Adidas: the size in the app still has to meet the foot

There is a particular confidence in saying, 'I always buy this size.' Then a different model arrives. The confidence remains; the toes have considerably less room. Size charts are useful, but nobody buys a chart. They buy something for walking, training and occasionally standing in a coffee queue for longer than planned.

Continue the conversation

An online assistant asks for foot length using clear measurement instructions, a familiar model and the customer's preferred fit. It explains uncertainty and suggests trying alternatives. Where model-specific information is missing, it says so. Personalisation should mean more than guessing while using somebody's first name.

The customer can save a short fitting card with two sizes to test. A store associate sees it only when the customer voluntarily presents its code. Instead of restarting the interview, they confirm the intended use, bring available options and ask what actually feels right during the fitting.

The customer must be able to correct the recommendation. The system should learn from an explicitly confirmed choice, not argue using its own probability score. Nor should that size become universal across the range. Carrying it unquestioningly into another model would simply automate the original mistake at greater speed.

Avoid building a biometric archive for a pair of shoes

Voluntarily supplied measurements, the model identifier and fitting outcome are sufficient. Foot photographs and permanent measurement storage are unnecessary defaults. Customers should be able to delete their fitting card. Each recommendation needs the relevant guide version and date, with an associate checking any conflict.

The first trial could cover several models in stores carrying adjacent sizes. Without those alternatives, the test is misleading: someone buys the only available option and appears satisfied in the dataset. Record separately when a person declines because the appropriate size cannot be supplied, rather than blaming the recommendation.

A seamless journey needs evidence: examine size-related returns, repeated clarification across channels and fittings completed with a suitable option. Examine post-purchase contacts as well as the initial transaction. Fewer questions are only positive if they do not mean customers have stopped expecting useful assistance from the retailer.

Adidas publishes a footwear size guide. That is a reference to check, not a complete solution to fit. My independent proposal concerns AI in omnichannel sales: preserve the reasoning between online selection and a store visit without presenting a probable size as a guaranteed result.

This AI does not replace trying shoes on or make human feet standardised. It removes the need to tell the same story three times. The retail opportunity is less friction and fewer avoidable sizing mistakes. The human benefit is simpler: the store remembers the conversation without deciding it knows the customer better than they know themselves.