APS-1.0 methodology
This page sets out every step from the question a buyer asks to the number on the report. It is published in full, so anyone with the same answers can reach the same score.
| Pillars scored | 3 of 4 |
| Scale | 0 to 100, for each pillar and for the composite |
| Cadence | Weekly, reported as a 12-week trend |
| Minimum evidence | 12 data points for each pillar |
The pipeline
Five steps, run every week. Each step is stored, so any score can be traced back to the answers behind it.
- Questions
A fixed set of buyer questions, branded and unbranded.
- Engines
Each question is asked in every measured AI engine.
- Answers
Every answer is stored with its engine and date.
- Pillar scores
Knowledge accuracy, perceived value and competitive loyalty, each from 0 to 100.
- APS
The weighted composite, once every pillar has enough evidence.
Question sets
Some questions name the company and some don't. They measure different things, so each pillar is scored on the set that can answer it.
| Branded questions | Unbranded questions | |
|---|---|---|
| Names the company | Yes | No |
| What it tells you | How the company is described when the buyer already knows its name. | Whether the company is the one recommended. Any company named was chosen by the engine. |
| Examples | What does Acme cost? Is Acme a good choice for me? | Best CRM for a 10-person team? Which airline for a group booking? |
| Feeds | Knowledge accuracy, perceived value | Competitive loyalty |
Pillar I: Knowledge accuracy
Every factual claim an engine makes about the company is extracted from its answers and checked against a fact base the company has verified.
- A claim is marked correct or wrong only when the fact base holds the fact.
- Claims the fact base can't check are reported, not scored.
- At least 12 checked claims are needed before the pillar counts.
// share of checked claims that are right knowledge accuracy = correct ÷ checked × 100
Pillar II: Perceived value
Each mention of the company is classed as positive, neutral or negative. The ratings are combined into a perceived value score from 0 to 100, overall and for each engine.
- At least 12 mentions are needed before the pillar counts.
// how positively engines speak about you perceived value = rating(mentions) → 0–100
Pillar III: Competitive loyalty
On unbranded questions, loyalty combines how often the company is named with where it lands in the list.
// one answer: n brands named, you at rank r position = (n − r + 1) ÷ n // one question loyalty = visibility × average position // the pillar competitive loyalty = average over unbranded questions
At least 12 unbranded questions are needed before the pillar counts. Being named every time at the bottom of five is worth about the same as being named a fifth of the time at the top.
The composite score
The three scored pillars are combined into one number, but only when each has earned its place.
APS1.0 = ⅓·I + ⅓·II + ⅓·IIIThe final split is published with APS-1.0 and applies to every company, because scores only compare if the method is the same. Pillar IV, strategic outcome effectiveness, is in development and not part of the score.
Minimum evidence
| Pillar | Minimum | Counted as |
|---|---|---|
| Knowledge accuracy | 12 | Checked claims |
| Perceived value | 12 | Mentions |
| Competitive loyalty | 12 | Unbranded questions |
Reporting
Every number can be cut and traced back to the answers behind it.
| Cut | What it shows |
|---|---|
| By engine | Every pillar and the composite for a single AI engine, such as ChatGPT or Gemini. |
| Over time | Scores pulled weekly and shown as a 12-week trend. |
| By question | Each question with its own scores. Each score opens the answers behind it. |
What APS is not
- Not a visibility score. Being named is one input to loyalty, never the whole score.
- Not a single screenshot. One answer from one engine is never reported as an APS.
- Not adjustable. The method is the same for every company, and scores are never changed by hand.