Personalisation and Loyalty Examples Across the Customer Lifecycle

A traveller filtering hotels by electric-vehicle charging, a grocery shopper getting money off products they actually buy, and a subscriber being helped before they cancel are all personalisation. None of them creates loyalty simply because customer data was involved.

The difference is context. Useful personalisation reduces effort, recognises a real need, or helps someone get more value from what they already have. Poor personalisation repeats itself, exposes an uncomfortable inference, or leaves the customer feeling that information about them is being used against them.

This guide works through eight examples, every one of them documented by the company that built it. For each one it sets out the customer context, the signal being used, the personalised action, the intended loyalty outcome, a sensible measurement method, the main privacy or trust risk, and the lesson.

Named examples come from first-party company publications. Those results are self-reported and have not been independently verified, so treat them as illustrations of what a company chose to build and publish, not as proof that personalisation caused loyalty or as outcomes you should expect to replicate.

Four lifecycle stages are deliberately absent: first visit, first purchase, lapse prevention and win-back. Few companies publish what they do at those moments, and I would rather leave a gap than fill it with an invented scenario dressed up as a case. For campaign designs covering those stages, see our guide to personalised engagement tactics that increase customer lifetime value.

A word on which of these to copy. In enterprise account meetings I lost count of the times a brand asked for the most impressive-looking example in the deck rather than the one that fitted their business. When I asked what was actually wrong with their current experience, the room would often go quiet. They knew something was off, usually from their growth numbers, but nobody could name it. Read what follows as a menu of problems, not a menu of features.

For the psychology and evidence behind all of this, see how personalisation can affect brand loyalty. For programme design, segmentation, triggers and operating models, see how to build a personalisation strategy for loyalty growth.

Personalisation examples at a glance

Customer stage	Example	Industry	Main loyalty opportunity
Discovery	Preference-led hotel search	Travel and hospitality	Make the brand easier to choose
Programme onboarding	Vodafone VeryMe preferences	Telecommunications	Establish an explicit value exchange
Second purchase	Sainsbury’s Your Nectar Prices	Grocery retail	Make repeat purchasing more useful
Repeat use	Netflix recommendations	Subscriptions and media	Help customers keep finding value
Loyalty milestone	Starbucks Rewards tiers	Food service	Make progress and recognition visible
High-value recognition	Hilton Honors upgrades	Hospitality	Recognise valuable customers when it matters
Ongoing service	Bank of America’s Erica	Financial services	Turn personalisation into practical help
Advocacy	Spotify Wrapped	Digital subscriptions	Help customers express identity

1. Discovery: match the experience to an explicit travel need

Named example: Hilton search filters

Customer context. A traveller is comparing hotels with a practical requirement: electric-vehicle charging, a pet coming along, or breakfast included.

Data or signal used. The strongest signal here is an explicit filter selected during the current search. It reflects what the customer is asking for right now, not an assumption drawn from an old profile.

Personalised action. Hilton added search filters for attributes including electric-vehicle charging, pet-friendly properties and free breakfast. The company reported that these three were among the most popular filters on Hilton.com and its app leading to confirmed stays in 2022, and that the electric-vehicle charging filter alone resulted in more than 30% of users completing a booking (1). That is a booking-conversion figure the company chose to publish, not evidence that the feature produced long-term loyalty.

Intended loyalty outcome. Reduce search effort and make the brand easier to choose. Over time, a discovery experience that reliably works may pull the traveller’s future searches back to the same brand.

How to measure it. Filter-to-property-view rate, filter-to-booking conversion, search abandonment, cancellation rate, subsequent direct-booking rate among filter users, and repeat use of the same declared preference.

Conversion alone is not enough. A hotel chosen because it appeared to meet a requirement and then failed to meet it produces a booking and damages loyalty at the same time.

Privacy or trust risk. A current-session filter is about as transparent as personalisation gets. The risk arrives when the brand starts silently inferring household circumstances, mobility requirements, religion or family status from past behaviour instead.

Worth naming what this is. A search filter is preference-led relevance rather than algorithmic personalisation: no predicted profile is involved, and the customer selected it themselves. That is a feature, not a shortcoming. Not every personalised experience needs an inference behind it, and explicit preference selection is the most transparent form of relevance available.

Lesson. Start with what the customer has deliberately told you in this journey. Explicit intent is usually more relevant, easier to explain and less intrusive than a hidden prediction.

2. Programme onboarding: make the personalisation choice explicit

Named example: Vodafone VeryMe

Customer context. A mobile customer is joining a rewards experience and needs to understand what value is on offer and what information shapes it.

Data or signal used. Vodafone states that when you first sign up to VeryMe Rewards it asks what you like, that what you tell it helps choose which rewards you get, and that the more you use it the more personalised it becomes. It also states that permission settings can be changed through the My Vodafone app or by logging in (3).

Personalised action. The programme adapts the rewards shown to declared preferences and subsequent interactions, with permission controls the customer can revisit.

Intended loyalty outcome. Make the programme feel relevant from the start, on the basis of something the customer actively said rather than something the operator worked out.

How to measure it. Onboarding completion, consent and permission rates, reward view and claim rates, return visits to the rewards area, permission withdrawal, complaints about relevance or data use, and retention among comparable participants and non-participants.

Reward claims are not proof of loyalty. Someone can claim a free coffee every week and switch network the moment a better tariff appears.

Privacy or trust risk. Asking for broad permissions before demonstrating any value. Or leaving customers unsure whether declining personalisation costs them the whole proposition.

Lesson. Separate participation from personalisation wherever you can. Explain what the preference data buys the customer, and make the choice a real one rather than a consent gate with a decorative “no” button.

3. Second purchase: make familiar buying easier without narrowing choice

Named example: Sainsbury’s Your Nectar Prices

Customer context. A grocery customer has an established purchase history. The retailer can identify what they buy repeatedly and what adjacent products they might reasonably find useful.

Data or signal used. Shopping history within the Nectar relationship.

Personalised action. Sainsbury’s says Your Nectar Prices gives customers up to ten personalised offers each week, combining favourite products with new recommendations based on shopping habits, refreshed every Friday. In July 2025 it reported that more than one million customers were using the feature weekly, that this had produced over £60 million in customer savings in the previous year, and that over 17 billion personalised discounts had been generated since launch (4).

Read the footnotes on that release and you learn something more useful than the headline. The “over £150 per year” saving it quotes is calculated from the average weekly saving offered to customers between April 2024 and April 2025, assuming Your Nectar Prices is used once, multiplied by 52 (4). That is a modelled projection from an offered average, not an observed per-customer outcome. It is a well-documented figure, and documenting it that way is to the retailer’s credit, but a saving that is offered is not a saving that was taken.

Intended loyalty outcome. Make repeat shopping more valuable, and give customers a reason to identify themselves at the till.

How to measure it. Offer activation and redemption, incremental category purchasing against a holdout, eight- or twelve-week repeat-purchase rate, margin after the discount, the share of customers receiving at least one genuinely relevant offer, complaints about fairness, and behaviour after promotional support ends.

An offer redeemed on something the customer would have bought anyway produces a saving and no incremental behaviour at all. The reverse is also possible: a programme can strengthen perceived value while short-term spend stays flat.

Privacy or trust risk. Shopping histories reveal or suggest pregnancy, health conditions, religion, financial pressure and household composition. The danger is at its highest when the retailer lets the inference show in the wording of the offer.

Lesson. Use purchase history to support familiar needs and controlled discovery. Do not confuse redemption with loyalty. Relevance, incremental behaviour, margin and perceived fairness have to be read together.

4. Repeat use: optimise for sustained value, not the next click

Named example: Netflix recommendations

Customer context. A subscriber opens a very large catalogue and needs help finding something worth watching.

Data or signal used. Netflix documents that its recommendation system draws on signals including what members watch, how far they get through titles and their ratings, and that it adapts which titles appear, which rows they appear in and the order within them (5).

Personalised action. Continuous reorganisation of how the catalogue is presented, around the account’s evolving behaviour.

Intended loyalty outcome. Help the subscriber keep finding value, and reduce the effort of navigating a catalogue that is too big to browse.

There is a measurement problem here that Netflix describes openly. Retention is the outcome the business cares about, but it is affected by seasonality and personal circumstances, is mainly sensitive among members already close to cancelling, may follow a run of poor experiences rather than one, and produces only a single signal per account per month. Netflix’s conclusion is that optimising for retention alone is impractical, so it uses more sensitive proxy rewards while keeping retention as the north star (6). That distinction is the whole game. The most clickable recommendation is not necessarily the one that sustains the relationship.

How to measure it. Successful play after opening the service, abandonment without watching, completion and satisfaction signals, breadth of content discovered, return frequency, subscription retention, and performance over a long enough experiment window.

Optimising purely for thumbnail clicks or immediate viewing time rewards curiosity, sensationalism and habit without necessarily improving satisfaction.

Privacy or trust risk. A shared household profile produces inaccurate or embarrassing recommendations. A single historic interaction becomes a permanent label nobody asked for.

Lesson. Weight recent behaviour more heavily, give people a way to correct the profile, and evaluate long-term value rather than immediate response.

5. Loyalty milestone: make progress feel like recognition, not pressure

Named example: Starbucks Rewards tiers

Customer context. A member is approaching or has reached a programme milestone. The brand has to decide how to recognise that without turning the relationship into a permanently moving sales target.

Data or signal used. Stars earned, tier status and eligible activity within the programme.

Personalised action. In January 2026 Starbucks announced a redesigned Rewards programme with three levels, Green, Gold and Reserve, launching in March, describing it as designed to strengthen connection and drive growth. In the same release it reported an all-time high of 35.5 million 90-day active members in the US in its first fiscal quarter of 2026 (7). The growth language is a stated objective, not a demonstrated causal result, and the membership figure covers the whole programme rather than any particular personalised feature.

The UK picture is worth putting next to it. In accounts for the financial year ended 28 September 2025, Starbucks reported that UK Rewards sales were up 45% against the prior year, that Rewards accounted for 42% of total UK sales in FY25 against 31% the year before, and that active membership grew 41% to 2.4 million. The same filing reported an operating loss of £29.8 million for Starbucks Coffee Company (UK) Limited, against £27.5 million the year before, and described a tougher and more competitive market (8).

Those numbers sit in the same document, and that is the point. A loyalty programme can be pulling a rapidly growing share of sales through a channel the company controls while the business underneath it is losing money. Programme engagement is a measure of programme engagement. It is not a measure of business health, and reporting it as though it were is one of the most common mistakes in loyalty marketing.

Intended loyalty outcome. Make recognition visible, encourage continued participation, and demonstrate that the customer gets more as the relationship develops.

How to measure it. Analyse behaviour around the milestone rather than comparing all members with all non-members: progress towards the threshold, incremental visits before and after qualification, reward cost, benefit utilisation, sustained activity after the milestone, drop-off among customers who narrowly miss it, and complaints about changing thresholds or devalued benefits.

Privacy or trust risk. Progress messages turn into pressure messages. A customer feels manipulated when the brand keeps stressing how little time remains, or moves the qualification criteria after they have already invested.

Lesson. A milestone should make recognition clearer, not manufacture anxiety. Measure whether the benefit produces sustained value after qualification, not just a burst of activity before it.

6. High-value recognition: provide certainty at a moment that matters

Named example: Hilton Honors upgrades

Customer context. A frequent guest is preparing for a stay. Recognition means most when it changes something they actually care about, and room upgrades have historically been the least certain benefit in hotel loyalty.

Data or signal used. Eligible status, qualifying nights, the reservation and available inventory.

Personalised action. Hilton has described making upgrade options visible to eligible members during digital check-in, alongside a new benefit it calls Confirmable Upgrade Rewards, which lets eligible members lock in a guaranteed upgrade of up to a one-bedroom suite at the time of booking, for stays of up to seven nights, on both paid and reward stays at participating properties (9). Members receive their first on reaching the programme’s new top tier, with the option of a second at a 120-night milestone (10).

Intended loyalty outcome. Make status tangible, remove uncertainty, and show that continued engagement produces a benefit you can actually plan around.

How to measure it. Upgrade availability and acceptance, satisfaction among eligible customers who do and do not receive one, repeat direct booking, service contacts about upgrade eligibility, perceived fairness, and incremental retention against comparable high-value customers.

Privacy or trust risk. Recognition backfires when treatment is inconsistent. Tell a customer their status is valuable and then repeatedly fail to deliver the benefit, or push a paid upgrade harder than the one their status promised, and you have made the relationship worse than saying nothing.

Lesson. Recognition is not a personalised message. It is reliable fulfilment, transparent eligibility, and a credible fallback when the preferred benefit is not available.

7. Ongoing service: personalise assistance, not just promotions

Named example: Bank of America’s Erica

Customer context. A banking customer has recurring questions and could benefit from timely information about transactions, balances, subscriptions or account management.

Data or signal used. Account activity, transaction patterns, recurring charges, and interactions with the assistant itself.

Personalised action. Bank of America reported in August 2025 that Erica had assisted nearly 50 million users since launch, surpassed 3 billion client interactions, and was averaging more than 58 million interactions per month, with clients having received more than 1.7 billion proactive personalised insights. The examples it gave include flagging which way balances are trending over the next seven days, highlighting cash-back deals based on spending, and notifying clients of rewards eligibility (11).

Those figures demonstrate adoption at very large scale. They do not demonstrate that the service caused retention or loyalty, and the bank does not claim they do.

Intended loyalty outcome. Make the banking relationship more useful, reduce avoidable effort, and build a habit around the provider’s digital service rather than a competitor’s.

How to measure it. Successful task completion, reduction in repeated service contacts, escalation to a human adviser, accuracy and helpfulness ratings, action taken after an insight, opt-outs and complaints, and account attrition among comparable exposed and unexposed customers.

Privacy or trust risk. Financial data is about as sensitive as it gets. A mistimed or inaccurate insight creates anxiety, exposes information on a shared screen, or simply makes someone feel watched by their bank.

Lesson. In a high-trust sector, utility and customer control come before promotional ambition. Explain the basis for an insight where you can, and make it easy to dismiss, correct or switch off.

8. Advocacy: turn personal data into customer-controlled expression

Named example: Spotify Wrapped

Customer context. A listener reaches the end of the year with a history of songs, artists and genres that can be turned into a personal retrospective.

Data or signal used. Eligible annual listening history, rendered as top artists, songs and genres, in cards designed to be shared.

Personalised action. Behavioural data converted into a visual story the customer can explore and choose to distribute.

Spotify reported that its 2025 Wrapped campaign finished with more than 300 million engaged users and more than 630 million shares on social media globally, across 56 languages (12). Those are participation and distribution figures, not independently verified referral or retention effects.

Intended loyalty outcome. Make the service part of how the customer describes themselves to other people. Unlike a referral request, the shareable object is meant to have value to the customer in its own right, which is why they distribute it without being asked.

How to measure it. Experience completion, voluntary share rate, referral traffic and attributed trials, subsequent engagement, retention over an appropriate period, negative feedback and correction requests, and performance against historical or experimental comparisons where that is feasible.

A share is an advocacy signal. It is not automatically a recommendation, and it is certainly not a new customer.

Privacy or trust risk. Listening histories reveal sensitive interests, and shared accounts produce summaries that are wrong or embarrassing. A retrospective the customer does not recognise undermines confidence in everything else the service claims to know about them.

Lesson. Advocacy personalisation works when the customer controls the disclosure and the artefact helps them express something about themselves. Do not auto-publish, do not preselect an audience, and do not make sharing a condition of receiving the value.

Examples of personalisation that can damage loyalty

Personalisation reduces loyalty when it makes your use of customer information more visible than the value it creates.

Excessive frequency

Poor example. A customer views a product once and then receives a sequence of emails, app notifications, display adverts and text messages about it.

Better practice. Apply a contact policy across channels, suppress after purchase, and reduce frequency after non-response.

What to measure. Complaints, unsubscribes, notification disabling, conversion after each additional contact, and total incremental response rather than attributed conversions.

The ICO’s guidance is explicit that people can object to direct marketing, that you must stop when they do, and that it is better to suppress details than delete them so that a preference is not accidentally overwritten later (13).

Inappropriate inference

Poor example. A retailer implies from purchasing patterns that a customer is pregnant, unwell or under financial pressure.

Better practice. Prefer explicit preferences, avoid unnecessary sensitive inference, and ask whether the message would surprise or distress the person receiving it.

What to measure. Complaints, opt-outs, message-level unsubscribes by campaign type, and qualitative feedback.

This is the gap that would genuinely surprise most shoppers. In my experience people broadly assume a brand records what they buy. What they have not thought about is what gets derived from it: when their salary arrives, how many people they are shopping for, which triggers move their decisions. Purchase history is not a list of products. It is a behavioural model of a household, and a personalised offer is the moment that model becomes visible to the person it describes.

Aguirre and colleagues found that when firms collected information overtly, more personalised advertising produced greater click-through intentions, while covert collection increased customers’ feelings of vulnerability and lowered response. Trust-building cues could offset that effect (14). Overtness is not a compliance detail. It changes whether the personalisation works at all.

Inaccurate recommendations

Poor example. A shared household account is repeatedly served recommendations based on one person’s behaviour, with no way to correct them.

Better practice. Weight recent behaviour more heavily, allow customers to remove items or signal “not interested”, and provide profile controls.

What to measure. Recommendation rejection, profile resets, repeated exposure to rejected categories, and long-term satisfaction rather than clicks.

Inconsistent treatment across channels

Poor example. An email promises a personalised benefit that a shop assistant, hotel desk or contact-centre colleague cannot see or honour.

Better practice. Confirm eligibility before you send, and give the customer a clear fallback when fulfilment is uncertain.

What to measure. Failed redemptions, service contacts, manual adjustments, complaints, and subsequent purchasing among affected customers.

Manipulative urgency

Poor example. A personalised countdown or “only for you” scarcity claim that does not reflect a real deadline or a real limitation.

Better practice. State the actual basis and duration of an offer. Do not use what you know about someone to intensify pressure on them.

The Competition and Markets Authority has issued compliance advice to online businesses on urgency claims, meaning scarcity, popularity, “act fast” or time-limited claims, and on price reduction claims. Its position is that these have a legitimate place when they alert consumers to genuine offers, but that claims which mislead or apply unfair pressure may be illegal under consumer protection law (15). Personalisation makes this worse rather than better, because a pressure tactic aimed precisely at a known weakness is more effective and less defensible than a generic one.

Unclear data use

Poor example. A message is strikingly specific and the customer cannot work out how you knew, or how to make it stop.

Better practice. Give a concise explanation at the moment it is relevant, link to fuller information, and provide accessible controls.

The ICO’s guidance requires that where you obtain information about people, including through profiling their interests and habits, the processing is fair and you tell them about it in language they can understand (2).

Highly personalised messages can also provoke reactance when the recipient cannot see why that level of personalisation is justified. White and colleagues found less favourable responses under exactly those conditions, particularly where the perceived utility of the offer was low (16).

What the strongest examples have in common

The eight vary enormously by sector and customer stage, but the ones that work share five characteristics.

The signal is relevant to the immediate context. A filter selected during a hotel search is easier to justify than an unexplained prediction derived from unrelated browsing.

The action solves a customer problem. Setup help, easier discovery, a timely service insight. A message is not useful because it contains someone’s name.

The intervention is measured over an appropriate period. Clicks, claims and redemptions are diagnostics. Loyalty needs evidence of continued value, retention, repeat behaviour or advocacy.

Customers keep meaningful control. They can correct a recommendation, change a preference, decline personalisation, or choose not to share a personalised result.

The brand has thought about how the action feels once its use of data becomes visible. Bleier and Eisenbeiss found that personalisation effectiveness depends on what is personalised, when it appears and where it is delivered, and that higher personalisation can work well early while decaying faster and becoming intrusive sooner (17).

There is a sixth thing, and it is the one that trips up the measurement rather than the experience. People who join your loyalty programme, download your app or open your Wrapped were disproportionately your most engaged customers before any of it existed. I spent years watching brands celebrate app metrics that comfortably beat their website metrics, and the explanation was almost never the app. If someone downloads your app, they were already loyal. The programme collected the behaviour; it did not create it. Every participation figure quoted in this article carries some version of that problem, which is why holdouts and comparable non-participants matter more than the headline number.

So the practical question is not “can this interaction be personalised?” It is this: does the signal justify the action, does the action help the customer, and can we show a better relationship without spending trust to get it?

The psychology and evidence behind all of this sit in the brand loyalty pillar. The segmentation, rewards, triggers and operating processes sit in the loyalty programme guide.

Frequently Asked Questions

References

1. Hilton. How Hilton’s tech innovations deliver frictionless travel, Stories From Hilton, 17 January 2023. https://stories.hilton.com/innovation/hiltons-consumer-centric-tech-innovations-delivering-frictionless-travel-experience
2. UK Information Commissioner’s Office. Collect information and generate leads, direct marketing guidance. Accessed 27 July 2026. https://ico.org.uk/for-organisations/direct-marketing-and-privacy-and-electronic-communications/direct-marketing-guidance/collect-information-and-generate-leads/
3. Vodafone. VeryMe Rewards. Accessed 27 July 2026. https://www.vodafone.co.uk/business/vodafone-veryme-rewards
4. Sainsbury’s. Unlock, shop and save: Sainsbury’s rolls out Your Nectar Prices across tills nationwide, 16 July 2025. https://corporate.sainsburys.co.uk/news/press-releases/unlock-shop-save-sainsburys-rolls-out-your-nectar-prices-across-tills-nationwide/
5. Netflix. How Netflix’s recommendations system works, Netflix Help Center. Accessed 27 July 2026. https://help.netflix.com/en/node/100639
6. Netflix Technology Blog. Recommending for Long-Term Member Satisfaction at Netflix. 29 August 2024. https://netflixtechblog.com/recommending-for-long-term-member-satisfaction-at-netflix-ac15cada49ef
7. Starbucks. Starbucks unveils reimagined loyalty program to deliver more meaningful value, personalization and engagement for members, 29 January 2026. https://about.starbucks.com/press/2026/starbucks-unveils-reimagined-loyalty-program-to-deliver-more-meaningful-value-personalization-and-engagement-to-members/
8. Starbucks. Starbucks files UK and EMEA accounts for the fiscal year ended September 2025, Starbucks Stories EMEA. https://stories.starbucks.com/emea/stories/2026/starbucks-uk-and-emea-accounts-fiscal-year-ended-september-2025/
9. Hilton. Hilton Honors redefines the room upgrade experience for members with more choice and flexibility, Stories From Hilton, 8 June 2026. https://stories.hilton.com/hilton-honors/hilton-honors-redefines-the-room-upgrade-experience-for-members-with-more-choice-and-flexibility
10. Hilton. Loyalty upgraded: Hilton Honors introduces faster path to elite status and reveals new premium tier, Diamond Reserve, Stories From Hilton, 18 November 2025. https://stories.hilton.com/releases/loyalty-upgraded-hilton-honors-introduces-faster-path-to-elite-status-and-reveals-new-premium-tier-diamond-reserve
11. Bank of America. A decade of AI innovation: BofA’s virtual assistant Erica surpasses 3 billion client interactions, 20 August 2025. https://newsroom.bankofamerica.com/content/newsroom/press-releases/2025/08/a-decade-of-ai-innovation—bofa-s-virtual-assistant-erica-surpas.html
12. Spotify. Spotify reports fourth quarter 2025 earnings, 10 February 2026. https://newsroom.spotify.com/2026-02-10/spotify-q4-2025-earnings/
13. UK Information Commissioner’s Office. Respect people’s preferences, direct marketing guidance. Accessed 27 July 2026. https://ico.org.uk/for-organisations/direct-marketing-and-privacy-and-electronic-communications/direct-marketing-guidance/respect-peoples-preferences/
14. Aguirre, E., Mahr, D., Grewal, D., de Ruyter, K. and Wetzels, M. Unraveling the Personalization Paradox: The Effect of Information Collection and Trust-Building Strategies on Online Advertisement Effectiveness. Journal of Retailing, 91(1), 34-49, 2015. https://doi.org/10.1016/j.jretai.2014.09.005
15. Competition and Markets Authority. Urgency claims and price reduction claims: compliance advice for online businesses, 29 March 2023. https://assets.publishing.service.gov.uk/media/64232b153d885d000cdadd30/OCA_business_open_letter_FINAL.pdf
16. White, T. B., Zahay, D. L., Thorbjørnsen, H. and Shavitt, S. Getting Too Personal: Reactance to Highly Personalized Email Solicitations. Marketing Letters, 19(1), 39-50, 2008. https://doi.org/10.1007/s11002-007-9027-9
17. Bleier, A. and Eisenbeiss, M. Personalized Online Advertising Effectiveness: The Interplay of What, When, and Where. Marketing Science, 34(5), 669-688, 2015. https://doi.org/10.1287/mksc.2015.0930