RETAILTECH & ECOMMERCE

eCommerce funding snapped back hard in 2024. Investors are concentrating on tech-led models, not brand plays. The technical differentiation is harder to see from the outside.

RetailTech companies compete on conversion, personalisation, and operational efficiency. Those advantages live in the data architecture, the AI layer, and the ability to perform at peak load — not in the feature list. A pitch deck that shows conversion uplift numbers doesn't tell you whether those numbers hold at scale, or whether the underlying system will survive Black Friday.

Request an X-Ray See example report

Why it’s different

Retail runs at peak loads that expose every technical weakness simultaneously.

RetailTech is one of the few sectors where technical failure is immediately visible to the customer's end consumers — and where a single peak load event can directly cost the customer revenue. The architecture that handles 1,000 transactions per hour will not handle 50,000 without explicit engineering to support it. Personalisation AI that works on 100,000 products will not scale without architectural changes to 10 million. These are the failure modes that determine whether a retailtech company retains its enterprise customers after the first peak season.

Peak load architecture is the most important technical variable in a retailtech investment

Retail operates on seasonal peaks — Black Friday, Cyber Monday, Christmas, back-to-school — that can represent 20–40% of annual transaction volume concentrated in hours or days. A retailtech platform that has never been stress-tested at projected customer peak loads is carrying performance risk that will surface at the worst possible moment: during a customer's highest-revenue period. Architecture that 'scales automatically' on paper may have database connection pool limits, session storage bottlenecks, or third-party API rate limits that create degradation at exactly the wrong time.

AI personalisation is often the core commercial claim — and the least independently validated

Conversion rate improvement, basket size uplift, and customer lifetime value impact are the metrics that sell retailtech products. They are also the metrics most susceptible to attribution errors, benchmark manipulation, and favourable test conditions. A/B test designs that compare AI-powered personalisation against a no-personalisation baseline rather than against a competitive product are not evidence of differentiation. AI models trained on a single large retailer's data may not perform equivalently for a new customer with different product categories. We evaluate the evidence behind the commercial claims.

Merchant and retailer data creates GDPR obligations that are frequently under-managed

RetailTech platforms process consumer purchase behaviour, browsing patterns, and personal preferences on behalf of their merchant customers. This creates a data controller/processor relationship with GDPR obligations that are often not fully operationalised: data retention policies, consumer rights request handling, cross-border data transfer compliance, and consent management for personalisation. Consumer data processed for personalisation has specific legal basis requirements, and 'legitimate interests' claims for behavioural tracking are being challenged with increasing frequency by EU data protection authorities.

Assessment Areas

Where we focus in RetailTech & eCommerce engagements.

Area

What we assess

What it means for the investment

Peak load architecture & stress testing

Load test results, database scaling design, CDN strategy, third-party API rate limits

Whether the platform can survive a peak load event without degradation — and whether it has actually been tested at that load

Peak load architecture & stress testing

Load test results, database scaling design, CDN strategy, third-party API rate limits

Whether the platform can survive a peak load event without degradation — and whether it has actually been tested at that load

AI personalisation & recommendation

Model architecture, training data quality and volume, attribution methodology, A/B test design

Whether the conversion uplift claims are real, reproducible, and generalisable to new customer segments

AI personalisation & recommendation

Model architecture, training data quality and volume, attribution methodology, A/B test design

Whether the conversion uplift claims are real, reproducible, and generalisable to new customer segments

Integration architecture with eCommerce platforms

Shopify, Salesforce Commerce Cloud, Magento, SAP Hybris connector depth

Whether customer acquisition requires a bespoke integration project — and whether the connector library covers the target customer's stack

Integration architecture with eCommerce platforms

Shopify, Salesforce Commerce Cloud, Magento, SAP Hybris connector depth

Whether customer acquisition requires a bespoke integration project — and whether the connector library covers the target customer's stack

Data pipeline quality

Event tracking accuracy, data freshness, missing data handling

Whether the AI layer is running on good data — or whether tracking gaps are silently degrading model performance

Data pipeline quality

Event tracking accuracy, data freshness, missing data handling

Whether the AI layer is running on good data — or whether tracking gaps are silently degrading model performance

GDPR & consumer data compliance

Consent management, data retention, consumer rights handling, cross-border transfers

Whether the data processing practices are defensible under current EU enforcement — and whether a customer's DPA audit would create problems

GDPR & consumer data compliance

Consent management, data retention, consumer rights handling, cross-border transfers

Whether the data processing practices are defensible under current EU enforcement — and whether a customer's DPA audit would create problems

Multi-tenancy & data isolation

Per-merchant data separation, cross-merchant learning controls

Whether a merchant's customer behavioural data can be used to benefit a competing merchant on the same platform

Multi-tenancy & data isolation

Per-merchant data separation, cross-merchant learning controls

Whether a merchant's customer behavioural data can be used to benefit a competing merchant on the same platform

AI in RetailTech & eCommerce

AI is the conversion story. It is also where the moat either exists or doesn't.

Every retailtech pitch in 2025 has an AI personalisation story. The differentiation between companies is increasingly about data depth, model quality, and the ability to demonstrate performance across genuinely different customer contexts. Here is what actually creates a durable AI advantage in retail — and where the risk is.

Opportunities we verify

Behavioural data accumulated over years and across diverse merchant types. A personalisation platform running on significant transaction volume for multiple years, across multiple retail categories, has a training dataset that newer entrants cannot immediately replicate. The key questions are whether the data is structured and usable for training, whether it is legally clean for model training under GDPR, and whether model accuracy actually improves with data volume or has plateaued.

Search and discovery AI on large catalogues. Semantic search, visual similarity, and intent-based product discovery AI create measurable conversion uplift for large-catalogue retailers where traditional keyword search fails. This is a high-value, high-switching-cost application where AI quality creates direct commercial differentiation. We assess the search quality on genuinely novel queries.

AI-native inventory and pricing optimisation. Dynamic pricing and inventory optimisation tools that use AI to balance sell-through rates, margin, and stock availability create quantifiable financial impact for merchants. Companies with strong inventory AI tied to their commerce data have a stickier value proposition than front-end personalisation alone.

Risks we surface

Conversion claims that don't survive scrutiny. Headline conversion uplift numbers in retailtech pitches are almost never comparable across companies. '35% conversion improvement' might be measured against a no-personalisation baseline, might include assisted conversion that doesn't add up to incrementality, or might reflect a single successful pilot. We evaluate the statistical methodology behind the commercial claims — because these numbers are the entire investment thesis.

Platform dependency on a small number of eCommerce hosts. Many retailtech companies are built almost exclusively for Shopify or one other major platform. A pricing change, API restriction, or competitive product launch by that platform can materially affect the addressable market overnight. We assess platform dependency concentration and the cost to expand to additional platforms.

AI models degrading silently as consumer behaviour shifts. Retail consumer behaviour shifts with trends, seasonality, macroeconomic conditions, and channel mix changes. AI personalisation models trained on historical behaviour can drift without surfacing any obvious error. A model well-calibrated in 2023 may be underperforming in 2025 in ways that only show up as gradual NRR erosion. We assess whether the company has an active model monitoring and retraining pipeline.

Know what you’re backing before you commit.

X-Ray delivers a full product and tech verdict on any retailtech or eCommerce platform target in one business day — validating the AI claims, testing the scalability architecture, and reviewing the data compliance posture.

Request an X-Ray See example report

250+ European engagements · 100% partner repeat rate