Nothing to paste? — a fictional Series B startup’s bands and a 13-person roster with real problems planted in it. Or , rendered from a saved result with no run and no charge.
Paste the roster, then the bands
CSV, TSV or markdown tables, with whatever column names your sheet uses. The band-math scan runs in your browser as you type — free, no sign-in, nothing sent anywhere — and flags the pure arithmetic: outside-band placements, compressed midpoints, inversions, missing cells.
Reviewed from your numbers — and only yours
The review weighs what the arithmetic cannot: which placements matter, what the band geometry does to your next cycle, who the data says is a retention risk, and what each fix costs. It never invents a market benchmark — if you did not supply one, “is this competitive” lands under Data gaps.
Check the verdict, then plan the cycle
A verdict binds it together: any outlier or retention risk rules out “Healthy”, and the recommendations arrive sequenced and costed — under-min fixes first, then structure. Copy the review or download the whole brief as markdown.
| Flag | The arithmetic | Why it matters |
|---|---|---|
| below min | base < band minimum for the person’s role and level | The classic stale-comp retention risk — usually long tenure that raises never caught up. |
| above max | base > band maximum | Either the hire was priced above the band or the band is stale — both need a decision. |
| compression | adjacent level midpoints less than 10% apart | Promotions stop paying — the ladder loses its meaning at exactly the level you promote into. |
| inversion | a lower level paid above a higher level in the same role | Usually a hot-market hire leapfrogging tenured staff — corrosive once it is discovered internally. |
| no band / no base | a roster row no band covers, or an empty pay cell | The review cannot place what it cannot match — these land under Data gaps. |
Derived from the comp-analysis skill’s framework — band placement, outliers and retention risks from the data you already have. Market percentile benchmarks need market data; paste yours into Context if you have bought some.
Does Comp Check compare our pay to the market?
No. It has no market data and never invents any. It reviews your roster against the bands you paste, and against itself — placement, compa-ratios, compressed levels, pay inversions. If you ask whether pay is competitive and supplied no benchmark, the answer lands under Data gaps rather than as a guessed percentile. If you have bought survey data, paste those figures into Context and the review may use them, saying they are yours.
What do I have to paste in?
An employee roster — CSV, TSV or a markdown table, with whatever column names your spreadsheet already uses (name, role, level, base, bonus_target, equity_annual, start_date and so on are matched loosely). The band table (role, level, min, mid, max) is optional, but per-person band placement needs it; without bands the review covers internal structure only.
What runs for free, without signing in?
The band-math scan and the placement table. Both run entirely in your browser as you type: below-min and above-max placements with the dollar gap, compa-ratios, adjacent midpoints closer than 10%, pay inversions inside a role, missing cells, per-level medians, and the total cost of lifting everyone below minimum up to it. Nothing is sent anywhere and nothing is charged. Only the written review is a metered run.
Can I test a raise before committing to it?
Yes, and it costs nothing. Type a proposed base into any row of the placement table and the entire scan re-derives in the browser — that person's placement and compa-ratio, the inversion and compression checks, the level medians and the running cost of your changes. "Lift everyone to band minimum" fills the whole column in one click.
Where does the result go afterwards?
You arrived from a spreadsheet, so you can leave for one: the placement table downloads as CSV or copies as tab-separated text you can paste straight into Sheets or Excel. The written review copies as markdown or downloads as a .md file.
Does it remember previous reviews?
Yes. Every review is saved to your SkillSafe account, not just to one browser, so a review you ran on your laptop is still there on your desktop. You can reopen an earlier one, or compare it with the review on screen to see exactly which people moved in or out of band, what it cost, and which risks cleared.
Is my compensation data used to train a model?
The free band-math scan never leaves your browser. When you run the written review, the text you pasted is sent to the model to produce that review and nothing else — there is no training on it and no sharing with anyone. Reviews saved to your history are stored against your own account.