Too small to test.
Too important to guess.

Too small for research. Too small for A/B testing. Too important to guess. That’s what Prior.Run is for.

Every consumer team has two ways to learn what customers think.

• —A study — eight people, three weeks, real budget.
• —Or a test — real traffic, six weeks, a statistically detectable lift.

Both have a floor. Below it, the evidence costs more than the decision is worth, so the decision gets made in a meeting by whoever is most senior.

This third category is actually the largest. Don’t believe us? Do this: sort the decisions you made last month into three buckets — Tested, Researched, Neither. This third pile of “Neither” is always bigger than people expect. And it currently runs on gut.

Pick an audience, ask one question, and a simulated panel of buyers reacts in their own words — including the parts you didn’t want to hear. Each panelist has a persona, a purchase history, and a set of priors about how they shop.

A/B tests tell you which. Never why. B won by 3%. Nobody knows the mechanism, so nobody can apply it to the next forty decisions. The finding dies with the test.

Studies answer why beautifully — and cost three weeks. Worth it for a new category launch, absurd for an out-of-stock message.

Our cost is about ninety seconds. That’s the entire difference: it clears the bar for decisions neither method can profitably touch.

Calibrate before you believe anything.

1. 01Pull three A/B tests you already ran and already know the winner of. Pick ones where the result surprised you.
2. 02Hand both variants to the panel without telling it which won.
3. 03Ask which it prefers, and why.

A home goods brand ran this on three Q1 tests:

Two out of three — and the miss draws the line for you. The first two were questions about what a customer thinks: does this claim land, does this subject line create urgency. It got both. The third was a question about where a thumb travels on a 6-inch screen. Nobody in the panel felt differently about the sticky bar; they just used it more. That’s an interface question, and interface questions belong to eye-tracking, session replay, and a real test. Ask this panel what someone believes, fears, wants, and won’t admit. Don’t ask it where to put the button.

For product

• —The default choice: the iPhone page pre-selects 256GB before anyone touches it. Ask a panel: “You land on this page and the storage is already chosen for you. What do you assume about why that one?” You’re checking whether it feels like a recommendation or a nudge toward the more expensive one.
• —Reseller portal: Apple Authorized Reseller has maybe a few thousand accounts. Statistical testing is off the table permanently — but these are your highest-value relationships and their onboarding has never been researched.
• —Empty promo field: the empty promo code field sits above the total at checkout. Design wants it gone, growth wants it visible. Ask: “There’s a box for a discount code and you don’t have one. What goes through your head?” Nobody has evidence today, so it resolves by seniority.
• —Tiers people can’t reach: you’re launching a bundle with a top tier most customers will never need. Ask if that’s motivating or just annoying.

• —Customization preview step: engraving an AirPods case is an extra step on a low-traffic flow. Does seeing your monogram render feel premium, or does it feel like a mall kiosk? Never enough volume to test, and it’s pure reaction.
• —Capacity comprehension: “256GB” vs. “about 50,000 photos.” A first-time buyer has no intuition for either. This is a comprehension question, not a preference one.
• —Storage chooser: 128 vs. 256 vs. 512GB. Gigabytes mean nothing; “four years of photos without deleting anything” means everything. Small lift, real confusion, nobody’s giving it a test slot.
• —Out-of-stock language: a $1,199 phone is unavailable. “Back in stock soon” vs. “Notify me” vs. a date. How does a customer feel about the brand at that moment? Low traffic, high emotional stakes.

• —The $1,199 question: what makes a price gap over a $400 competitor feel earned versus exploitative. You can test price. You cannot test why.
• —Discount tone: Apple rarely discounts. When it does, does a sale email make a customer feel lucky, or make them feel like the product was overpriced?
• —Trade-in framing: the trade-in estimate drops the sticker price by half before checkout. Does that read as generous, or does it make the original number feel invented?
• —Upgrade cadence: you’re asking a customer to replace a device that still works fine. Ask what makes that feel like an upgrade rather than a tax.

The free panels are generic. They’re a decent stand-in for how a shopper thinks, but they don’t know anything about your buyers specifically — your reviews, who comes back and who doesn’t, what people complained about last quarter.

The paid version is a panel built on your CRM data. “E-commerce shopper” becomes Sally, who actually bought your $1,199 phone in October and came back in March.

So start with the free panels. Run tests you already know the answer to. If it gets them right, you have got your case for the calibrated version of the model.