What problem does it solve?
A loyalty programme only works if members actually use it. Many programmes still treat every member the same: the same point multiplier, the same birthday coupon, the same generic email, whether a member visits weekly or has not come back in months. That leaves two things on the table: members who are about to lapse get no different treatment than loyal ones, and members close to a reward they would value do not hear about it until it is too late to matter.
The job covers the loyalty programme itself. It runs inside the fixed rules of a membership programme (points, tiers, a redemption catalogue and a promise that status and benefits mean something) for the programme's own members. Salesforce reports that 65% of consumers say they will remain loyal to companies that offer more personalized experiences, a high bar for a programme that still emails every member the same offer on the same day.
- 65% of consumers say they will remain loyal to companies that offer more personalized experiences.IHG Hotels & Resorts Aims to Boost Guest Loyalty with Salesforce (2024)
How does it work?
- Unify the member profile. Purchase history, app and web behaviour, loyalty tier and redemption history come together into one profile per member, across every brand and channel the programme spans.
- Predict who needs attention. A model scores each member's risk of lapsing, or their readiness for a specific reward, from recency, frequency and engagement signals.
- Pick the action for the member. Instead of one offer for everyone in a segment, the system chooses among many possible offers, rewards or messages the one most likely to matter to that specific member, within the programme's point and budget rules.
- Deliver it where the member already is. The action reaches the member through the app, a push notification or email.
- Test, learn and stay inside the programme's promise. Results feed back into the model, and the programme keeps its published rules (how points are earned, what a tier guarantees) intact, so personalization only shapes which message a member receives.
- Audience
- Customer facing
- Autonomy
- Copilot
- Adoption
- Emerging
- Channels
- Mobile app, Email
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Conversion uplift | Too few to pool | 2x | 1 | 1 vendor |
Value drivers: Customer experience, Revenue growth, Lower cost to serve.
Indicative value
A retail or restaurant loyalty programme with 5 million active members
USD 3 million to USD 20 million
Member spend retained per year per year
How this is calculated
Formula: activeMembers * atRiskShare * retentionLift * avgAnnualSpendPerMember. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Active members activeMembers, members | 5,000,000 | 5,000,000 | The reference programme. |
| Average annual spend per active member avgAnnualSpendPerMember, USD per member per year | 200 | 500 | Editorial assumption, replace with your own average member spend. |
| Share of members at risk of lapsing in a given year atRiskShare, fraction of active members | 0.1 | 0.2 | Editorial assumption, replace with your own lapse rate. |
| Additional share of at risk members retained with personalization retentionLift, fraction of at risk members | 0.03 | 0.04 | Braze reports a 5% retention lift among at risk guests for Panera Bread from an AI powered, Braze delivered campaign, but the source does not say whether that is percentage points or a relative change. The range is set below that figure, with a high end of 0.04, because it covers one campaign around a single menu change, not a full year of personalization, and it is Braze's own reported number, not an independently measured result. |
What it leaves out: Gross retained spend only. It leaves out the cost of the platform and of any reward or discount given to retain a member, the share of members who would have stayed without any intervention (only a control group removes that), and any effect on members who are not flagged as at risk.
Who already uses it?
3 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Starbucks
United States · Retail and ecommerce · 2024
Starbucks Chief Executive Officer Laxman Narasimhan told investors that Starbucks activated new capabilities within its proprietary Deep Brew data analytics and AI tool to identify and incentivize specific Starbucks Rewards member cohorts, part of a set of initiatives aimed at bringing occasional customers back into the loyalty programme. Narasimhan said the company had already seen a positive impact from these initiatives, with occasional customers beginning to rebound in December 2023. US 90 day active Starbucks Rewards members grew 13% year over year to a record 34.3 million in the quarter.
No outcome disclosed.
IHG Hotels & Resorts
United Kingdom · Travel and hospitality · 2024
IHG Hotels & Resorts announced an expanded relationship with Salesforce to standardize its guest CRM on the Einstein 1 Platform (since renamed the Salesforce Platform) and adopt Salesforce Loyalty Management, which Salesforce says will make IHG "the first global hospitality company to standardize on Salesforce Loyalty Management."
Salesforce describes IHG as being able to grow its IHG One Rewards membership "with predictive AI capabilities" once the rollout is in place, and names generative AI for guest stay management as a further step it is laying the foundation for.
The one guest facing capability the announcement describes as already running is plain segmentation, not AI: when a guest starts but does not complete a booking, a segment is created in Data Cloud and sent to Marketing Cloud to prompt a return visit.
No outcome disclosed.
Panera Bread
United States · Retail and ecommerce · 2024
For its largest menu transformation, Panera Bread combined its own AI powered decision engine with the Braze customer engagement platform to personalize MyPanera loyalty communications during the change. AI decisioning segmented guests likely to lapse because of the menu changes; Panera then created over 4,000 unique combinations of personalized offers and product recommendations delivered across email, app and web, and used Braze's Custom Events to trigger reactivation messages to guests who browsed new items but did not buy. Braze reports the campaign achieved a 5% retention lift among at risk guests, a doubling of loyalty offer redemptions, and a doubling of purchase conversion rates from abandoned order follow ups.
- Conversion uplift: 2x, April 2024 menu transformation campaign, abandoned order follow ups
"By leveraging Braze, Panera achieved a 5% retention lift among at-risk guests, 2X loyalty offer redemptions, and 2X purchase conversion rates from abandoned order follow-ups."
Claimed by: vendor
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- A unified member profile across every brand, channel and touchpoint the programme spans
- Programme rules and the current reward or redemption catalogue, kept in sync with what members were promised
- At least a year of member transaction and engagement history with lapse outcomes
Systems to integrate
- Loyalty and rewards management platform (points, tiers, redemption)
- Customer data platform or CRM holding the unified member profile
- Email, push notification and in app messaging systems
- SMS and WhatsApp messaging, where members have opted in
Complexity: Medium
The personalization logic is well understood. The member data takes more work: unifying profiles across brands, channels and sometimes acquired loyalty programmes, while keeping the published programme rules (how points are earned and what a tier guarantees) completely consistent for every member regardless of what the AI recommends.
- 1
Unify the profile before personalizing anything
A single view of the member, across every brand and channel the programme spans, has to exist before any model can decide what to say to them.
- 2
Separate the programme's rules from the personalization layer
Points earning, tier thresholds and what a reward is worth must stay exactly what was published to members; the AI decides what to communicate and when, and the programme rules decide what a member is entitled to.
- 3
Score lapse risk and reward readiness, not just RFM segments
Build models that flag members close to lapsing and members close to a reward they would value, so the same member can get a different message next month as their behaviour changes.
- 4
Pilot on one moment that matters
Panera Bread built its personalization around one concrete event, a menu transformation that risked confusing or losing guests, rather than personalizing everything on day one.
- 5
Keep a holdout and measure against it
Compare personalized members against a matched group that gets the standard treatment, so a real lift can be told apart from members who would have stayed or redeemed anyway.
Guardrails
- Programme rules (points, tiers, redemption value) enforced identically for every member, independent of any AI recommendation
- A cap on how many personalized offers or messages a member receives in a period
- Marketing consent and channel preferences enforced before any personalized outreach
- A member can always see and reach a human about their actual points balance and tier status
KPIs to instrument
- Lapse rate among at risk members reached with personalization versus a holdout
- Reward redemption rate versus a holdout
- Member complaints about incorrect points, tier status or reward value
- Message frequency per member, to catch fatigue before it shows up in lapse
Human in the loop
A loyalty or CRM team owns the reward catalogue, the message cadence caps and the model's performance, reviewing lapse and redemption results against a holdout regularly. Member service handles any dispute about points, tier status or a reward, since those are programme commitments the AI does not have authority to change.
Common failure modes
- Personalization contradicts the programme's own rules
- A message implies a reward or tier benefit the programme does not actually grant. Generate messages from the same rules engine that governs the programme, never from a separate, driftable copy of it.
- Over messaging the members most worth keeping
- The best, most engaged members get the most personalized attention and end up fatigued by it. Cap frequency per member and monitor unsubscribes and complaints by engagement tier, not only by segment.
- No holdout, no real lift
- Comparing treated members to their own past behaviour overstates the effect, since overall demand and seasonality move too. Keep a matched holdout group for as long as the programme runs.
- A single lapsed brand poisons the unified profile
- When a loyalty programme spans several brands, a bad signal from one (a returns spike, a service outage) can misclassify a member's risk everywhere. Validate signals per brand before they feed a cross brand profile.
What are the risks and rules?
EU AI Act
Depends on design
Deciding a loyalty offer or reward for a customer is not listed in Annex III and does not itself evaluate creditworthiness, employment or access to an essential service, so a back office personalization engine is minimal risk. If a conversational channel is added to talk to members directly, Article 50 would apply and members must be told they are dealing with AI. Personalization that deploys manipulative or deceptive techniques, or exploits a member's vulnerabilities due to age, disability or a specific social or economic situation, in a way that causes or is reasonably likely to cause significant harm, is prohibited under Article 5(1)(a) and (b).
Rules that apply
Guidance
- Article 5: Prohibited AI practices (European Union, Europe). Bans manipulative techniques and the exploitation of vulnerabilities, including those due to a person's age, disability, social or economic situation.
Controls to put in place
- Programme rules (points, tiers, redemption value) served from one authoritative source, never duplicated into the personalization layer
- Message frequency caps enforced per member across every channel and brand
- Consent and channel preference checks before any personalized outreach
- Regular review of personalization outcomes by member segment, to catch fatigue or unequal treatment early
Frequently asked questions
- How is this different from marketing personalization at scale and from an offers and rewards agent?
It is scoped to a membership programme's own members and its published rules: points, tiers and a redemption catalogue that promise something specific to each member. A general marketing personalization engine picks the next offer or message for any customer across any campaign. This use case personalizes what a loyalty member sees inside their programme and leaves what a member has earned to the programme rules. The offers and rewards agent covers the same kind of job for a bank or card issuer's own rewards programme; this page covers retail, restaurant and travel membership programmes instead.
- What results has AI loyalty personalization actually shown?
Braze reports that Panera Bread achieved a 5% retention lift among at risk guests and a doubling of loyalty offer redemptions from an AI powered, Braze delivered campaign around a menu change. Starbucks Chief Executive Officer Laxman Narasimhan told investors that Starbucks had activated new capabilities in its Deep Brew data analytics and AI tool to identify and incentivize specific Rewards member cohorts, as one of several initiatives he said were already having a positive impact on occasional customers. IHG Hotels & Resorts has announced plans to standardize its guest CRM on Salesforce's platform and grow IHG One Rewards membership with predictive AI, but this is a stated intent, not yet a published result.
- Does AI loyalty personalization need to be disclosed to members under the EU AI Act?
Choosing an offer is minimal risk, so no disclosure is required for that alone. Article 50 requires telling a person they are interacting with AI, unless that is obvious from the context, where a conversational channel talks to members directly. Article 5 bans manipulative techniques and the exploitation of vulnerabilities in any personalization.
How to cite this page
Blits.ai AI Use Case Library, "AI loyalty and rewards programme personalization", last verified 29 September 2026, https://www.blits.ai/ai-use-cases/loyalty-rewards-personalization-agent. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 13 October 2026: First published