Methodology

How Zyobit measures

Understand what goes into your scores and how to read them. Explore the signals behind each finding and use them to plan your next action.

Formula market-v1Updated

At a glance

Five dimensions
Discovery, understanding, trust, competition, and conversion signals.
Scores from 0–100
Read each score alongside its evidence and local market context.
Public signals
A view of what customers can find, rather than private sales or booking data.

Data sources and refreshes

Measurements combine public business listings, including business details, ratings and review counts; checks of publicly reachable business websites; and sampled AI answers with web search enabled. Comparisons use the businesses collected for the selected category and area. Provider coverage and collection limits mean this is not a complete census of local businesses.

The free report is a one-time snapshot. Paid workspaces with tracked markets are scheduled for a weekly market measurement. Each result includes its measurement date, and individual sources may use cached data. Read findings alongside their measurement dates.

What gets measured

We follow the whole path a customer takes — searching, asking an assistant, reading reviews, comparing a couple of websites, deciding — and score five points along it.

FindabilityDo you appear in sampled AI answers?

AI mentions relative to the market leader, coverage across the sampled questions, and share of sampled responses. These contribute 45%, 35%, and 20% respectively.

The response-share contribution reaches its maximum at one third of sampled responses. The combined score is rounded and bounded between 0 and 100.

Business clarityAre your business details clear?

Eight equally weighted presence checks: website, phone, opening hours, address, business category, reachable website, hours on the website, and at least one service page.

Trust & reviewsWhat do your reviews show?

Review volume relative to the market median (45%), rating quality (35%), and rating relative to the market median (20%).

Review recency and review topics are not inputs to this formula version. Comparisons depend on the businesses included in the measured market.

Local competitionHow do you compare in this sample?

Your position in the scored market, normalized from 0 to 100. First place scores 100 and last place scores 0; a market with one business scores 50.

Contact & bookingWhat contact and booking options were detected?

A presence checklist: booking link, click-to-call, contact form, mobile viewport, load time, service-page count, hours shown.

Each detected website signal contributes equally to the conversion score.

How to read your score

Each dimension scores 0–100 and falls into one of three bands. The bands are a judgement about what is worth telling you, not a measurement — they are listed here so you can disagree with them specifically.

  • Strong70–100/100

    Maintain these signals and watch for changes.

  • Fair50–69/100

    Review the evidence for opportunities to improve.

  • Weak0–49/100

    Investigate the gaps and prioritize your next action.

From signals to a score

For a reachable website, the conversion dimension counts seven equally weighted checks. Here is how one illustrative result is calculated.

Illustrative exampleConversion signals
  • Mobile viewportPresent
  • Click-to-call linkPresent
  • Booking linkPresent
  • Contact formNot met
  • Opening hoursPresent
  • At least three service pagesPresent
  • Loads in under three secondsNot met
5 of 7 checks passed5 ÷ 7 × 100 ≈ 71
71/100Strong

This measures detected website signals, not the percentage of visitors who book. An unreachable website scores 0 in this dimension.

How AI answers are measured

We generate the questions a real customer would ask in your area, put each one to an AI assistant with web search enabled, and record which businesses it names.

  • Every question runs more than once. We record the businesses named across repeated answers and report the average.
  • Web search is always on. That is what a real customer gets. A model answering from memory alone invents local businesses, which would make the numbers meaningless.
  • Names are resolved to real places. “Smile Dental” and “Smile Dental Clinic Pvt. Ltd.” are one business. We match extracted names against verified local listings above a similarity threshold of 0.60; anything below it is held for review rather than quietly discarded.
  • We compare a measured market. The comparison includes businesses collected for your category and area, rather than only the business requesting the report. It does not include every local business.

A version behind every score

Scores are stored with the version of the formula that produced them. When we change how something is scored, past results keep their original version rather than being silently rewritten. Compare formula versions alongside measurement dates: scores can change because public signals, competitors, sampled answers, or the formula have changed.

Raw source data is retained, which means a corrected formula can be applied to past measurements without re-measuring.