Blog · November 3, 2026 · 5 min read

Your average score hides your best fix: design for your weakest buyer

By Pavan, Founder, FirstReaction · Updated

In short

An ad's average score hides a 20 to 40 point gap between its most and least convinced buyers. In five live ads, the least convinced buyer type gained at least 10 points from a fix, and the most of any type in three. The catch: fixes aimed at the weakest buyer sometimes left the strongest one flat.

When you test an ad, the number everyone looks at is the average: the CTR, the score, the cost per result. In the five live ads we simulated for What's Wrong With This Ad?, the average hid the most useful thing in the result: how far apart the different kinds of buyer were, and where the easy gains sat.

The gap inside every audience

Each simulation splits the audience into up to 5 buyer types (archetypes) and scores each one. In every original ad, the most and least convinced buyer types were 20 to 40 points apart on likelihood to act:

AdLeast convinced (before → after)Most convinced (before → after)Gap before
Snitch plus-size jeansSavvy Deal Seeker, 40 → 82Loyal Brand Advocate, 80 → 7940
Lola Blankets, $325 blanketValue-Driven Splurger, 20 → 40Cozy Comfort Seeker, 53 → 7333
Sahajan cleansing oilPractical Purist, 45 → 62Socially Influenced Explorer, 72 → 7627
Lola Blankets, holiday offerDiscerning Decorator, 33 → 63Value-Driven Splurger, 58 → 5725
Bella Vita perfume bundleSpecial Occasion Impressor, 43 → 53Savvy Gifter, 63 → 7220

An ad "scoring 61" might really be one that works brilliantly for loyal fans and barely at all for people hunting a deal. Those need different fixes.

The weakest buyer is where the upside is

The least convinced buyer type gained at least 10 points in every rebuilt ad, and the most of any type in three of the five (Snitch, Sahajan, and Bella Vita, where the two weakest types gained the most). In the two Lola ads, the weakest type doubled or nearly doubled.

That's not a coincidence. The weakest type usually has one specific, answerable objection, and the simulation names it:

  • Snitch's deal seekers (40) were looking at a hot-audience carousel with prices but no offer. Adding Snitch's own app offer took them to 82.
  • Lola's value-driven shoppers (20) saw a $325 blanket with no reason it's worth $325. Adding the rating, the review count and practical features doubled them to 40.
  • Sahajan's practical purists (45) needed a reason the $41 price was worth it, from a brand they didn't know yet.
  • Bella Vita's special-occasion buyers (43) wanted to know what the scents smell like and that others rated them.

Meanwhile the most convinced type often had nothing left to gain.

Watch the buyer you already had

The flip side: a fix aimed at the weakest buyer can cost you the strongest one.

  • Snitch's loyal brand advocates went from 80 to 79. They already loved the brand; an offer didn't add anything for them.
  • Lola's value-driven shoppers in the holiday ad went from 58 to 57. The discount code was already working for them; telling the product story added little.
  • Sahajan's first re-cut dropped two buyer types by 8 to 10 points before a second round won them back (more in fix, then test the fix).

None of those are disasters, but they're decisions. If the buyer you'd lose is your core customer, the "better" version might not be better for you.

How to use this on your own ads

  1. Look past the average. Find the buyer type with the lowest score and read its reason.
  2. Fix that one objection, using only claims you can back: a rating, a detail, an offer, a demonstration.
  3. Re-test, and check the top of the table too. Make sure the buyer you already had didn't drop.
  4. Decide who matters most. If one buyer type is most of your revenue, weight its score above the average.
  5. Then confirm live with the version that wins for the buyers you care about.

For how buyer types are built and what the research says about simulated audiences, read what is AI audience simulation in advertising?

Sources

  1. Meta Ad Library