

Hi {first name}, it's Mia from {company}. You spoke to my colleague Sean about solar for the roof a while back. Has he been in touch about the new rebate changes yet?

Roofing variant:

Hi {first name}, it's Mia from {company}. You spoke to my colleague Sean about the roof last spring. Has he been in touch since the storms came through?

Real-estate variant:

Hi {first name}, it's Mia from {company}. You spoke to my colleague Sara about selling in {suburb}. Has she been in touch about how the market's moved since?

What comes after the opener runs on one track. "Have they been in touch?" gets a yes, or an instant no from anyone who no longer needs the work. A positive or ambiguous reply gets the next question: did the {job} end up getting done? A realistic share of an old quote list quietly had the work done elsewhere, and this question finds them; for everyone else, the answer usually names what stopped them. Whatever stopped them then gets related to something new: a new offer, stock clearance, a new season, new legislation, a price increase coming. Then a soft ask: would it hurt to have a look again?

Do not send promotions to an old list. A "limited time offer" from a business the lead has not heard from in a year reads as spam, and enough spam reports will typically get the number filtered. And do not text contacts who never agreed to hear from you; old lists are where the compliance rules below apply hardest. In real estate, most sellers who asked for an appraisal said there was no hurry, and usually they meant it. The re-ask carries no pressure, so it still works months or years later, when they finally sell.

Across the meta-analysis's 13 randomized trials, the pooled odds ratio for attendance was 1.62.^[9] The research comes from healthcare, so reminders raising attendance is well established, although the best wording and timing for a sales appointment have not been studied.

What to actually write comes from sales practice. Alex Hormozi reports one funnel's show rate rising from 74% to 91% after three changes made at once: removing false-urgency copy, booking appointments for the same or next day, and replacing automated reminders with personal ones.

Two-option near-slot offer:

We can have Sam out {day/time A} or {day/time B}. Which works better for you?

Confirmation:

All set: {day/time A} at {address} with Sam from {company}. Anything he should know before he heads out?

Morning-of reminder:

Morning {first name}, Sam from {company} here. Looking forward to the {job} at {address} at {day/time A} today.

Reschedule:

No stress at all, {first name}. The next spots are {day/time A} or {day/time B}. Which works better?

No-show recovery:

Hey {first name}, looks like we missed each other today. Want me to hold the same time tomorrow instead?

The two-option offer turns the booking into choosing a time, so the lead never has to decide whether they want an appointment at all. A same-day or next-day time gives a lead less time to reconsider, double-book or forget. The reminder reads human because it carries specifics no blast would bother with, and the recovery treats the miss as a scheduling problem rather than something to apologize for, so replying stays easy. If a lead never confirms, act rather than wait. Offer the same time tomorrow and give the near slot to someone responsive. How an agent books inside the conversation, and what that replaces, is shown on the appointment booking page.

Five findings, each from published research on real conversations:

1. Warmth belongs at the start and end, substance in the middle. Analysis of 23,958 moments from live service conversations, plus four experiments, found warmer language raised satisfaction at the open and close of a conversation, and problem-solving language mattered most in the middle. The same research found questions helped satisfaction except in the very opening moments of a conversation.^[10]
2. "I" beats "we." Across five studies, customers were more satisfied and bought more when the person helping them said "I'll take care of that" rather than "we'll take care of that," because "I" tells the customer one specific person has taken their problem on. Wording built around "you" made little difference and sometimes backfired.^[11]
3. A first name is not personalization. A randomized trial of reminder texts found adding the recipient's name moved response from 84% to 90%, and pooling it with earlier trials the effect disappeared.^[2] Adding a name is easy, although on its own it does not really move anything. What personalizes a text is the job, the address, the original conversation.
4. Restating what the customer said makes a sender feel human. Four studies of conversational AI found that taking proper turns, moving the conversation forward, and briefly rephrasing what the customer just said made an AI interface feel more human, and lifted recommendation acceptance, willingness to pay a premium, and loyalty.^[12] This is how Vendita's agent is built to converse. It acknowledges what the lead said, then asks the next question.
5. An instant reply is what makes automation acceptable. Across six studies, identical service was rated worse when customers believed a bot delivered it, because they read automation as the business cutting costs at their expense. The result reversed when the bot's advantage was unmistakable. An identified chatbot replying instantly scored 5.92 on service evaluation, against 3.82 for a human replying on roughly a minute's delay and 2.87 for the same bot slowed to that pace.^[7] Faking typing delays isn't a recommended strategy, as it removes speed and convenience as an advantage for the consumer.

The short version of what the law requires: consent you can prove, accurate sender identification, a working opt-out, and marketing hours of 8 am to 9 pm in the US.^[13][15][17] The carriers add their own rules on top. US business traffic needs A2P 10DLC registration or it is blocked,^[14] and since 1 July 2026 an unregistered brand name on an Australian text is replaced with the word "Unverified" on the customer's phone.^[16]

Vendita handles all of it before a message leaves the platform: numbers and sender IDs registered, identification and opt-out wording attached, STOP honored instantly, a conversational "don't text me again" treated the same way, and every send paced inside quiet hours. This is general information, not legal advice.

Reply rate moves easily, and for bad reasons as well as good. A provocative opener lifts replies and opt-outs together.

When you test, change one thing at a time: the opener, the question, the send time, or the follow-up gap. Change two at once and you cannot tell which one moved the number.

Volume is required to show a statistically relevant difference between two results. Below these volumes, a "winner" is usually luck. For a campaign replying at around 30%:

If the maths behind those volumes is something you genuinely care about, they come from one equation, at 95% confidence:

Contacts per variant ≈ 16 × rate × (1 − rate) ÷ difference²

• rate × (1 − rate) measures how much the number jumps around on its own. A rate near 50% jumps the most, so it needs the most volume; a rate near 0% or 100% barely moves.
• ÷ difference² is there because a small gap is harder to tell from luck than a big one. Halve the gap you want to trust and the volume needed quadruples.
• × 16 is the safety margin that holds the test to 95% confidence, so a false winner slips through in at most about one test in twenty.

Vendita campaigns run this natively. An A/B split sends each variant to part of the list, the results tab reports the chain per variant, and the platform warns you off declaring a winner while the sample is too small.