Salon client win-back automation: why 90 days misfires
A flat 90-day win-back rule chases loyal clients and misses lapsed ones. Build it on each client's own service interval instead, with honest offer advice.
What you get from this guide
- Why a flat 90-day rule chases loyal clients and misses lapsed ones
- How to compute each client's own visit interval, and what to do when the history is thin
- A lapsed definition built on a multiple of that interval, with a floor and a ceiling
- Three touches of real copy, and an honest answer on whether to discount
- How to measure it against a control group instead of claiming every rebooking
Why a flat 90 days misfires
"Set up a campaign for clients who have not visited in 90 days" is the standard advice, and it is the reason most salon win-back campaigns underperform. The number is not wrong because it is 90. It is wrong because it is one number applied to a client base that does not share a rhythm.
Consider two clients. The first has a colour appointment every six weeks, has done for two years, and you could set your calendar by her. The second gets a cut roughly every twelve weeks and always has. Today both of them are ten weeks past their last visit.
| Client | Their normal interval | At ten weeks | What a 90-day rule does |
|---|---|---|---|
| Six-week colour client | 42 days, very consistent | Clearly overdue. Something changed - a new salon, a bad visit, a move. | Nothing yet. Waits until day 90, by which point she has been somewhere else twice. |
| Twelve-week cut client | 84 days, also consistent | Entirely normal. Probably books next week. | Messages her at day 90 to say you miss her, six days after a normal gap. |
The flat rule fails in both directions at once. It is far too slow for your most frequent - which is to say most valuable - clients, and it insults your infrequent ones. And the frequent client is the expensive miss: a six-week client who leaves costs you eight or nine visits a year, and by day ninety she has already found somewhere else and been twice.
The fix is not a better number. It is a different kind of rule: lapse is relative to the client's own rhythm. Everything else in this guide follows from that one change.
Working out each client's interval
The interval is the median number of days between a client's consecutive completed appointments. Median, not mean. One client goes away for three months every summer, and an average will decide her normal rhythm is eleven weeks when it is actually five.
- Three or more completed visits: use the client's own median gap. This is the number you want and it beats every heuristic.
- One or two visits: no reliable personal interval exists. Use the median interval for the service they last booked - colour, cut, nails, brows and lashes all have genuinely different natural rhythms.
- First-time client who never returned: this is not a win-back problem, it is a first-rebook problem. Treat it as a separate, shorter sequence with different copy.
- Clients whose interval has been visibly lengthening: the trend matters more than the median. Someone drifting from six weeks to eight to ten is leaving slowly, and that is the most recoverable client on your list.
Compute the service-level medians from your own data rather than importing someone else's rule of thumb. The gap between a root-touch-up rhythm and a balayage rhythm is real, but its size depends on your pricing, your clientele and your area, and you already have the data to measure it.
One practical note: compute intervals from completed appointments only. Cancellations and no-shows in the history will make a client look like they visit more often than they do, and you will start chasing people early. If no-shows are a significant slice of your book, that is its own problem with its own fix - see AI appointment reminders and the no-show rate entry.
Defining lapsed properly
With an interval per client, the lapsed rule becomes a multiple rather than a constant:
The lapse rule
interval = median gap between that client's
completed visits (or service median)
lapse_threshold = clamp(interval * 1.5,
min = 21 days,
max = 180 days)
lapsed = days_since_last_visit
> lapse_threshold
AND no future booking
AND not opted out
AND not excludedThe 1.5 multiple is a starting point, not a discovered constant. The reasoning is straightforward: at exactly one interval the client is merely due, and a meaningful share of them have already booked or are about to, so messaging then is noise. At two intervals they have visibly gone elsewhere and the message arrives as an afterthought. Halfway is early enough to catch a drift and late enough not to nag someone who is simply a week late.
The clamp matters as much as the multiple. Without a floor, a client with two visits eight days apart gets a twelve-day threshold and receives a "we miss you" text a fortnight after seeing you. Without a ceiling, an annual client generates a threshold well over a year out, by which point the exercise is pointless. Twenty-one days and six months are sensible bounds for most salons; adjust them to your services.
Then subtract the people who must never receive it. This exclusion list is not a nicety, it is the difference between a campaign that reads as caring and one that reads as automated:
- Anyone with a future appointment already booked. Obvious, and still the single most common mistake in these campaigns.
- Anyone who opted out of marketing, on any channel - see the opt-in and stop-handling rules, which are legal requirements and not preferences.
- Clients flagged after a complaint, a refund or a deliberate parting of ways. Your team knows who these are; give them a field to mark it.
- Clients whose stylist has left. These need a different message entirely, from a named person, ideally not automated at all.
- Staff, family, trade accounts and test records.
- Anyone who already received a win-back sequence in the last twelve months.
If you are not on a big platform
This guide is written for salons that are not on Boulevard, Vagaro or Fresha - and that is deliberate, because on those platforms your first move is to check what the software already does before building anything.
What we could and could not verify, August 2026
The reason to build anyway, even where a native feature exists, is the one this whole page is about: a weeks-since-last-visit setting is a flat rule. If your platform lets you set "email clients who have not visited in 12 weeks", you have the 90-day problem with different arithmetic. Per-client intervals need a computation the booking software is not doing for you.
If you are on a smaller booking system, a spreadsheet, or a diary and a card file, the build below works regardless, because it only needs one thing from your system: a list of completed appointments with client, service and date. Almost everything exports that. Where systems differ is what they let you write back, and if yours has no API at all, you can still run the analysis and hand the front desk a call list - which is frequently the better outcome for a salon anyway. Broader context for the sector is on the salon automation page.
The offer question, honestly
Every template you will find leads with a discount. Twenty percent off your next visit, we miss you. It works in the narrow sense that it produces bookings, and this is the part the templates leave out: discounting trains discount-seeking.
A client who disappears, receives twenty percent off, and comes back has learned something specific: going quiet is rewarded. Some of them will never pay full price again, and you have converted a full-price client into a discount client while congratulating yourself on a recovery. Worse, if you run the campaign on a schedule, your most price-sensitive clients will work out the pattern faster than you would like.
| Policy | What it does | When it is the right call |
|---|---|---|
| No offer | Asks a genuine question and makes rebooking easy. Recovers clients who simply drifted. | Default. Especially for high-value regulars, where price was almost certainly not the reason they left. |
| Value-add | A treatment, a longer consultation, a product sample. Costs you less than the equivalent discount and does not reset the price anchor. | Most situations where you want to offer something. This is the middle path and it is usually the right one. |
| Discount | Money off. Highest response, clearest cost, real behavioural side effect. | Final touch only, clearly one-off, time-boxed, and ideally not on your best clients. |
There is also a question the offer cannot answer. If a client left because of a bad cut, a stylist who was rushing, or a booking system that failed them, a discount is an insult with a number attached. That is why touch one below asks rather than offers: some share of your lapsed list left for a reason you would want to know, and the only way to find out is to make it easy to say.
The three-touch sequence and copy
Three touches, roughly three weeks, then stop. Timings are from the day the client crosses their own lapse threshold, not from a fixed date, so different clients enter on different days - which is exactly the point.
| Touch | Timing | Job of the message |
|---|---|---|
| 1 | Day 0 of lapse | A short, personal check-in from their stylist by name. No offer, no marketing language. Invites the honest answer. |
| 2 | Day 10 | A reason to come back that is not a discount: availability with their stylist, something new they would like, a seasonal hook. |
| 3 | Day 21 | Close-out. Say you will stop, make rebooking one tap, and place the offer here if you are using one at all. |
Touch 1 - day 0 (SMS, from the stylist)
Hi {first_name}, it is {stylist} at {salon}.
I noticed it has been a while since your last
{service} and wanted to check in - everything
ok?
If you would like to book, here you go:
{booking_link}. And if something was not right
last time, tell me honestly, I would rather know.Touch 3 - day 21 (SMS)
Hi {first_name}, last one from me - I will
leave you be after this.
{stylist} has space on {day_option_1} and
{day_option_2} if you would like your
{service} back in the diary: {booking_link}
Either way, thank you for coming to us. The
door is open whenever.Send touch one from the stylist by name, not from the salon. The relationship in a salon is with a person, and a message from "the team at" is the fastest way to signal that nobody actually noticed she was gone. If the stylist has left, do not send this sequence at all - that client needs a different, human conversation about who takes over.
Keep every message short enough to read on a lock screen, include a booking link that opens straight into booking rather than a homepage, and make sure a reply reaches a human within the hour during opening times. A win-back message that gets a reply nobody answers for two days is worse than no message. If you are sending automated SMS in the US, brand and campaign registration comes first - see the A2P 10DLC guide and the opt-in and stop handling entries.
The build, step by step
Export the appointment history
Two years minimum, ideally three. One row per completed appointment: client id, name, mobile, email, service, stylist, date, and value if you have it. Every booking system exports this, and if yours only gives you a PDF, that is the first thing to fix. Filter to completed appointments so cancellations do not flatter the interval.
Compute intervals and service medians
For each client, sort visits by date, take the gaps, take the median. Store it. Separately, compute the median interval per service across the whole book, for the clients with too little history. This is a script or a spreadsheet, not an automation platform job - do it once, look at the distribution, and sanity check it against what your team believes about the client base.
Score everyone and look at the list before sending anything
Apply the lapse rule and the exclusions, then read the resulting list with your senior stylist. This step is not optional. They will recognise names, and they will catch the three categories a rule cannot: the client who moved away, the one who is unwell, and the one you parted with on purpose. Fix the exclusion flags from what they tell you, then re-run.
Hold back a control group
Randomly exclude ten to twenty percent of the eligible list and send them nothing. Some lapsed clients rebook by themselves, and the control group is the only way to know how many. Without it your reporting is a story rather than a measurement, and you will keep spending on a campaign whose real effect you have never seen. Related: ROI and payback period.
Write the state table and run the first batch small
One row per client in sequence: client id, interval, threshold, lapse date, status = open, step = 0, next_touch. Run the first batch at fifty clients, not five hundred. You will find a data problem - a bad phone number format, a duplicate client record, a service name that did not map - and you would rather find it on fifty.
Build the stop signals first
Any inbound SMS or email stops the sequence and assigns it to a person. Any new booking stops it. Any opt-out stops it and writes back to the client record so no other campaign picks them up. Check the booking system for a new appointment immediately before every send too, because a client who booked online this morning must not get a "we miss you" text this afternoon - see idempotency for the general habit.
Run it on a schedule, not as a one-off blast
Once the back catalogue is worked, a daily scheduled job recomputes who has newly crossed their threshold and enters them. This is the ongoing version, and it is a trickle rather than a blast - a handful of clients a day, which is also the volume a front desk can actually follow up by hand if a reply comes in.
Report returns against the control, monthly
Clients entered, messages sent, replies, bookings made within thirty days, and the same booking rate for the held-back control. The difference between those two rates is what your automation did. Report it that way even when the honest number is smaller than the headline one, because the alternative is a campaign everyone stops believing in around month four.
Where these builds go wrong
- One flat lapse threshold for a client base with wildly different natural rhythms.
- Computing intervals from an average instead of a median, so one holiday distorts the rule.
- Including cancelled and no-show appointments in the history, which makes clients look more frequent than they are.
- Not checking for a future booking immediately before sending.
- Sending from the salon rather than the stylist the client actually sees.
- Leading with a discount, then wondering why the returning clients never pay full price again.
- Re-entering the same clients every quarter until they opt out of everything you send.
- No control group, so every self-motivated rebooking gets counted as a recovery.
- Nobody watching the inbox, so a reply to a personal-sounding message sits unanswered for two days.
- Messaging a client whose stylist has left without telling them who is taking over.
The interval computation is genuinely the whole game here - the sequence and the copy are the easy part, and most salons could improve their win-back results without sending a single extra message just by changing who gets one and when. If you want the sector view, the salon automation page covers the other workflows worth building, the automated review requests guide covers the other side of the visit, and AI automation cost covers the money. To have it scoped properly, start with the free automation audit and see automation services for how we build and hand over.
FAQ
- Why is a 90-day win-back rule a bad default?
- Because 90 days means completely different things to different clients. A client who has come in every six weeks for two years is meaningfully overdue at ten weeks and long gone at ninety days - you have waited more than double their normal gap before saying anything. A client whose natural rhythm is a cut every twelve weeks is not lapsed at ninety days at all; they are on time, and a message telling them you miss them is embarrassing. One flat number chases the loyal and misses the leaving.
- How do I calculate a client's service interval?
- Take every completed appointment for that client, sort by date, compute the gap between each consecutive pair, and take the median of those gaps. Median rather than average, because one long gap for a holiday or an illness drags an average badly. Three visits gives you two gaps, which is thin but usable. Below that, fall back to the median interval for the service they booked rather than inventing a personal number from a single data point.
- What multiple of the interval means lapsed?
- Around 1.5 times their own interval is a defensible starting point, and it is the multiple most commonly suggested in salon retention writing. The mechanism is what justifies it rather than any published study: at exactly one interval a client is merely due and may already have booked; at twice the interval they have visibly moved on and the message arrives late. Halfway between is early enough to matter and late enough not to nag. Put a floor of about three weeks and a ceiling of about six months on the result so a strange interval cannot produce an absurd threshold.
- Do Boulevard, Vagaro and Fresha already do this?
- Partly, and it is worth knowing exactly what you would be rebuilding. Vagaro documents a Lost Customer campaign type for both email and text, described as automatically messaging customers who have not visited in a while, configured in weeks since their last checked-out appointment. We could not read those support articles directly in August 2026 because support.vagaro.com returned 403 to our automated fetch, so confirm the current behaviour in your own account. We could not confirm equivalent documented win-back automations for Boulevard or Fresha from their public pages - Fresha's marketing page rendered no feature detail to us, and we found no Boulevard support documentation naming a win-back campaign - so treat those as unverified rather than absent. The general point stands either way: a weeks-since-last-visit setting is a flat rule, which is precisely the thing this guide argues against.
- Should the win-back message include a discount?
- Be careful. Discounting works in the sense that it produces bookings, and it has a cost beyond the margin on that visit: a client who returns for twenty percent off has learned that going quiet produces a discount, and some of them will do it again. That is a real behavioural risk and any page that tells you to lead with an offer without mentioning it is selling you something. Our default is no discount on the first two touches, and if you use one at all, put it on the final touch, make it clearly one-off and time-boxed, and prefer a value-add - a treatment, a longer consultation, a product sample - over money off the service price.
- What response rate should I expect from a win-back campaign?
- We are not going to give you a number, and you should distrust the pages that do. The reactivation percentages that circulate for salons come overwhelmingly from vendor marketing rather than published research, and they vary with client base, price point, why people left and how long the list has been sitting. Measure your own, and measure it against a held-back control group of lapsed clients who get nothing. Without a control you will count every client who was going to rebook anyway as a win, and the number will look excellent and mean nothing.
- How often should a lapsed client be contacted?
- Once as a sequence, then leave them alone. Three touches across roughly three weeks, then close the record and do not re-enter the same client for at least a year unless they visit again in between. A win-back list that gets worked every quarter is a nuisance list, and the people on it will opt out of your marketing entirely, which costs you the ability to reach them when you actually have something to say.
- Is this worth automating for a small salon?
- It depends on whether anyone is doing it now. If your front desk already calls lapsed regulars and it works, automation mostly buys you consistency and a record. If nobody has ever done it, the first run over a two-year client list is where the value is, and that run is a one-off analysis rather than an automation. Our honest advice is to do the analysis first - compute the intervals, see how many genuinely lapsed clients you have - and only build the ongoing automation if that number is big enough to matter every month.
Done for you
Want your win-back list built on each client's own rhythm instead of a flat 90 days?
We compute the intervals from your booking history, build the sequence with a control group so you can see what it really did, and hand you the whole thing documented. The audit comes first, and it is free.
The consult is free, and you keep whatever we build.
Keep building
The rest of the cluster
AI Appointment Reminders That Cut No-Shows
Two-way confirm and reschedule, waitlist backfill, cadence, and SMS compliance basics.
Read the guideAutomated Review Requests After Every Job
Completion vs payment vs delay, what Google's policy forbids, and a two-message sequence that catches complaints first.
Read the guideAutomate Unsold Estimate Follow-Up for Contractors
The estimate-state trigger, five touches of real copy, stop-on-reply and stop-on-acceptance detected from the system itself.
Read the guideFree AI Automation Audit Template
Score every recurring task and rank the top three worth automating before you build anything.
