A next-service suggestion is most useful when it fits the client, not just the appointment before it. Your booking history can help you find services that clients often book in sequence. But a popular pairing is only a starting point: the client’s current needs and your eligibility rules should determine whether you recommend it.
Recent recommendation research offers a useful principle: customer history can inform what to suggest next. A separate line of research on AI agents points to another requirement: suggestions should reflect the user’s current situation. Neither research line proves that a particular pairing will increase bookings at your business. Together, they offer a practical way to structure a small, testable approach.
Start with booking sequences
Recommendation research studies how past interactions can help predict what someone might want next. One paper describes “sequential recommendation” as predicting future interests from a person’s interaction history (LSF-SR paper). Its proposed method combines item identifiers with language-generated information about what those items mean. In plain terms, it explores using both what someone interacted with and what those items represent.
The authors tested their method on five public benchmark datasets and reported that it performed better than comparison methods in those tests (LSF-SR paper). That is an early research result, not evidence that a service pairing will lift revenue or retention at a salon, clinic or studio. The useful takeaway for an owner is more modest: look at the order of past bookings, not just which services are popular overall.
For example, review whether clients who book a particular gateway service—the service that can lead naturally to a follow-on visit—often book another service later. Choose a reasonable time window for your business and compare like with like. A pairing that appears often may be worth testing; it is not automatically right for every client.
Add the client’s current context
A second recent paper looks at AI agents that need to account for live user state, such as session history. Its authors call a mismatch between a response and the user’s current state “direction drift” (State-Grounded Conditioning paper). The paper tests its approach with an in-game coaching agent, not a booking or service business, so its findings should not be treated as proof about client recommendations. It does reinforce a useful design question: does the suggestion fit what is happening now?
For a service business, that means a sequence alone should not trigger an automatic recommendation. Consider the client’s existing bookings, service history, stated preferences and any relevant eligibility rules before making a suggestion. A discussion of enterprise AI context also lists previous interactions, transaction data and constraints as useful context; the article presents Google Cloud’s view, so treat it as a vendor perspective rather than independent evidence about service businesses (Diginomica interview).
You do not need an AI system to apply this principle. A booking report and a clear staff guide may be enough to identify a few candidate pairings. If you later use AI to draft suggestions, give it defined inputs and rules rather than asking it to invent a recommendation from a client’s entire history.
Make a small test
Choose a few pairings that make sense for your services and write down who is eligible for each one. Include situations where staff should not make the suggestion or should check with the client first. Keep the rule simple enough that your team can apply it consistently.
Then compare a specific suggestion with your usual approach. Track whether the client books the follow-on service and whether the booking is completed. You can also review return visits and revenue per client, but keep the same measures and comparison period for each test. The point is to learn which suggestions deserve a place in your process, not to add more offers to every conversation.
What to do
- Export booking history and group visits by service sequence.
- Pick a small set of pairings to review with your team.
- Write clear eligibility rules and note when staff should not suggest a service.
- Test the suggestions against your current approach and review the results before expanding.
A careful pairing strategy starts with your own client patterns, then adds the rules and context that keep each suggestion relevant. That gives you a practical test before deciding whether more automation is worth the effort.
- client retention
- service pairings
- booking data
- recommendations
- service optimization
Alan Jordan, founder and principal solutions architect. Data engineering, data science and business analytics for appointment-based businesses.
