Lead Follow-Up Automation Examples: 9 We Run, With the Numbers
Quick answer. The lead follow-up automations worth copying: deliver what they asked for in seconds, ping the owner with where the lead came from, ask one question, send a Day 3 proof email and a Day 7 close, then stop on any reply. Ours sends the kit in a median 3 seconds. Three emails, then silence.
Which one are you? Same nine. Different next step.
- Setting it up yourself. Start with the timeline and the table, then steal the subject lines in each entry. They are the real ones.
- Want it built into your business. Read what it costs, then send me what happens after someone fills your form. I reply within 24 hours.
- Agency setting this up for clients. We build it white-label, you keep the client. Send the brief.
Most lead follow-up advice is a template pack. This is the other thing: nine automations running on dappasol.com right now, with what each one logged between July and November 2026, the copy it actually sends, and the part that broke.
The setup is boring on purpose. A form posts to one PHP file. That file sends the kit, pings my phone on Telegram and emails me a copy. A job on the server runs every morning, sends the Day 3 and Day 7 emails, checks consent and reads a stop list. No CRM, no Zapier. The same nine work in HubSpot, Make or n8n, because the timing and the stop rules are the product, not the tool.
How we measured: 43 real leads from 13 July to 11 November 2026 (19 test submissions removed), 110 rows in the follow-up ledger, the mail server's transport log from 18 August, and the scheduler's run log. Every number below is a count or a median from those files. No names, no addresses, aggregates only.
Free, and from a human
Send me your follow-up and I'll tell you where it leaks.
Paste what happens after someone fills your form, or forward the emails your system sends. I read it myself and reply within 24 hours with the leak and the fix. Free. If it's already good, I'll say so.
Something went wrong. WhatsApp us instead?
Got it. If you gave an email, a confirmation is on its way, and I reply within 24 hours. Paste your follow-up into that reply and I'll start there.
Not sure follow-up is even the job to automate first? Run your numbers through the Automation ROI Calculator: What to Automate First.
The whole sequence, Day 1 to Day 7
Everything a kit lead can receive from us, on one line, with what stops it underneath. Read the stop rules first: they are the part most sequences forget.
Someone who writes to us instead of downloading something gets a different path: one acknowledgement with one question, then a human. That is automation 3 below.
The nine at a glance
| Automation | Trigger | Timing | Logged | Stop rule | Human step left | What broke |
|---|---|---|---|---|---|---|
| Instant kit email | Kit form submitted | At capture, median 3 s, slowest 20 s | 39 people; 34 of 35 sent at capture landed first try | One send, resent only if it failed | None | 30 Oct permissions bug, resent 8 h 21 min later |
| Owner alert | Any form | Same request, Telegram first, email copy median 2 s | One per lead since July | None, it is for the owner | Decide who gets a personal reply | Email copy bounced for almost four weeks; Telegram did not |
| One-question acknowledgement | Contact, reach-out, URL check, wizard | At capture | 18 sent, 4 variants, 3 to real messages | One email, no sequence | Read the answer, reply in 24 h | Two forms fell into a catch-all before 8 Sep |
| Day 3 proof email | 2 days after Day 1 | Due day 3, logged median 2.6 days after Day 1 | 23 came due: 11 sent, 12 withheld | Reply, “no”, unsubscribe, no consent | Answer WhatsApp replies | Laptop scheduler: 17 runs in October |
| Day 7 close | 4 days after the actual Day 3 | Logged median 5.8 days after Day 3 | 8 came due: 7 sent, 1 withheld | Last email, then silence | Answer one-word replies | Late Day 3 pushed Day 7; the gap never shrank |
| Consent gate | Any Day 3 or Day 7 due | Every run | 13 of 31 marketing emails withheld | Unclear counts as no | None | Nothing broke; it costs volume by design |
| Stop on reply | Reply, “no” or unsubscribe | Next run reads the list | 29 of 42 lead addresses on the list | Permanent, every stage | Add a replier to the list | Depends on someone reading the inbox first |
| No-show recovery | A booked call nobody joins | Touch 2 planned at 3 days, logged at about 5 | Every touch logged | Any reply, or after the breakup | Write touch 1 from a real look | Touch 2 sent two days late |
| Catch-up and redelivery | Daily server job | 09:00 IST, backup 20:00 IST, 20 h throttle | 59 runs since 31 Aug, 4 backfills, 2 redeliveries | 3 failed tries; one email per lead per run | Read the run log | 17 runs in October; then the throttle ate 4 Nov |
The nine automations, one by one
Same parts every time: trigger, timing, stop rule, the human step left, the number, the real copy, what broke. Then the same pattern for a dentist, a contractor, a clinic or an agency.
1. The kit lands before they close the tab
- Trigger
- Someone asks for a kit
- Timing
- At capture: median 3 seconds, slowest 20
- Stop rule
- One send. The daily job resends only if this one failed
- Human step left
- None
- Logged
- 39 people asked for a kit. 35 got it at capture, 34 on the first try. 4 earlier sign-ups were backfilled
Speed to lead starts with the thing they asked for, not with a sales call. The kit email goes out inside the same web request that saved the form, so there is no queue to fall behind. Across the 35 kit requests we could match against the mail server's log since 18 August, the median gap between the form and an accepted email was 3 seconds. The slowest took 20.
The case for speed is old. Harvard Business Review's study of online lead response put it bluntly: “most companies are not responding nearly fast enough” (Oldroyd, McElheran and Elkington, HBR, March 2011). Fifteen years later the fix costs nothing. Send the thing in the same request.
It comes from a person's address, signed with a name, in plain text. No logo header, no tracking pixel. It reads like a person sent it because a person wrote it, once.
One person submitted the same form twice, 8 seconds apart, on 8 August, and got the kit twice. Harmless, slightly silly. The form now hides itself after the first successful send.
Subjectyour automation opportunity finder
Fill in the hours column first, and be honest. The job you hate is rarely the job that costs you the most time.
Want a second opinion? I'll give you 15 minutes.
You asked for the kit, so I sent it. Want nothing else from me? Reply “no” and that's that.
Lowercase subject on purpose. Every kit lane has its own subject and its own “one move” line, so nobody gets the generic version.
What broke: On 30 October a server permissions change stopped the site writing files. One Day 1 email failed at 01:21 UTC. It went out at 09:42 the same day, 8 hours 21 minutes late, and the person got a personal note on top. Guard: after any server change, submit a real test lead as the web server's own user, not yours, and watch it arrive.
Same pattern elsewhere: A dentist's new-patient form sends the forms pack and the parking map. A contractor's quote request sends the price sheet and three photos of recent jobs. A clinic's intake form sends what to bring and how to cancel. An agency's brief form sends the process doc and one past case. All of it in seconds.
2. The owner alert, with where the lead came from
- Trigger
- Every form, every tool
- Timing
- Same request. Telegram first, email copy median 2 seconds
- Stop rule
- None. It is for me, not the lead
- Human step left
- Deciding who gets a personal reply, within 24 hours
- Logged
- One alert per lead since July
A lead nobody sees is a lead you didn't get. The alert goes to Telegram first because a bot message skips the whole email stack: no SPF, no spam folder, no inbox I forgot to check. It is one HTTPS call to the Telegram Bot API. A Slack incoming webhook is the same pattern. The email copy follows for the record.
The line that matters is Via. It names the search engine or assistant that sent them and the page they landed on first. That is how I know 22 of our 43 real leads came from Google, 12 arrived direct or with the referrer hidden, 2 came from ChatGPT and 1 each from Gemini, Yahoo and Meta AI. Without it, every lead is from “the website”.
Telegram, to my phoneReal format, sample data
New lead (magnet:automation) Email: name@example.com Details: page=/guides/what-to-automate-first-in-a-small-business/ wa= marketing=yes Via: Google -> /guides/what-to-automate-first-in-a-small-business/ Form on: /guides/what-to-automate-first-in-a-small-business/ IP: 203.0.113.24
What broke: From 30 August to 24 September the email copy went to an address on our domain that did not exist. Every alert email in that window bounced. The Telegram ping does not use email at all, so it kept arriving. Guard: alerts go to two channels that fail differently, and after any address change you send a test to every address and read it arrive.
Same pattern elsewhere: Dentist: the front desk gets a ping naming the treatment asked about. Contractor: the owner's phone gets the job type and postcode, so the nearest crew calls. Clinic: the ping says new patient or existing, so the right person picks it up. Agency: the ping names the page, so you know if they came from pricing or a blog post.
3. The acknowledgement that asks one question
- Trigger
- Contact form, reach-out box, URL check, project wizard
- Timing
- At capture
- Stop rule
- One email. Never a drip
- Human step left
- Reading the answer and replying within 24 hours
- Logged
- 18 sent across 4 variants since 18 August
These people wrote to us, so they get a reply, not a drip. The automation does one job: prove a human got it, and ask the question that makes the human reply useful. The URL check even says it hasn't opened their site yet, which is true, and saves an awkward email later.
18 acknowledgements since 18 August. Only 3 answered real contact-form messages. The rest went to our own test submissions and to URL-check requests. Small volume, same code path, and every variant has fired at least once.
Subject{their site}: one question before I look
Got it. {their site} is in front of me. I haven't opened it yet, so nothing in this email is a price or a plan yet.
What's the one number this site has to move?
SubjectYou asked us to reach out, one question first
What are you trying to get done, and by when?
SubjectGot your message, {first name}
If there's a link to what you're working on, a live site, an app, anything, send it.
Three of the four variants. The first name is only used when it looks like a real first name.
What broke: Before 8 September two of our forms posted fields the lead handler didn't recognise. Both fell through to a catch-all while the page told the visitor a reply was on its way. Nobody got one. Guard: every form names itself, and anything unrecognised writes a “gap” row to the ledger. One gap row so far: a cost-estimator lead with no sequence, caught because the ledger said so.
Same pattern elsewhere: Dentist: “Is this for you or a family member?” Contractor: “Is this a repair or a replacement, and what's the address?” Clinic: “First visit or follow-up?” Agency: “What's the launch date you can't move?” One question, the one your reply depends on.
4. Day 3: proof, not a nudge
- Trigger
- Daily job finds a lead 2 days past Day 1
- Timing
- Due on day 3. Logged median 2.6 days after Day 1 (2.1 to 4.6)
- Stop rule
- Reply, “no”, unsubscribe or no consent
- Human step left
- Answering the WhatsApp replies
- Logged
- 23 came due for Day 3: 11 sent, 12 withheld for no consent
Day 3 is where most sequences say “just checking in”. Ours shows something: two live sites we built and one instruction. No “did you get a chance”. The ask is a WhatsApp message with their URL, and a human answers it.
Subjectthe $3,000 version of the sites in that kit
Quick one. Two of the sites in the kit are ours.
Open either on your phone. That is the whole pitch.
Subjectwhat the audit catches that the checklist can't
The checklist covers what you can see.
Subjectfixed price, in writing
Send me the one-paragraph version of your idea on WhatsApp.
Three of the five lanes that have Day 3 copy.
What broke: The scheduler lived on a laptop until 3 November. October had 17 runs in 31 days, with a 6-day hole from 30 September to 6 October, and every email due in that hole went out late. Guard: the job now runs on the server at 09:00 IST, the laptop ping at 20:00 IST is only the backup, and the scheduler catches up missed days instead of skipping them.
Same pattern elsewhere: Dentist: one before-and-after from a real patient who agreed to share it. Contractor: a 30-second video of a finished job in their area. Clinic: one outcome number from your own practice. Agency: one case with one number. Proof, then one way to reply.
5. Day 7: the close that lets them go
- Trigger
- 4 days after Day 3 actually went out
- Timing
- Logged median 5.8 days after Day 3 (4.0 to 5.8)
- Stop rule
- Last email. Then nothing, ever, unless they write
- Human step left
- Answering one-word replies
- Logged
- 8 came due for Day 7: 7 sent, 1 withheld
A good close does two things. It makes the offer once more, plainly, and it says this is the last email. Saying it is a promise, and the engine keeps it: there is no Day 14 in the code.
Subjectclosing the loop
Last one from me on this. If a cinematic site is a someday thing, keep the kit, no hard feelings.
Subjectlast nudge, then I leave your app alone
If the app is a side project, the checklist is enough, genuinely.
Subjectstill building it?
If the Mini App got shelved, ignore me.
Each one gives them permission to say no. That is the point.
What broke: Day 7 inherits Day 3's lateness. When Day 3 ran late, Day 7 did too, which is why the logged gap is 5.8 days, not 4. Guard, built in: Day 7 counts from the day Day 3 actually went out, not from signup, so a late run can never stack two emails a day apart.
Same pattern elsewhere: Dentist: “If now's not the time, here's the number for when it is.” Contractor: “If the job's on hold, the quote stands for 30 days.” Clinic: “Your referral stays valid, book when you're ready.” Agency: “If the project slipped, keep the plan. Last email from me.”
Have a question about your own follow-up? You don't have to book a call. Message me. I reply within 24 hours.
Something went wrong. WhatsApp us instead?
Got it. If you gave an email, a confirmation is already on its way and Ishan replies within 24 hours.
6. The consent gate: fail closed
- Trigger
- Any Day 3 or Day 7 coming due
- Timing
- Checked on every run
- Stop rule
- Anything not clearly yes counts as no
- Human step left
- None. The lead made the choice
- Logged
- 13 of 31 marketing emails withheld (12 Day 3, 1 Day 7)
The kit email is transactional: they asked for it. Day 3 and Day 7 are marketing, so they go only to people who ticked the box. If the consent field is blank, misspelled or from an older form, it counts as no. That cost us 13 of 31 marketing emails. Fine.
Gmail's sender guidelines ask senders to keep the spam rate under 0.3%, and bulk senders to offer one-click unsubscribe on marketing mail. We are nowhere near bulk. The cheapest way to stay under any spam threshold is not emailing people who never said yes.
What broke: Nothing, by design, but it is why the Day 3 numbers look small: older forms never posted a consent field at all. Guard: treat unknown as no, and log every skip, so a quiet sequence is visibly quiet, not secretly broken. This is how we run it, not legal advice.
Same pattern elsewhere: Dentist: the recall reminder is service, the whitening promo is marketing. Contractor: the visit confirmation is service, the spring offer is marketing. Clinic: appointment details are service, the new-clinic launch is marketing. Agency: the proposal is service, the newsletter is marketing. Gate the second kind.
7. Stop on reply, with one list for every stage
- Trigger
- Any reply, any “no”, any unsubscribe
- Timing
- The next run reads the list
- Stop rule
- Permanent, across every stage and every lane
- Human step left
- Adding a replier to the list. The one manual duty
- Logged
- 29 of 42 lead addresses are on the stop list
Replies always beat the sequence. The moment someone writes back, the automation's job is done and a person's starts. Our stop list is a plain text file. A reply goes on it by hand, the unsubscribe link adds itself, and every send checks it first.
29 of 42 lead addresses are on it now. That isn't a failure rate. It means either the person or a human ended the sequence, which is the sequence doing its last job.
Subjectevery marketing email, last line
If you'd rather not get emails from me, reply “no” and that's that.
Plus a one-click unsubscribe link in the headers, signed so nobody can unsubscribe someone else.
What broke: Nothing has broken yet, but this is the weak joint: the stop depends on a person reading the inbox before the next 09:00 run. Guard: suppress first, reply second. Reply first and get distracted, and Day 3 lands on top of your personal note, looking exactly as automated as it is.
Same pattern elsewhere: Dentist: a patient who calls to book drops off the reminder sequence. Contractor: a homeowner who replies with a date drops off the quote chasers. Clinic: a patient who cancels by text drops off the confirmations. Agency: a prospect who replies drops off the drip, and a person takes over.
8. No-show recovery, started by a human
- Trigger
- A booked call nobody joins
- Timing
- Touch 2 planned 3 days after touch 1. Logged at about 5, see below
- Stop rule
- Any reply, or after the breakup
- Human step left
- Writing touch 1 from a real look at their business
- Logged
- Every touch sent and logged by the same engine as the rest
A no-show is not a no. Usually it's a calendar. The worst follow-up is “sorry we missed you, rebook here”. The one worth sending is proof you looked: something specific on their site, worth two minutes of their time.
This is the only entry a human starts, because touch 1 needs a human to look. After that the engine sends and logs it like everything else.
SubjectFound something on {their business} worth 2 minutes
SubjectDo you know what you keep per {unit}?
Touch 2: one question about their own margins.
SubjectYour {trade} website, and an apology
The version for when the miss was ours.
Templated here. In the real ones the braces hold their business name and their product.
What broke: Touch 2 went out two days late. The staged script had the wrong home directory typed into it, so it failed until someone noticed. Guard: every scheduled send gets a dry run before it is live.
Same pattern elsewhere: Dentist: “We held your slot for 15 minutes. Here are two times this week.” Contractor: a photo of a similar finished job and two visit slots. Clinic: “Your referral is still valid until {date}.” Agency: one finding from their site and a link to rebook.
9. The catch-up job: late, never lost
- Trigger
- Server cron at 09:00 IST, laptop backup at 20:00 IST
- Timing
- Daily, throttled to one real run per 20 hours
- Stop rule
- 3 failed tries per address; one email per lead per run
- Human step left
- Reading the run log
- Logged
- 59 runs since 31 August, 21 duplicate pings throttled, 4 kits backfilled, 2 redelivered
This one makes the other eight trustworthy. It sends nothing clever. It notices what should have gone out and didn't.
When the engine went live on 5 August it backfilled 4 kit emails for people who had signed up before it existed. Before 27 August our mail failed authentication; two days after moving to authenticated SMTP the job re-sent the two most recent kits.
What broke: October: 17 runs in 31 days, because the trigger lived on a laptop. Then the fix ate a day: the new server run came too soon after the laptop backup and was skipped. Nothing ran on 4 November. Guard: one primary schedule on the server, the backup later in the day, and a scheduler that catches up, so a missed day is late, never lost. November so far: 11 runs in 12 days.
Same pattern elsewhere: Any business with any of the eight above needs this one. Dentist, contractor, clinic or agency, the question is the same every morning: what should have gone out yesterday, and did it?
What 110 ledger rows taught us
- Speed is solved. Attention isn't.The kit lands in a median 3 seconds and the alert reaches a phone in the same request. What decides whether a lead goes anywhere is the human step after the ping, and no automation does that part.
- Consent cuts the sequence by about 40%.13 of 31 marketing emails were withheld. Write Day 3 and Day 7 for the people who said yes, and make the Day 1 email good enough to stand alone for everyone who didn't.
- Schedules drift, so measure the gap you ship.The rule said 2 days and 4 days. The log said 2.6 and 5.8. Nobody would have known without timestamps on both ends.
- Copy is what's missing, not code.Only 5 of our 13 kit lanes have Day 3 and Day 7 copy. The other 8, including the opportunity-finder kit, get the kit email and then silence. The engine handles any lane. Nobody has written the emails.
- Two channels that fail differently beat one good one.The alert email bounced for almost four weeks. The Telegram ping kept arriving because it never touches email. Same logic for the scheduler: server first, laptop second.
- A stop list is a feature.29 of 42 leads are on it. A sequence that never stops anyone is just a newsletter nobody asked for.
How many follow-ups before you stop?
Three, for someone who downloaded something. The thing, the proof, the goodbye: Day 1, Day 3, Day 7. Each gap is longer than the last. Someone who downloads a kit is telling you something: interested, not ready to talk to a human yet. Three emails respects that. A fourth is a newsletter they never signed up for.
One, for someone who wrote to you. An acknowledgement with one question, then a human. Putting a person who typed a message into a drip is the fastest way to look like a machine.
Zero more, after any reply. A reply, a “no”, an unsubscribe: the sequence is over, on every stage, for good.
The count matters less than the stop rules. If you want a longer sequence, earn it: ask for the newsletter separately, with its own box, and keep the follow-up at three.
What it costs
Ours runs on the server we already pay for: one PHP file for capture, one scheduled job, a Telegram bot with no fee, and authenticated SMTP on a mailbox we already had. The real cost was the build time and the copy. What agencies charge for the same thing is in what business automation costs and AI automation agency pricing.
Before paying anyone, check follow-up is even the right first job. The Automation ROI Calculator: What to Automate First ranks your manual jobs by hours back, so you fix the biggest leak first, which might not be this one.
If you'd rather not build it yourself, there are two ways to work with me:
- AI Game Plan, from $500, credited 100% to the build. A week mapping where follow-up and the rest of your admin leak time, with each fix priced. We find you a day a week to claw back, or the Game Plan is free.
- The Residency, from $8,000, build included. I embed for 3 to 5 weeks and build the follow-up, the alerts and the rest of your ops automation inside your own accounts, myself.
More on how the build side works: business automation. Or start smaller: show me your Monday and I'll walk through your week with you, free, in 20 minutes.
Keep going
- Small business automation examples: the wider set, beyond follow-up
- Build log: running DappaSol on our own automations: the log these nine came out of
- What to automate first in a small business: if follow-up isn't your biggest leak
- Automation ROI Calculator: What to Automate First: hours back per job, ranked
- Business automation cost in 2026: what the market charges for this
- AI automation agency pricing: what sits inside each tier
- Business automation: built into the tools you already pay for
- Show me your Monday: a free 20-minute walkthrough of your week
FAQ
What is lead follow-up automation?
Software that answers a new lead for you: it sends what they asked for, tells you they arrived, and sends a short, timed sequence until they reply or the sequence ends. Ours sends the kit in a median 3 seconds, alerts the owner on Telegram in the same request, and stops after three emails.
How fast should you follow up with a new lead?
Seconds for the automatic part, under 24 hours for the human part. Our kit email goes out in a median 3 seconds and the owner alert in the same web request. Harvard Business Review's 2011 study of online lead response found most companies were not responding nearly fast enough, and that was before instant email was free.
What should the follow-up email after a lead magnet download say?
Deliver the file, give one thing to do with it, offer a human, and give an easy out. Ours: the link, one move (“Fill in the hours column first, and be honest”), a 15-minute second opinion, and “Reply no and that's that.” Plain text, from a named person, no logo header.
How many follow-up emails before you stop?
Three for a download: Day 1, Day 3 and Day 7, and the last one says it is the last. One acknowledgement for someone who wrote to you, then a human. None after any reply, any no, or an unsubscribe. 29 of our 42 lead addresses are on the stop list.
How do I set up a Telegram or Slack alert for new leads?
Have the form handler call the Telegram Bot API's sendMessage method, or a Slack incoming webhook, in the same request that saves the lead. Include where the lead came from and the page they landed on. Send it before the email copy, with a short timeout, so a slow chat service never delays the lead's own email.
What should a no-show follow-up email say?
Something specific you found on their business, plus a way to rebook. Not “sorry we missed you”. Ours is built as three touches: a finding (“Found something on {their business} worth 2 minutes”), one question about their numbers, then a breakup that gives the fix away. Any reply stops it.
Do I need consent to send automated follow-up emails?
Sending what someone asked for is transactional. Promotional follow-ups are marketing, and the rules depend on where your leads live. We send Day 3 and Day 7 only to people who ticked a marketing box and treat anything unclear as no, which withheld 13 of 31 marketing emails. Gmail also asks bulk senders for one-click unsubscribe. This is how we run it, not legal advice.
What does lead follow-up automation cost?
The tools can cost almost nothing: ours runs on an existing server, a Telegram bot with no fee and an existing mailbox. The cost is setup and copy. Done for you, the AI Game Plan starts at $500 and is credited 100% to the build, and the Residency, where I build it inside your business, starts at $8,000.
Have a question about your own follow-up? You don't have to book a call. Message me. I reply within 24 hours.
Something went wrong. WhatsApp us instead?
Got it. If you gave an email, a confirmation is already on its way and Ishan replies within 24 hours.