Methodology

Both engines are deterministic arithmetic. AI is used only to propose starting assumptions and to draft narrative — it never computes, adjusts or overrides a number. Every figure on screen can be reproduced by hand from the formulas on this page.

Nothing is estimated to fill a gap. If a required input is not in the filed data, the dependent metric is reported as Data unavailable, its weight is removed from the composite denominator, and the header reports what percentage of the framework was scorable.

Every input carries a provenance tag — Verified (filed statement or live feed), Consensus, Guidance or Assumption (analyst entry) — and that tag travels through to the report.

The model starts from the most recent verified historical year and rolls forward one year at a time, independently for each scenario (bear, base, bull). Each scenario owns its own complete assumption set, so the three paths never share intermediate values.

Terminal value is deliberately the average of two independent methods, so a single aggressive input cannot dominate the answer.

If terminal growth is greater than or equal to WACC the perpetuity term is set to zero rather than returning a negative or infinite value, and the exit multiple carries the terminal value alone.

The headline price target is the probability-weighted average of the three scenario fair values. Weights are normalised by the probability actually assigned, so the target stays valid even if the three probabilities do not sum to exactly 100%.

The sensitivity grid re-runs the entire DCF — not a shortcut approximation — across five WACC steps (−0.5, −0.25, 0, +0.25, +0.5 percentage points) crossed with five terminal-growth steps, producing 25 fully rebuilt valuations per scenario.

Every metric, whatever its unit, is converted to points by one shared linear mapping. There is no curve, no judgement and no peer-relative ranking inside the score itself.

Inverted metrics (where lower is better, such as net-debt/EBITDA) simply set worst above best; the same formula then runs downward. Each metric card in the app shows the band used, the arithmetic actually performed, and the source of every input.

An analyst may override any single metric, but the override requires a written justification, is stored with the report, and is displayed beside the calculated value so a reviewer can always see what was changed and why. Reverting restores the computed figure instantly.

Management Execution Score is the mean of six graded categories. Five are analyst grades against a published definition; share dilution is scored automatically from filed share counts so the most gameable input cannot be graded by opinion.

Thesis Break Score starts at 100 and only ever moves down. Every deduction must carry a date, a category and a written explanation, and the card shows the current score, the previous score and the direction of travel.