The First Job of a Shopping Agent Isn’t Just Discovery. It’s Preventing Regret

Retail has no shortage of “artificial intelligence shopping agent” demos that promise to get shoppers to the right product faster. And discovery is still hard. Shoppers still get lost in endless options, retailers still struggle with relevance, and the stakes are high when the product is physical. But in apparel and footwear, the failure that hurts the business most often happens at the moment of decision. The hard part is the last 30 seconds before checkout, when a shopper is thinking, "Will I regret this? Will this fit the way I want?"

In 2024, retailers projected $890 billion in returns, roughly 16.9 percent of annual sales, per the National Retail Federation (NRF) and Happy Returns (a UPS company). NRF also reported that 51 percent of Gen Z shoppers engage in “bracketing,” buying multiple items or sizes with the intent to return some. EMARKETER forecasts suggest online returns are on track to approach half of all retail returns by the end of the decade, even though e-commerce sales will represent only about 20 percent of total retail sales. That mismatch is the tell: what breaks online is confidence.

If we want shopping agents to be more than a novelty layer, we should judge them by a simple question tied to the shopper’s purchase: Did they keep it?

The KPI Evolution Steers From Intent to Validated Outcome. From Desire to No Regrets. From 'Add-to-Cart' to 'Keep Rate'

Add-to-cart (ATC) and conversion remain critical signals. They measure interest, intent, and demand. However, when shopper confidence is low, those signals can mask a deeper issue: shoppers buying multiple options “just in case,” expecting some to be returned.

That’s where keep rate becomes especially valuable. It measures validated outcomes, such as whether the shopper felt confident enough to buy one item in one size and keep it.

These metrics aren’t in conflict, they’re complementary: ATC tells you what shoppers want. Keep rate tells you what actually worked.

In high-stakes categories like apparel and footwear, the standard isn’t just how effectively you drive demand. It’s whether that demand converts into confident, successful purchases.

One way to make this actionable is to pair keep rate with the “why” behind it: return reasons at a decision-point level. Instead of “didn’t fit,” capture the failure mode the agent could have prevented: waist too tight, fabric too sheer, heel slipped, length too long, occasion mismatch.

That taxonomy becomes a road map for what the agent needs to answer, and keeps teams honest about whether “helpful” guidance actually reduces regret, not just accelerates checkout.

Map the 'Confidence Moments' That Shape Decisions

Teams often treat fit like a single sizing problem. In reality, hesitation occurs at specific moments when the shopper lacks decision-grade context. A practical way to design (and measure) an agent experience is to map these confidence moments by category: fit and proportion, fabric and feel, comfort and movement, occasion-fit, and edge cases.

These are the moments when shoppers hesitate and where discovery breaks, not because they cannot find options, but because they can’t trust the outcome.

That is the design principle retailers should hold the line on: an agent doesn't need better taste than the shopper. It needs to remove the barriers that cause avoidable mistakes. Fit is a barrier. Style is personal.

That gap shows up in very human behavior: impulse buys, second-guessing, and the quiet decision to order “just in case.”

An Operator Plan: Crawl, Walk, Run

You do not need a full re-platform to improve confidence and reduce regret. Follow these steps:

1. Crawl: Get the foundations clean enough to trust.

• Quantify bracketing and track both ATC and keep rate.
• Fix return reasons (get past “didn’t fit”).
• Clean fit-relevant attributes.

2. Walk: Close the loop between what shoppers ask and what they keep.

• Connect pre-purchase questions to outcomes.
• Define escalation rules. When uncertain, ask or escalate.

3. Run: Apply precision, not blanket generosity.

• Segment the return experience so you can use the right lever: education, exchange, or guardrails.

The Point

Retailers will not succeed in agentic commerce by optimizing discovery alone. They will win by making discovery trustworthy and reducing regret at the moment of decision. When you design around confidence moments and balance intent metrics like ATC with outcome metrics like keep rate, shopping agents stop being a shiny interface and become an operational lever.

Jessica Arredondo Murphy is co-founder and CEO of True Fit, the leading AI provider of size and fit technology for fashion retailers.