From 0% to 100% of agent choices.

The starting point: agents chose other options

In an early research round, agents compared the original offer with alternatives. The original offer received no selections across 108 recorded choices: a 0% selection rate. The alternatives received all of the choices.

The practical question was how to give agents a stronger reason to choose our offer without simply lowering the price. The company, industry, tasks, prompts and actual offer construction remain private.

What we changed

We connected the buyer’s desired outcome with what the product could deliver, then reconsidered the promise, pricing and packaging together. We tested the resulting offers against competing options as the research progressed.

The later comparison tested a redesigned offer at three prices against a generic option and other alternatives. This account shares the results, not the confidential offer wording or implementation.

The later result: 100% of recorded choices

The later test produced 20 readable choices from 24 attempted runs. Every one of those 20 choices went to the redesigned offer: 100% selection. Four answers could not be reliably interpreted and are not counted as wins.

Compared with the 0% result for the original offer in the early research, the later result is 100 percentage points higher. This is the observed difference across research stages, not an isolated causal estimate from a matched before-and-after experiment.

The later test used two AI model configurations. Offer positions were rotated and identifiers remapped to reduce presentation and interpretation errors.

What “50% higher” means

The redesigned offer was compared with a generic option at the same listed price, at a 25% higher listed price and at a 50% higher listed price. At that highest price, the new offer’s listed rate was 1.5 times the generic option’s rate.

The redesigned offer won all six readable choices at the same price, all seven at the 25% higher price and all seven at the 50% higher price. The cheaper generic option did not receive a recorded choice.

The 50% refers to the listed rate in that comparison. It is not a 50% profit increase, a measured rise in customer spending or proof that an unchanged product can raise its price by 50%. The offers differed in what they promised and provided.

What the comparison does and does not show

The early and later rounds used different tasks, offers, competitors and model configurations. The 0% to 100% comparison describes the progress of the research programme; it is not a like-for-like A/B test or a universal expected lift.

The early round had 108 recorded choices. The later round had 24 attempted runs, 20 readable choices and four excluded answers. Each higher-price condition had only seven readable choices. These are controlled research results, not independent customers or live marketplace sales.

No realised revenue or margin improvement was measured. Delivery quality, costs, payments and repeat purchasing still need separate validation.

Why this matters for your product

A cheaper price is not the only reason an agent might choose you. This research identified an offer worth testing further because it earned the choice despite a higher listed rate.

The commercial next step is to check whether the product can deliver that promise profitably and whether live buyers make the same choice. That connects customer research and offer design to product requirements, pricing and a measurable growth opportunity.