A Compensation Benchmark and Band Check
Answer 'what should we pay for this role?' with structure instead of a gut number — benchmark ranges, band placement, and the internal-equity check that prevents a quiet pay problem.
'What should we pay?' is three questions, not one. Answer all three before you make an offer.
The market range, where in the band this person lands, and whether that's fair against who you already employ — skip any one and you create a problem later. Store your bands and philosophy, and the check becomes structured instead of a gut number you'll defend awkwardly.
A number pulled from instinct feels fine until a current employee finds out a new hire is paid more for the same work. That's the quiet problem a band check prevents.
A pay number you can't explain is a number you'll have to walk back.
This structures the decision; it doesn't make it. Comp is judgment about budget, value, and fairness — the recipe makes sure the judgment has the full picture. Not financial or legal advice.
| Complexity | Easy |
| Tools needed | Claude Pro or above, Claude Desktop → Cowork mode, on macOS or Windows, Claude Projects |
| Time to build | ~25 min first time · ~10 min per role after that |
| Best for | Founder or HR lead who owns offer numbers — Small companies making competitive hires |
What this solves
Offer numbers pulled from instinct — never checked against the market, your bands, or what your current team already earns.
The problem
A great candidate is in final rounds and the hiring manager asks what to offer. You reach for a number that feels right — close to the last person you hired, adjusted for vibes. It might be fine. But you didn’t check it against the market, you didn’t place it deliberately in a band, and you definitely didn’t compare it to what your current team in that role earns. Six months later, someone discovers the new hire makes more for the same job, and now you have a retention problem and a fairness problem that a few minutes of structure would have caught. Comp decisions get compared whether or not you meant them to.
The fix is running the same structured check every time, so the number is defensible before it’s offered.
Ingredients
- Claude Subscriptions — Cowork and Projects aren't available on the free plan
- Platforms & Modes — Cowork runs in the desktop app only — not on web or mobile. This recipe structures comp data and checks it against your team — multi-step work that runs in Cowork
- Claude Projects — Your comp philosophy, your bands, and your team data persist. A Project stores them so every comp question is answered against your real structure, not a generic salary survey
How it works
You make comp decisions repeatedly, and your philosophy, bands, and team data are constant. A Project stores them so every decision is checked against your real structure — and your internal equity — instead of a one-off guess.
- Open Claude Desktop and click Cowork in the mode selector across the top (Chat · Cowork · Code).
- In the left panel, find Projects and click the + button.
- Choose Start from scratch. Name the project “Compensation” and let Claude set up its folder.
- You’ll know it worked when the project appears in the left panel with its own folder and an instructions field.
Your project has a folder on your computer — that’s where the three working files live.
- Ask Claude, right in the project: “Create three files in this project’s folder:
comp-philosophy.md,bands.md, androster.md. Leave them empty — I’ll fill them in.” (Or create them yourself in any text editor and save them into the project folder. Keeproster.mdprivate — it’s pay data.) - Fill in all three files using the descriptions below.
- Confirm Claude can see them: ask “List the files you can see in this project.” All three filenames should come back. If they don’t, see If It Doesn’t Work.
comp-philosophy.md
How you decide pay — where you aim against market (e.g. 60th percentile), how you weight experience, your stance on negotiation. This is what turns a range into a placement.
Example: “We target the 60th percentile of market for our region. We pay for demonstrated impact, not years. We don’t lowball and negotiate up — we open near our real number. Bands are public internally.”
bands.md and roster.md
Your salary bands by level, and a current roster with roles and pay (keep this file private and local). Together they let Claude place a number in a band and check it against who you already employ.
Example bands: “L3 IC: 90–115k. L4 IC: 110–140k.” Roster: “Two L4 ICs at 122k and 128k.”
Open your Project in Claude Cowork. Give Claude the specifics for this run, ask for the main output, then follow up for any additional pieces you want. The exact wording for each prompt — with what it’s asking for and why — is in What you actually type below.
Check three things before any number reaches a candidate:
- The market range is sourced or flagged as an estimate. Benchmark data drifts and varies by region. If Claude gives a confident range without saying where it comes from, treat it as an estimate to validate against a real survey — not a fact. Push for the caveat.
- The equity check actually compared the team. The most valuable output is “this would put the new hire above your two tenured L4s — here’s the gap.” Confirm Claude used your roster, not a generic statement. That’s the part that prevents the quiet problem.
- The reasoning is visible. A number with no shown logic can’t be defended in a comp committee or to the candidate. Make sure you can see how it was derived, so you can stand behind it or correct an assumption.
Before you offer: this is structure, not a decision. Comp involves budget, the candidate’s value, and fairness judgments that are yours to make — and the numbers here are not financial or legal advice. Validate market data against a real source, and run anything sensitive past whoever owns comp policy.
What you actually type
Name your files explicitly the first few runs, and ask Claude to show its work on anything that matters.
“We're hiring an L4 individual contributor for [role]. Using `comp-philosophy.md`, `bands.md`, and `roster.md`: give me a benchmark range for this role and region, where in our band a strong candidate should land per our philosophy, and an internal-equity check against our current L4s — flag if the proposed number would create a fairness issue. Show your reasoning.”
“If the candidate counters at 135k, where does that put them relative to the band ceiling and the current team?”
A three-part comp picture: a benchmark range for the role and region with its source named or flagged as an estimate, a recommended placement in your band per your philosophy, and an internal-equity check that compares the proposed number against your current people at that level. The reasoning behind each part is shown. If the range arrives with no source caveat, or the equity check is a generic statement instead of a comparison against your actual roster, it missed — push back.
- It doesn't supply current benchmark data. Market ranges drift and vary by region — validate any range against a real, current salary survey, and run sensitive calls past whoever owns comp policy. The recipe replaces neither the source nor HR/legal judgment.
- It structures the decision; it doesn't make it. Budget, the candidate's value, and the fairness call are yours.
- The equity check is only as good as `roster.md`. Stale or partial pay data means the fairness check compares against the wrong team.
If it doesn’t work
- No Cowork tab in Claude Desktop — update the app to the latest version and confirm you’re on a paid plan; Cowork isn’t on the free tier. On Windows, Cowork also needs the Virtual Machine Platform feature enabled — if the tab still won’t appear, that’s the fix.
- Claude can’t see
comp-philosophy.md,bands.md, orroster.md— the files aren’t in the project’s folder, or they’re in a different folder than the one the project owns. Open the project, check which folder it points to, and move the files there. Then re-run “list the files you can see.” - The market range arrives with no source — a confident range with nothing behind it is an estimate dressed as a fact. Ask: “Where does this range come from? If you can’t source it, label it an estimate.” Then validate against a real survey before the number reaches a candidate — the recipe structures the check; the survey is still on you.
- The equity check is generic — “ensure internal fairness” instead of “this puts the new hire above your two L4s at 122k and 128k” means Claude didn’t use
roster.md. Name the file in the prompt — “check this against the people inroster.md” — and confirm the file holds current pay data, not placeholders. The comparison against your actual team is the whole point.
Extra credit
Small additions that pay back the next time you run it.
- Keep roster data local and private — store
roster.mdin a connected local folder, not a shared drive, given its sensitivity. See the Platforms and Modes guide on local files. - Band review — once a year, paste the whole roster and ask Claude to flag where people have drifted below band or compressed against new hires.
- Offer letter — once the number’s set, hand it to an offer-letter workflow so the package and the letter stay consistent.
“Every comp decision is eventually compared to another one. Make it defensible now or explain it later.”
What this teaches you about Claude Cowork
The recipe is one application. The principles apply to everything you’d hand to Claude.
A consequential decision deserves a repeatable check. The reason comp problems happen is that the decision gets made by instinct under deadline. Running the same three-part structure every time — market, band, equity — is what turns a gut number into one you can defend.
Internal equity is the part everyone skips. Market data is easy to find; comparing a new number against your actual team is the work that prevents resentment and turnover. Storing the roster makes that check automatic instead of the thing you meant to do.
Show the reasoning on anything sensitive. For decisions that get scrutinized, a visible chain of logic matters more than a clean answer. The recipe insists on shown reasoning so the number survives the comp committee — and so you catch a bad assumption before it becomes an offer.
Who this is for
Founder or HR lead who owns offer numbers in Small companies making competitive hires (10–150 employees).
The pain: Offer numbers set by instinct that surface later as internal fairness problems
The outcome: Every offer is placed in a band and checked against the team before it's made
Published June 22, 2026 · 0 views