See what you would get back.

A short report should explain what we did, what the evidence shows and what you can decide next.

Illustrative report — entirely fictional data

This is a teaching example, not a customer case study, a live engine run or evidence of improved performance. Engine labels are generic. No real customer or provider results are represented.

1. The agreed scope

Example business: a fictional independent HVAC company. Service: furnace repair. Area: one agreed service area.

Protocol example-v1: ten discovery questions, three repetitions per engine, four selected engines. That means 30 planned answers per engine and 120 per phase. A real report includes exact business identity, question text, run dates, engine/model settings, location settings and analysis version.

2. The approved change

The fictional business’s existing service page did not explain whether it repaired furnaces. The approved edit clarifies the service, supported equipment, area and request process. A real report links to the published page and stores approval, before-and-after content and publication time.

3. Results by engine

Fictional counts below use complete batches of 30 answers. “Website cited” means an explicit citation to the verified client website. A retrieved background URL does not count.

Fictional data · Report preview

What changed?
A mixed picture.

One engine stayed the same. One rose. One fell.
The fourth needs a complete follow-up.

BaselineFollow-upEvery chart: 0–30 answers

Example A

No observed change

Business mentioned

0102030Baseline6/30Follow-up6/30

Website cited

0102030Baseline2/30Follow-up2/30

Example B

Higher observed counts

Business mentioned

0102030Baseline3/30Follow-up5/30

Website cited

0102030Baseline1/30Follow-up2/30

Example C

Lower observed counts

Business mentioned

0102030Baseline8/30Follow-up7/30

Website cited

0102030Baseline3/30Follow-up2/30

Example D

Comparison withheld

28 of 30 follow-up answers usable. Two requests failed. The baseline is complete, but there is no valid comparison yet.

This strip shows data completeness—not visibility. Missing answers are never treated as zero citations.

Illustration only. These are not measured results, customer leads, or evidence that an edit caused a change.

View the exact example counts
Fictional data — counts out of 30 answers per phase
EngineMentions: baseline → follow-upCitations: baseline → follow-up
Example A6/30 → 6/302/30 → 2/30
Example B3/30 → 5/301/30 → 2/30
Example C8/30 → 7/303/30 → 2/30
Example DWithheldWithheld

There is no combined visibility score. Example D remains incomplete; failures are not counted as zero visibility. In this illustration, the protocol and analysis version match for A–C. A mismatch would prevent a valid comparison.

4. Evidence you can inspect

No raw answer download is provided here because these measurements never happened. A real pilot includes its supporting evidence.

5. What this would mean

The illustrative results are mixed. They do not establish a reliable improvement or show that the edit caused the differences. No automated statistical significance test or causal attribution is included. API measurements also differ from personalized consumer chatbot experiences.

6. Business outcomes and next decision

Calls and booked jobs: not measured in this example. A real customer keeps a separate inquiry log; an increase in citations is not a lead count.

Example decision: keep the factual page correction because it helps customers understand the service. Investigate the incomplete batch, preserve failed attempts and agree the next checkpoint before making another change. Do not claim a success story from these counts.

Review the proposed $500 pilot · Read the full measurement method

Start with a useful question.

Which service do you want more customers for—and how do they find you today?

Discuss a pilot