A portable “shopping passport” promises to let customers carry useful preferences and purchase context between retailers, giving personalisation systems a richer starting point than a new visitor’s sparse browsing history. For e-commerce and omnichannel leaders, the relevant question is not whether this sounds convenient. It is whether an external identity layer can generate enough incremental commercial value to justify integration costs, privacy exposure and dependence on another platform.
EU-Startups reports that Stockholm-based eComID has raised €14.6 million to develop a shopping passport for AI-powered personalised commerce, with strategic backing involving retailers including H&M Group and Stadium. That participation makes the proposition operationally noteworthy: established retailers appear interested in exploring identity-based personalisation beyond their own customer accounts. It does not, however, demonstrate higher conversion, lower returns or durable customer adoption. Those remain hypotheses that each operator should test.
What a portable shopping passport changes
Conventional retail personalisation is usually bounded by one company’s data. A retailer may connect an account, loyalty record, order history, product views and store transactions, but it knows little about a new customer and loses continuity when identification fails across devices or channels.
A portable profile proposes a different arrangement. With the customer’s permission, an identity layer could make selected shopping information available across participating merchants. Depending on the eventual product design and permissions, useful signals might include brand or category preferences, sizing information, prior purchases, fit feedback or other declared preferences. The source describes eComID’s ambition, but retailers should confirm exactly which attributes are stored, inferred, shared or merely referenced before evaluating the business case.
The most plausible benefit appears early in discovery. A participating retailer could potentially offer more relevant product ranking before accumulating extensive first-party behaviour. Continuity could also improve when a shopper moves from mobile research to desktop checkout or visits a physical store. For repeat customers, portable information might supplement rather than replace loyalty and transaction data.
There is an important limit: identity is not intent. A profile built from historical choices may miss a gift purchase, a changed budget, a new size or an unfamiliar category. Retailers should therefore treat passport data as one input to ranking and recommendations, not as an authoritative description of what the person wants today.
Where the commercial case may—and may not—work
The opportunity is strongest when the retailer’s current personalisation suffers from sparse data and when portable attributes materially narrow the assortment. Fashion is an intuitive test environment because size, fit and style preferences can influence both discovery and purchase confidence. Strategic interest from H&M Group and Stadium is relevant in that context, although backing should not be interpreted as proof of production-scale results.
A passport may add less value where purchases are infrequent, preferences change rapidly or the assortment offers little meaningful personalisation. It can also be redundant for a retailer with high login rates, mature loyalty coverage and strong first-party recommendations. In those circumstances, the external layer must outperform data the operator already controls.
Retail leaders should formulate the investment thesis by customer segment. New anonymous visitors, recognised non-buyers, occasional customers and loyal members begin with different data gaps. If a pilot mixes them together, an aggregate conversion result may conceal both a valuable use case and a harmful one.
Returns require particularly careful analysis. Better sizing or preference signals could plausibly reduce avoidable returns, but more persuasive recommendations might also increase ordering volume or encourage experimentation. The pilot should measure return rate, return reasons and retained margin by cohort rather than assuming that personalisation automatically improves product fit.
The integration burden behind the customer experience
A useful passport cannot sit as an isolated sign-in button. It needs controlled connections to the systems that determine what the customer sees and what the retailer can fulfil.
- Identity and access management: account matching, authentication, duplicate resolution and rules for linking an external identity to existing customer records.
- Consent management: auditable records of which attributes the customer authorised, for which purposes, for how long and how withdrawal propagates.
- Customer data and loyalty platforms: reconciliation between passport attributes and first-party purchase, preference and membership data.
- Product information: consistent taxonomy, sizing, fit, brand and category metadata so external preferences can map to the retailer’s catalogue.
- Search and recommendation services: decision rules for using portable signals, including confidence thresholds and fallbacks when data is missing or contradictory.
- Commerce and store systems: web, app, point-of-sale, clienteling, order and returns connections if the proposition is genuinely omnichannel.
- Analytics infrastructure: cohort assignment, exposure logging and outcome measurement needed to establish incrementality.
Data quality is likely to determine more value than the novelty of the identity layer. “Medium” in one brand may not correspond to another; taxonomies and fit conventions differ; inferred preferences can become stale. Operators need provenance, timestamps and confidence indicators, plus rules specifying whether first-party observations, customer declarations or passport attributes take precedence.
Consent must be part of the product, not a compliance screen
Portable profiles introduce a sensitive trust exchange: customers disclose information because they expect less effort and better choices. Broad or ambiguous consent can undermine that exchange even if the implementation satisfies formal requirements.
The customer should be able to understand which data will move, which merchant receives it, why it is needed and how to revoke access. Granular controls are preferable to a single permission covering every possible personalisation purpose. Retailers also need clear responsibility for access, correction, deletion, retention, security incidents and responses to data-subject requests. Legal and privacy teams should validate the applicable obligations rather than relying on the vendor’s positioning.
Adoption is itself a commercial constraint. Customers may not recognise the provider, may see little benefit before sharing data or may prefer an existing retailer account. A pilot should test the wording and value exchange without using coercive defaults. Consent acceptance alone is insufficient: measure completed setup, successful matching, subsequent reuse and revocation.
Poor personalisation can damage trust more visibly than no personalisation. Provide a way to correct a size or preference, explain why a recommendation appeared where appropriate, and retain a non-passport experience. Sensitive inferences or opaque automated decisions should receive additional review.
Control, interoperability and third-party dependence
A shared identity network becomes more useful as participation grows, but that network effect can increase switching costs. Retailers should examine whether profiles and consent records can be exported in usable formats, whether identifiers work with alternative providers, and whether recommendation logic remains inside the retailer’s own stack.
Contract and architecture reviews should cover data ownership, permitted secondary use, model-training rights, sub-processors, service levels, breach obligations, pricing changes and termination assistance. The operator also needs a contingency plan for vendor failure or strategic change. The storefront should fall back gracefully to first-party or contextual personalisation rather than losing core functionality.
Interoperability should be demonstrated, not inferred from the word “portable.” Ask whether portability applies across participating retailers only, how catalogue attributes are normalised, and whether customers can transfer their profile elsewhere. Avoid sending the full customer record when a narrow, purpose-specific attribute or query would suffice.
A limited pilot with explicit go/no-go thresholds
Start with one market, one channel or journey, and a category where passport attributes could plausibly improve selection. Compare eligible customers offered the passport with an appropriate control, while separately analysing people who adopt it. That distinction reduces self-selection bias: willing adopters may already be more engaged than average shoppers.
Before launch, record baseline performance and agree thresholds for:
- offer visibility, setup completion, successful identity matching and reuse;
- consent withdrawal, preference correction and support contacts;
- product discovery, recommendation engagement and add-to-basket rate;
- incremental conversion and revenue or contribution per eligible visitor;
- repeat-purchase rate over a suitable observation window;
- return rate, retained-item rate and return reasons;
- integration, licence, maintenance, privacy and customer-support cost;
- latency, availability, matching errors and data-quality exceptions.
Proceed beyond the pilot only if the passport produces incremental value after implementation and operating costs, customers repeatedly use it, and privacy and reliability indicators remain acceptable. Pause if gains exist only among a narrow self-selected cohort with no scalable acquisition path. Stop if matching errors, revocations, support demand or dependency costs outweigh the improvement.
eComID’s funding and retailer backing make the shopping-passport model worth structured evaluation, not automatic adoption. The responsible decision is a reversible pilot that tests a precise customer problem while preserving first-party capabilities and an exit route. Until controlled results establish adoption, commercial uplift and trust, the promised benefits should remain labelled as claims rather than outcomes.
