top of page
Oct 6
10 min read

Updated: 10 hours ago

Browse guides: Pricing · Hiring · Audit · Klaviyo agencies · Apparel · Footwear

Progressive Personalisation: How Customer Journeys Change After Each Purchase

By Trayan Hristov | Updated 13 November 2026 | 10 min read

Thrivelia guide: Progressive Personalisation: How Customer Journeys Change After Each Purchase

Progressive personalisation means using what you learn about a customer to change the next useful message, product suggestion or service interaction. A new subscriber, a first-time buyer and a regular customer should not remain in the same journey simply because they share a segment label.

At Thrivelia, we use the term for a lifecycle approach that develops with the relationship. The objective is a better next decision, supported by reliable information and a clear customer need. Adding a first name to a campaign is only a small part of that work.

What makes personalisation progressive?

A static segment describes a customer at one point: interested in a category, located in a market or assigned to a spending band. Progressive personalisation asks what changed and what that change means for the next interaction.

The customer might answer a preference question, buy a first product, receive the order, need help, return an item or purchase again. Each event can change the appropriate message. The programme should use the information it actually has, rather than pretend to know a motivation or a product experience it cannot observe.

Relationship stage

What the brand may know

How the next interaction can change

New subscriber

Stated interest, source or an answered preference

Introduce relevant products and the information needed to choose

First-time buyer

Purchased product, order context and available delivery information

Replace acquisition messages with useful expectations and product guidance

Customer adopting the product

Available usage, service or feedback signals

Offer appropriate help and identify a sensible next step

Returning buyer

Actual purchase sequence and time between orders

Recommend a more informed next product or use case

Established customer

Reliable history, preferences and engagement

Recognise the relationship and tailor relevant opportunities

These are examples of available signals and possible actions, not a claim that every account tracks every stage automatically.

Order count changes the question

Before the first purchase, the question is often whether the product and brand deserve trust. After the first purchase, the question becomes whether the customer gets the expected value. After another order, the programme can learn more about the products and occasions that matter to that person.

Order count is a useful organising variable, but it is not a complete customer model. Two orders placed years apart differ from two orders placed within a short product cycle. A return, an exchange or a subscription rebill can also require different interpretation. Combine the count with the product, timing and context that can be established reliably.

Avoid turning a sophisticated-looking score into unexplained certainty. If a segment identifies a valuable customer, define what value means, which data informs it and what action follows. Keep the criteria understandable to the people managing the account.

Worked example: how we designed Scent Intelligence for ESAS


When we worked through ESAS’s fragrance data, I wanted to answer a practical question: once someone tells you what they like, what should you actually do differently? A quiz result is a starting point. The useful part is remembering it when the customer samples, buys, wears and eventually chooses another fragrance.


We combined quiz answers, Klaviyo customer profiles, first-product purchase cohorts, Hiro lifecycle performance and Outersignal’s customer-persona analysis. That became the Scent Intelligence strategy: a way to connect the product someone might enjoy with the message that helps them choose it. The analysis and journey design are the work described here; the proposed flows and scoring model are not presented as proven revenue improvements.


Watch: The Mathematics of Desire


This 4-minute walkthrough illustrates the strategy and proposed customer journey. The Scent DNA score is an illustrative matching framework, not a scientifically validated model. The video depicts the intended experience; it does not establish that every element has been deployed.



Two numbers need context. The 98.8% engagement figure means quiz starters who answered at least one question. The 77.8% figure is the share of eventual quiz conversions occurring within 24 hours, not a 77.8% purchase rate among all quiz takers. Figures in this example come from the August 2026 analysis.


1. Separate what customers tell us from what we infer


The quiz captured desired feeling, preferred note families, intensity, wearing occasion and visual cues. Someone could choose Comforting / Cozy, gourmand notes and a subtle, close-to-skin fragrance. Those answers could change the product selection, the description of how it wears and the image used to introduce it.


Outersignal supplied a different view: modelled customer types and their associated purchase economics. We used that to prioritise audience and creative hypotheses. It could suggest whether formulation proof, everyday wearability or sensory storytelling deserved more attention. It could not tell us exactly why an individual bought.


The person-level Outersignal field was present on roughly 6% of the full quiz export and about 14% of current-schema profiles, with coverage concentrated among buyers. That ruled out making persona labels the foundation of every journey. Quiz and purchase data were the default; persona enrichment could sharpen the proof angle where it was available.


2. Compare preference with customer value


We checked which answers were associated with conversion, buyer value and repeat purchase. Comforting / Cozy stood out in the current-schema analysis: $66.93 lifetime revenue per quiz profile, including non-buyers at zero, and $213.63 average lifetime revenue among buyers. Those are different denominators. Neither is a forecast for the next person who selects that answer.


We also kept quiz-recommendation cohorts separate from first-purchased-product cohorts. Being recommended Cherry Cloud does not mean someone bought it first. Likewise, a product with high 365-day customer revenue can owe much of that value to an expensive first order. We examined value added after day 30 to understand the later relationship.


This helped us choose what to test. A smaller group could justify more attention because of its observed economics, while the largest audience still mattered for scale. The analysis did not prove that changing an email to match a preference would cause higher LTV.


3. Turn the answers into a recommendation people can understand


We proposed a 100-point Scent DNA matching score: 60 points for note families, 15 for desired feeling, 10 for visual cues, 10 for occasion and 5 for energy or identity. These were transparent starting weights to test and refine against clicks, purchases and later feedback. A score of 80 would mean a match under those rules, not an 80% probability of liking a fragrance.


The customer-facing explanation matters as much as the score: “You chose warmth, gourmand notes and close-to-skin wear. Here is why this fragrance fits.” Visual answers such as white curtains or warm almond pastry could guide imagery and language. They should not override product fit or become claims about universal psychology.


4. Change the journey when the customer learns something new


For a new lead, the proposed immediate sequence explains the result, why the product fits and how to choose between finalists. The first-day conversion pattern gave us a reason to put useful guidance early. It did not justify sending the entire sequence in one day.


For an existing customer taking the quiz, the recommendation should account for fragrances already owned. The useful question becomes which occasion, mood or season the next scent could serve. Recommending the same bottle again would waste the history we already have.


For a Discovery Set buyer, the next job is helping them try the samples properly. We designed wear guidance first, then one-click favourite capture, followed by a short calibration quiz around 7 to 12 days after delivery. It asks which samples they reached for again, which felt too strong and which occasions they actually wore them for.


That creates a better recommendation: “Before wearing them, your answers pointed to X. After trying them, you kept reaching for Y.” If the favourite stays the same, the message can explain the consistency. If it changes, the customer’s experience should update the recommendation. The planned conversion email presents one finalist and one runner-up, with any sample-set credit explained only where confirmed.


For a full-size owner, we designed a further change: first-wear guidance, help building a habit, specific review prompts and then a second-scent recommendation. The second bottle should fill a different job, such as an evening fragrance alongside a daily favourite. Owned products and observed co-purchase patterns help narrow that choice.


5. Keep the customer memory detailed and the flow structure manageable


We recommended normalising raw quiz answers into stable Klaviyo fields for feeling, occasion, intensity, recommendations, sample favourites and owned fragrances. A new answer should refresh the relevant preference. Old tags should not keep sending someone down a journey that no longer fits.


The architecture used four core journeys: new quiz leads, existing customers retaking the quiz, Discovery Set calibration and full-size owners. Reusable modules could then carry the customer’s context into welcome, browse, cart, checkout and campaigns. We did not need a separate flow for every answer combination.


The operating rules were part of the design: marketing consent, deduplicated quiz events, purchase exits, owned-product exclusions and priority between overlapping journeys. Missing persona data needed a coherent fallback. A Discovery Set buyer needed sampling guidance instead of another broad pitch for the kit.


What we would measure


The design calls for first-purchase conversion within a consistent window, Discovery Set-to-full-size conversion, revenue per eligible profile and 60- or 90-day second-order rates. Where volume permits, holdouts can test whether the new journey adds orders. Compare cohorts with equal time to return and account for offer cost and returns.


That is the progression I want from personalisation: what someone says helps the first decision; what they experience improves the next one; what they buy changes what we recommend afterwards.


Source note: ESAS Scent Intelligence strategy, 21 August 2026, drawing on Visual Quiz Builder analytics, a 21,917-profile Klaviyo export, Outersignal enrichment, first-purchased-SKU cohorts and Hiro performance data. The example explains the analysis and recommended architecture, rather than claiming a measured uplift from the full proposed system.


An illustrative apparel journey

Consider a fictional first-time customer buying a core top. Before purchase, a useful message might explain fabric, fit and how the item works within the collection. After purchase, it might provide care guidance and styling ideas. A later message could introduce a companion product, if the customer's context and the catalogue make that relevant.

If the customer buys the companion product, the programme should change again. It can stop presenting the same introductory argument and use the developing purchase history to suggest a different use case or collection.

This is a hypothesis to test, not a claimed result from a specific account. The perfectwhitetee case provides a relevant analytical lesson: compare customer tenure fairly when evaluating second purchasing. The product path becomes more informative when the populations have had equal time to return.

An illustrative footwear journey

A fictional first-time footwear buyer might need help choosing size or width. After delivery, they may need fitting guidance, care advice and a route to support. If a service issue is visible, the next useful interaction may be help rather than another promotion.

Once the first pair has earned its place, the next purchase could serve a new occasion or use case. A returning customer then receives a different selection from someone still deciding whether to trust the first pair.

The VIBAe case connects product education, post-purchase support, loyalty and the next order. Read it alongside our footwear email marketing guide for the operational questions behind that journey.

Personalisation requires reliable handoffs

The flow architecture should change with the customer. Check what happens when someone subscribes, browses, starts checkout and purchases before an earlier sequence finishes. Inspect returning buyers and support contexts as well as the simple first-order path.

Personalised content cannot compensate for unreliable entry and exit logic. A buyer still receiving first-purchase recovery messages has a routing problem, even if each message contains an accurate product recommendation.

Use the Klaviyo audit checklist to review events, eligibility, exclusions and re-entry. Document the information needed for each decision, its source and the fallback when the field is absent or uncertain.

Start with a small set of meaningful signals

Choose information that changes an action. Purchased product, order count, a stated category preference and an observed time to another order may be more useful than a large profile that nobody maintains.

For each proposed field, ask three questions: is it reliable, is it appropriate to use for this interaction, and what would change if we had it? If the action stays the same, the field may not belong in the first implementation.

Define fallbacks. A message should remain coherent when the preference is missing, a product is unavailable or an event arrives late. Review the experience with sample customer histories before expanding the logic to a larger audience.

Measure the customer transition

Personalisation should be evaluated against the decision it was intended to improve. A welcome experiment may focus on qualified first purchases. An adoption journey may focus on a defined product or customer outcome. A second-purchase test needs an eligible first-order cohort and a consistent observation window.

Proposed change

Useful primary question

Supporting checks

Preference-specific welcome

Do eligible subscribers make the intended first purchase?

Audience balance, delivery and offer cost

Product-specific post-purchase

Does the guidance improve the defined customer outcome?

Timing, support context and actual exposure

First-product cross-sell

Do more first-time buyers return within the chosen window?

Contribution, returns and competing messages

Returning-customer selection

Does the new approach improve the chosen repeat behaviour?

Cohort age, product availability and frequency

Separate a plausible hypothesis from an observed result. Where a comparison is feasible, define it before launch and record the limits of the finding. Attributed flow revenue helps evaluate the message programme; it does not by itself establish that personalisation caused additional orders.

How this fits the Marketing Hourglass

The Marketing Hourglass makes the relationship visible from Know through Advocate. Progressive personalisation describes how the next interaction can change as a customer moves through that relationship.

The two ideas serve different purposes. The Hourglass identifies where the brand needs to help. Progressive personalisation connects the available customer context with what to do next. Together, they support a programme that continues to develop after the first order.

Test the next customer decision with your team

Start with one signal the customer can correct, one decision it changes and a fallback for everyone who skips it.

Frequently asked questions

Is progressive personalisation just segmentation?

Segmentation is one tool. The broader approach also changes timing, journey logic, product advice and service routing as the relationship develops. The action matters more than the number of segments.

Do we need AI to do this?

No. Start with reliable data and a small set of clear decisions. AI can assist appropriate analysis or content work, but it does not replace product truth, sound journey logic or human judgment about what the customer needs.

How much customer data is enough?

Enough to make the next interaction meaningfully better. Begin with fields you can maintain and use. Expand when a specific decision requires more information and the collection and use are appropriate.

How do we know whether it works?

Define the customer transition, the eligible audience and the observation window. Compare outcomes using a suitable design, then inspect relevant commercial and experience guardrails. Read our first-to-second-purchase guide and email revenue share guide for measurement examples.

Build the next stage of your customer journey

Pick one piece of information the customer has volunteered. Write the two messages you would send depending on their answer. If both messages lead to the same product, claim and next step, leave the question out of the form.

 
 
 

Recent Posts

See All

Comments


bottom of page