This tells us what to measure and, just as importantly, what you are actually free to do about it. Every answer feeds a named part of the report — the grey label on the right of each question is the field it becomes, so nothing you write here goes nowhere.
Skip anything you genuinely do not know. A blank is a finding; a guess is contamination. The required fields are the ones without which the report would test nothing.
So we measure the right entity. Large groups often have several — we need the one that hires.
companyNameNot the legal entity. The name someone searching for jobs would use.
Example Composio
legalEntityOnly useful as a last-resort search string — some job boards and AmbitionBox file postings under the registered name. Measured honestly: across our first three reports a legal name appeared in none of the 21 queries and none of the 11 review profiles, so leave it blank without worrying. The question below is the one that matters.
Example ROO India Pvt Ltd
reviewPopulationThis is the most consequential question on this page and it takes five seconds. If your Glassdoor or AmbitionBox rating comes from a different workforce — a global parent, a factory, a services arm — then reading it as evidence about the team you are hiring into is simply wrong. It happened in one of our own reports: a new engineering centre was scored on its parent group’s worldwide rating, and that was a quarter of its headline number.
Example different_workforce
aliasesComma-separated. Old names, local spellings, the parent brand. We search all of them; a company missed because we searched the wrong string reads as invisible when it is not.
Example Zeiss India, ZEISS Group India, Carl Zeiss
targetLocationOne city. Visibility is measured on that city's discovery surfaces, and a report spanning two cities measures neither.
Example Bengaluru
targetLocationOtherWe will tell you honestly whether we have a surface frame for it yet. If we do not, your report says so.
Example Coimbatore
companyTypeThis sets which surfaces a candidate would actually check. A GCC and a seed-stage startup are found in completely different places.
Example ai_startup_seed_to_a
localHeadcountSets what a realistic 90-day plan looks like. Advice for 25 people and advice for 2,500 are not the same advice.
Example u100
The single most useful part of this form. A report that answers a question you did not ask is a waste of your money.
reportTypeThese are genuinely different documents, not labels on the same one.
Example employer_brand
focusProblemPlain language. Not a brief. The thing you would say to a colleague over coffee — that sentence shapes the whole report.
Example We get plenty of applications from service companies but almost nothing from the product-AI engineers we actually want, and the ones we do reach seem never to have heard of us.
topQuestionsOne per line, up to five. We check every one of these is either answered or explicitly recorded as unanswerable — that check is a gate, not a promise.
Example Why do senior AI engineers never apply to us? Are we paying below the market we think we are in? What do candidates find when they search our name?
lossStageDifferent stages point at completely different causes. Never applying is a visibility problem; declining an offer is not.
Example never_apply
We only ever name companies you name here. We never invent a competitor, and we never add one we think fits.
competitorsTaggedAt least two, ideally three or four. This is the set every competitive number is computed against. Pinning fewer, or softer, rivals used to flatter the score — it no longer can, because the share is floored at eight naming events, but a thin set still makes the reading less useful to you.
Example Sarvam AI · rival
| Company | Relationship | |
|---|---|---|
| × | ||
| × | ||
| × |
hiringRolesDetailedStart with the hardest one to fill. Comp benchmarking, salary evidence and the candidate-questions section are all scoped to these.
Example Senior platform engineer · senior · 12
| Role title | Level | How many in 12 months | Key skills | |
|---|---|---|---|---|
| × | ||||
| × | ||||
| × |
We test each of these against what is publicly discoverable. Where the record backs you up, that is your strongest asset. Where it does not, that is the report.
evpClaimsOne line each. If you have never written these down, say what you would SAY in a final-round conversation — that is the claim, whether or not it is on a careers page. Leave one blank if you genuinely make no claim there; a blank is a finding too.
publicSurfacesOne URL per line — careers page, engineering blog, GitHub org, X, LinkedIn, a founder's post, anywhere you make a claim about working there. Four separate boxes used to ask for these and clients pasted the same link twice. A claim with a citable artifact scores far higher than the same claim made only in interviews, because a candidate can actually find it.
Example https://composio.dev/careers https://blog.composio.dev/engineering-culture https://github.com/ComposioHQ
Optional, and the most valuable thing you can give us. It is the only way to measure the gap between what you say and what people experience. Aggregates only — never individual responses, never below five people, and it never leaves your own report.
internalEngagementA number and its date. We report the say–do gap, never the raw score.
Example eNPS +18, March 2026, 78% response rate
internalStrengthsThemes, not quotes. Never paste free-text comments — we would only have to strip them, and they should not travel.
Example Autonomy; the quality of the engineering bar; direct founder access
internalPainsThemes, not quotes. This is the half that makes the report worth reading, and the half people are tempted to leave out.
Example Unclear promotion path; on-call load; compensation opacity
exitReasonsThemes, and roughly how many. Aggregate only.
Example Comp (4), no growth path (3), manager (2)
referralRateA revealed preference: people who like working somewhere tend to bring friends. We also flag the other side of it — referral-heavy pipelines concentrate on your existing employees' networks, so we tell you to watch representation before optimising for this.
Example About 16% last year
This decides which recommendations you actually get. A plan you are not allowed to execute is not advice, and we would rather know now.
engineerPostingPolicyMany regulated employers and GCCs prohibit it. If yours does, we will not recommend engineer advocacy — we will find another route.
Example encouraged
leaderVisibilityA named leader writing once is one of the highest-leverage actions available. If that is not going to happen, we should not build a plan around it.
Example willing
budgetBandHonest answers get better plans. Several of the highest-impact actions cost nothing but attention.
Example small
keyStakeholderA role, not a name. It changes what we lead with — a CFO and a Head of Engineering need different first pages.
Example CFO, then the India site leader
Optional, and only two questions now. These two are separate from the surfaces above because they answer a different question: WHICH review profile is yours. Attaching a parent's rating to a subsidiary is the most consequential measurement error this method can make, and it has happened.
glassdoorUrlThe one for THIS entity if you have several. Attaching a parent's global rating to a local site is a real measurement error and we would rather avoid it.
Example https://glassdoor.co.in/Overview/…
ambitionboxUrlExample https://ambitionbox.com/overview/…
contactEmailWhere the report goes, and how we resolve your account. This is the only personal detail we ask for.
Example you@company.com
consentUnderstoodScores are modelled estimates from counted public evidence, shown with sources and dates. They are not a transcript of any AI assistant's output, and no link to a hiring outcome has been established.
Example yes
consentInternalDataNo individual responses, no free-text comments, nothing below five people. If you supplied none, choose the second option.
Example confirmed