Export each client's treatment history (last visit, what they had, how often, spend, therapist, marketing consent), group clients into four to six patterns such as monthly facial regulars or one-off gift-voucher visitors, and have AI draft one offer per pattern, then personalise the wording per client. Keep health notes out of marketing entirely.
Personalisation here means relevance, not flattery. A client who had three hot stone massages this year should hear about the new hot stone ritual, not the teen facial. You'll leave with a map of which data is usable and which isn't, six patterns worth an offer each, the prompts to write them, a rule for where personal becomes creepy, and a holdout test that shows whether any of it beat your normal newsletter.
Which fields in your booking data are fair game for marketing
Spa software holds two very different kinds of information about a client: what they chose to buy, and what your therapists learned about their body. Only the first belongs in marketing.
| Field | Use for offers? | Note |
|---|---|---|
| Treatments booked and dates | Yes | The core of every pattern below |
| Visit frequency and total spend | Yes, for grouping | Never quote spend back to the client |
| Preferred therapist, day and time | Yes | Makes the offer easy to act on |
| Retail products bought | Yes | Useful for refill reminders |
| Gift voucher bought or redeemed | Yes, carefully | The buyer and the recipient are different people |
| Birthday | Only if collected for this purpose | Check how it was asked for |
| Consultation form: conditions, pregnancy, medication, allergies | No | Collected for safety, not selling |
| Therapist notes: skin concerns, tension areas | No | Use in the treatment room, not in an email |
| Marketing consent | Required filter | No consent, no offer |
Health information counts as special category data under data-protection law such as the GDPR, and using it to target marketing is a different purpose from the one it was collected for. If you're unsure where a field falls, ask your data-protection adviser before it goes anywhere near a campaign. Sorting fields like this is worth doing once for the whole business; classifying business data before using AI tools gives a simple scheme.
Six treatment-history patterns worth their own offer
Most spa client lists contain the same recurring shapes. Each gets one offer, not a menu of discounts.
1. Gift-voucher one-timers
Spot it: one visit, paid by voucher, no booking since. Offer: an introduction to booking for themselves, such as the treatment they had at a returning-guest price, or a shorter version at a lower price point. Watch: never mention who bought the voucher.
2. Single-treatment loyalists
Spot it: four or more of the same treatment in 12 months and nothing else. Offer: a complementary add-on with their usual booking, or a course at a package price. Watch: they like what they like; suggest, don't replace. Upselling add-on treatments without being pushy covers the pairings.
3. Seasonal regulars
Spot it: bookings clustered in the same month or two each year, often around a birthday or the end of the year. Offer: an early-booking reminder three to four weeks before their usual window, with their preferred therapist's availability. Watch: no discount needed; the value is getting the slot they want.
4. Drifting regulars
Spot it: their gap since the last visit is now about twice their usual gap. Offer: a friendly nudge from their therapist with two available times. Watch: hold any incentive for a second message.
5. Course finishers
Spot it: completed a series, such as six facials or a block of massages, in the last month. Offer: a maintenance plan or membership priced below the pay-as-you-go rate. Watch: time it within weeks of the last session, while the results are fresh.
6. Couples and pairs
Spot it: bookings that always sit alongside another client's in a couples room. Offer: a couples ritual or anniversary package. Watch: send to the person who books, and never send couples offers to someone whose pair bookings stopped.
Getting the export into shape
Export 12 to 24 months of appointments with a client ID, date, treatment, price paid, payment type (to spot vouchers), therapist, and the marketing consent flag. Leave names, contact details and every consultation field out of the file. Merge duplicate client records first, or the same person will land in two patterns. Then:
Attached: appointments.csv (client_id, date, treatment, price_paid,
payment_type, therapist, couples_room Y/N, marketing_opt_in Y/N).
Use only clients with marketing_opt_in = Y.
Assign each client to at most one of these patterns, in this priority:
1 course finisher: 6+ of the same treatment ending in the last 45 days
2 drifting regular: 4+ visits, current gap > 2x their median gap
3 seasonal regular: 60%+ of visits in the same 2 calendar months
4 single-treatment loyalist: 4+ visits, all the same treatment
5 couples: 3+ visits in a couples room
6 gift-voucher one-timer: 1 visit, payment_type = voucher
Return counts per pattern, then a file with client_id, pattern,
last treatment, usual therapist, usual weekday.
Show five example clients per pattern with the reasoning.Spot-check the examples against your booking system. Expect a chunk of active clients to fit no pattern; they get your normal newsletter, and that's fine.
The examples are where rule problems show. An illustrative extract from the seasonal-regulars list:
Client 5512: seasonal regular. 3 of 4 visits (75%) in November-December across 2024 and 2025. Usual therapist: [therapist]. Usual day: Saturday.
Client 6027: seasonal regular. 2 of 2 visits (100%) in December 2025.
The first is a real pattern. The second is one person who came twice in one December, perhaps with a voucher and then once more, and that says nothing about next year. The 60% rule needs a floor: "at least 3 visits, spread over at least two different years". Rerun with that and the seasonal group usually shrinks, which is the point: a smaller list of people who genuinely return at the same time each year.
Gift-voucher one-timers have their own trap. When a voucher is booked, whose contact details end up on the client record? If reception booked the appointment using the buyer's email, your "one-timer" offer goes to the person who paid for the gift, not the person who enjoyed it. Filter the pattern to records where the visitor's own contact details were captured at check-in, and fix the booking process so they always are.
Drafting one offer per pattern with AI
Give the model the pattern, the treatments you're willing to offer, and hard limits on discounts, so it can't invent a deal you can't afford:
Write a marketing email and a text message for one client group.
Group: [pattern description, e.g. "gift-voucher one-timers: visited
once on a voucher, never booked themselves"].
Offer: [exact offer, e.g. "60-min signature facial at $99 instead of
$120, valid 6 weeks, first self-booked visit only"].
Voice: calm, warm, plain English. Email 80-120 words, 3 subject lines.
Text under 160 characters with the booking link [link].
Personalise with these merge fields only: {first_name},
{last_treatment}, {therapist}.
Never: invent treatments, prices or deadlines; make health claims
("detox", "cure", "heal"); mention how much they've spent or how
many days since their visit; mention who bought a voucher.Check the output for claims your therapists wouldn't make. Spa copy drifts towards wellness promises quickly, and "detox" or "boost your immunity" in an email is a claim you'd have to stand behind.
An illustrative first draft of the gift-voucher email, even with the "Never" line in place:
Subject: Your glow is waiting ✨
Hi {first_name}, we hope you loved the {last_treatment} someone special treated you to! Now it's your turn to book for yourself: enjoy our detoxifying signature facial for just $99 (usually $120). Hurry, this offer ends Sunday!
Three rule breaks in four sentences. "Someone special treated you to" mentions the voucher's buyer. "Detoxifying" is a health claim. "Ends Sunday" is a deadline you didn't set; the offer runs six weeks. The edited version keeps the price and the warmth and loses all three: "Hi {first_name}, we hope you enjoyed your {last_treatment} with us. If you'd like to come back for yourself, our 60-minute signature facial is $99 instead of $120 on your first self-booked visit, any time in the next six weeks." Point each break out in the chat so the next pattern's draft starts cleaner.
Merge fields or individually written messages?
There are two levels of personalisation, and the right one depends on the size and value of the group:
- Merge fields (first name, last treatment, therapist) inside one well-written template. Right for large groups such as the gift-voucher one-timers, where a few hundred near-identical messages are fine and reviewing each one isn't practical.
- A line written per client by AI from their history, reviewed by a person. Right for small, valuable groups such as course finishers or long-standing regulars. For 45 clients, having the model draft one opening sentence each ("You finished your six-week facial course with [therapist] in August…") and a manager reading them all takes about half an hour.
Never send individually generated messages unread. One garbled line about the wrong treatment undoes the point of personalising. An illustrative three from a batch of 45 course-finisher openings:
- "You finished your six-week facial course with [therapist] in August, so this is the ideal moment to keep those results going."
- "Your massage block with [therapist] wrapped up last month, and she's kept Thursday mornings free for regulars."
- "Congratulations on completing your six-week facial course with [therapist]!"
The first is fine. The second assumes the therapist has kept slots free, which nobody told the model; check the diary or cut the clause. The third went to a client who switched therapists halfway through the course, so it names someone she last saw in June. The export only held the most recent therapist, and the model filled in a story around it. This is exactly what the half-hour read is for.
Timing, channel and how often
A relevant offer at the wrong moment still gets ignored. Rough timings by pattern:
- Course finishers: within a week of the final session.
- Drifting regulars: when they reach twice their usual gap, by text from their therapist.
- Seasonal regulars: three to four weeks before their usual month, by email, when they're planning.
- Gift-voucher one-timers: four to eight weeks after their visit, while they still remember it.
- Couples: ahead of the dates they've booked around before.
Use text for anything time-sensitive and short; email when the offer needs a picture or explanation. Set one hard limit across every pattern: no client gets more than one offer a month, whichever groups they could fall into. The priority order in the prompt above exists for exactly that reason.
Where personal turns creepy
The line is simple: mention what the client chose, never what you observed about them.
- Fine: "Since you enjoyed the hot stone massage with [therapist] in March, we thought you'd like to know about her new ritual."
- Too precise: "It's been 63 days since your last visit." Accurate, and it reads like surveillance. "It's been a while" does the same job.
- Not acceptable: "Your skin felt dehydrated at your last facial, so…" That's a therapist note used for selling.
- Not acceptable: anything implying pregnancy, a medical condition, or weight, even if the booking history hints at it.
- Not acceptable: "Your partner treated you to a spa day last year…" The voucher belongs to someone else's relationship with you.
What your spa software may already do
Before building this in a spreadsheet, check your platform. As of September 2026, Zenoti lets you create dynamic audiences using filters such as service history, visit recency or loyalty tier, describes a Smart Marketing engine that segments clients and sends personalised messages, and offers an assistant called Zeenie that drafts campaign content. Phorest runs automated SMS and email and an Expert Recommendations SMS that lets staff send tracked product recommendations. If your system can build the audiences, use AI only for the wording and keep the segmentation where the data lives. Segmenting an email list with AI covers the method if you're doing it outside the platform.
Six patterns in one day spa's client list
Illustrative figures only. Say a day spa has 2,300 clients on file, 1,400 of whom visited in the last 18 months and 1,050 of whom have agreed to marketing. The pattern prompt returns:
| Pattern | Clients | Offer |
|---|---|---|
| Gift-voucher one-timers | 310 | Signature facial, $99 instead of $120, first self-booked visit |
| Seasonal regulars | 180 | Early access to their usual month, no discount |
| Single-treatment loyalists | 140 | A 15-minute add-on with their usual booking |
| Drifting regulars | 120 | Therapist nudge with two times |
| Couples | 60 | Anniversary couples ritual |
| Course finishers | 45 | Monthly maintenance plan |
| No pattern | 195 | Normal newsletter |
Check the economics before sending. If the facial costs about $55 to deliver in therapist time, product and room, the $99 offer still leaves $44, against $65 at full price. If 8% of the 310 one-timers book, that's about 25 facials and $2,475, and the offer is worth it only if some of them come back at full price. Your response rates will differ; these numbers show the sum to run.
Proving it beat your usual newsletter
Hold back a random tenth of each pattern and send them only the normal newsletter. After 30 days, compare bookings per 100 clients between the group that got the tailored offer and the group that didn't. Track four things per pattern: booking rate, revenue per message sent, discount given away, and unsubscribes.
Here's the sum for the gift-voucher group from the example. Of 310 clients, a random 31 are held back and 279 get the facial offer. After 30 days, 22 of the 279 have booked, about 7.9 per 100, and 1 of the 31 has booked through the newsletter, about 3.2 per 100. The offer group booked about 4.7 more per 100, so roughly 13 of those 22 bookings wouldn't have happened without it. Revenue per message sent was about $7.80 (22 facials at $99, over 279 messages), and the discount given away was $21 on each of the 22, $462 in all. One booking in a group of 31 is a thin comparison, though; a single extra booking in the holdout would cut the apparent lift by more than two-thirds. That's why the rule below asks for two rounds before deciding.
A pattern that doesn't clearly beat the newsletter after two rounds isn't worth the effort; fold those clients back into the general list. The ones that win become standing automations. Once patterns prove themselves, they're the backbone of a loyalty scheme, which building a loyalty programme with AI personalisation covers. And whatever tool runs them, the rules in keeping customer data private when your team uses AI apply to every export you make.
Further reads
- How Day Spas Can Take Bookings From Instagram DMs With AI — Let an assistant answer the DMs your offers generate.
- How to Win Back Lapsed Customers With AI-Personalised Emails — A fuller sequence for clients who've stopped coming.
- Best AI Email Marketing Tools for Small Businesses (2026) — Email tools that handle segments if your spa software can't.
- Salon and Spa Software With Built-In AI: What to Compare — Compare spa systems on segmentation and AI copy features.
- Is It Safe to Put Customer Data Into ChatGPT? — Settings to check before uploading any client export.
- What Can AI Do for Your Email Marketing? 10 Practical Uses — Ten other jobs AI does well in email marketing.
- How Cafés Use AI to Fill Quiet Hours With Targeted Offers — Find your café's quiet hours in the till data, design offers for the people free at those times, and measure whether they brought anyone new in.
- How Salons Use AI to Rebook Clients and Fill Gaps in the Diary — Return intervals from your own history, a chair-side rebooking habit, and a way to match each empty slot to the five clients most likely to take it.
- How to Upsell Room Upgrades and Late Checkout With AI — Inventory rules, an upgrade price ladder, offer timing and the AI jobs that lift take-up, with the upsells that backfire on small hotels.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Zenoti marketing platform page (dynamic audiences by service history and visit recency, Smart Marketing, Zeenie AI Assistant); Phorest platform pages (automated SMS and email, Expert Recommendations SMS, Ads Manager). Checked September 2026. Client counts, prices and response rates in the example are illustrative.