Role Ledger — website to scorecard

Two inputs

Give it the company site and the job description. It reads the site, infers the OKRs with evidence, then maps the role against them.

Company website

Job description

Site won't read? Paste it instead

Some sites render entirely in the browser — React, Lovable, Framer — so a crawler sees an empty shell. Others are gated or too new to be indexed. Open the site, select the homepage, pricing, about and careers pages, and paste the visible text here. Objectives get inferred from this instead of from a search.

Nothing pasted — the site will be searched instead.

Running this outside Claude?

Inside a Claude artifact the model call is authenticated for you. Opened from your disk or your own server, it isn't — paste a key and it will call the API directly.

Anthropic API key

Kept in this tab's memory only — never saved to the file, never stored. Anyone you share the page with enters their own.

01Read the pages that carry intent
02Infer objectives & key results
03Pull KRAs and KPIs from the JD
04Link, weight, and find the gaps
05Break it into tasks and cadence

The ledger is empty.

Paste a URL and a JD on the left. Everything below is produced from those two things and nothing else — every inferred objective arrives with the page it came from, linked, so you can check it.

Objectives inferred from the site, each with its evidence and source link
3–5 KRAs from the JD, one KPI each, with target, source and cadence
Every KPI linked to a named KR, tagged BAU, or cut
Weights that sum to 100, and a score you can run

The method this follows

Settle the OKRs first — official ones if they exist, inferred and labelled as inferred if they don't.
Read the JD, not the OKR, to find the KRAs. Only what's quantifiable qualifies.
Limit to 3–5 KRAs. More and focus dilutes until nothing is prioritised.
Attach a KPI to each — a number, a target, a data source, a cadence.
Link every KPI to a specific KR, or tag it BAU, or delete it. Then read the map top-down: any KR with no KPI pointing at it is unowned.
Accept that enabler and innovation work won't map. Keep it on a separate list rather than inventing a number.
Weight the KRAs to sum to 100. Weight follows outcome ownership, not effort.
Score: achievement = actual ÷ target, capped at 120%. Final = Σ (weight × achievement).
KRAs hold for the year; KPIs get revisited quarterly.

A marketing site lags real strategy by a quarter or two and never shows what isn't working. What comes out here is a strong first draft to take to the founders — not a substitute for asking them.