Ulta Beauty: 'too expensive' is not an invitation to argue

A customer says, 'Too expensive.' Somewhere inside the associate's head, a forty-slide presentation starts loading. Nobody requested it. We hear about innovation, uniqueness and remarkable results, while the customer may simply be wondering how this product differs from the one already sitting in their bathroom.

Ulta Beauty describes a range of in-store services, with availability varying by location. That provides useful context for consultation-led selling. The training design below is my independent proposal, not a report on an existing Ulta AI programme or its commercial performance.

Find out what expensive means

A useful starting point is three virtual shoppers saying exactly the same words. One has a firm budget, another is uncertain about the shade, and the third doubts they will use the product often enough. Each needs a different conversation. A universal objection script mainly benefits whoever sells the script.

The first mechanism is branching role-play. The AI customer reveals the real concern only when an appropriate question is asked. Defending the price immediately does not make them more agreeable. Understanding the task creates room for a smaller selection, another product or a demonstration available in that location.

Next comes explaining value without miracles. The employee uses verified product features and application information rather than promising a different life after two uses. The trainer checks claims against approved material. Skin reactions and medical questions must trigger appropriate boundaries, not a competition in confident improvisation.

A third mechanism lets the employee replay only the difficult moment: moving from an objection to a useful question. A manager then reviews examples to see whether the person understands the reasoning or has merely memorised an elegant sentence. A patient virtual partner is helpful; an unaccountable examiner is not.

Include a shopper who still declines after receiving a clear explanation. The exercise ends with a respectful close, not a fourth attempt to force the decision. Otherwise we teach employees to recognise concerns only so they can ignore them more efficiently.

Nobody needs to defeat the customer

A pilot could cover a single category using a few common scenarios. Compare groups with similar experience and access to the same products and promotions. Otherwise a successful sale period will mysteriously become an AI achievement, when the software was simply standing next to a compelling discount.

My three primary measures would be correctly identified concerns in an independent assessment, returns linked to mismatched expectations and customer ratings of freedom from pressure. Together they say something about consultation quality. Conversion is a useful additional business signal, but cannot tell us what the employee actually learned.

Recorded real conversations need a separately agreed consent and data-handling process. They are unnecessary for an initial pilot: synthetic dialogues, verified product information and manager observation are enough. useful AI in sales does not make somebody persuasive at any price. It helps them judge whether persuasion is appropriate.