Skip to content
ChatCommerce
ENFR

Method

A measurement by draws, with figures and sources.

An AI assistant never gives the same answer twice. Here is how we get a figure that holds up from it: every share is published with its sample and its range.

The path

Three stages, from the first audit to tracking.

01

Free audit

You give us your store's address. You receive a grade from A to E, three priorities and a report to share.

02

Guided setup

A prioritized fix plan and a guided installation, by video call.

03

Tracking

A subscription to measure over time what changes and which fix to handle next.

Definition

What a draw is

A draw is a buying question put once to one engine, on a given date. We keep the engine, the question, the date, the text of the answer, the sources cited and the competitors named.

Your store counts as cited if its address appears in the answer's sources, or if its name appears in the text. The two cases stay distinct in the interface.

From the answer to the figure

A share, its sample, its range.

The share of presence is the number of answers that cite you, divided by the number of usable draws. It is published with that number and with a 95% Wilson interval.

From draws to the share of presenceFictional example: sixteen draws, seven of which cite the store, give a share of 44%, published with its sixteen draws and a range of 23% to 67%.01 DRAWSOne square per answer.Filled: the store is cited.02 SHARE7 out of 16Seven answers cite the store,out of sixteen usable draws.03 WHAT IS PUBLISHED44%16 draws. Range: 23% to 67%.Fictional example.

The five-draw rule. Below five draws, we do not show a share: the interface reads "Too few draws". A period with no draw is absent, never set to zero. An unknown position stays empty.

The questions

Where the questions come from

The questions are derived from your catalog and your categories, then you validate them. For a tracked product, the starting question uses its category and its brand.

You always compare comparable periods: same questions, same number of draws.

Reading

How to read a result

  1. Every share is published with its sample: from one week to the next, the same question can give different shares, and the range shows the margin of each figure.
  2. The engines are queried through a provider, DataForSEO. Behind every figure you can open the answer obtained, its sources and the competitors named.
  3. A product counts as cited if its address is in the sources, or if the beginning of its name is in the answer.
  4. The measurement is about citation: it tells you whether you are cited, engine by engine.
  5. Every result is dated: you follow what changes from one month to the next.

Collection

A controlled collection

A missing answer is left out of the calculation and flagged. The denominator stays visible, and the last known data stay displayed with their date.

See the sources and definitions

Start with the free audit.

One address, a minute and a half, a report to keep.

Run the free auditSee the prices