The Closelook Company Score — Three Layers, Not One Number
Most scoring products compress a company into one proprietary grade and ask you to trust it. Closelooknet deliberately does the opposite: every stock page separates the raw statements, the established metrics from the finance literature, and our own score — so you can rebuild the judgment yourself, or replace it with your own.
The Three Layers
Layer one is raw data: the statements themselves — revenue, margins, cash flows, balance-sheet items across the last fiscal years, refreshed from vendor filings. Layer two is the established toolkit: the ratios and composite scores the finance literature has validated for decades — P/E, PEG, EV/Sales, FCF yield, ROIC, accruals, Altman Z, Piotroski F, Beneish M — computed with their canonical formulas and shown with their standard interpretation zones. Layer three is ours: the Closelook Company Score. The layers never blend. If you disagree with our weights, layers one and two are complete without us.
The Seven Modules
Our score ranks every metric cross-sectionally across the ~334 companies in the five Closelook index families, then averages the percentile ranks into seven module scores: Valuation (is the price reasonable against growth and quality), Quality (is the business structurally profitable — ROIC, margins), Growth (is it expanding), Balance Sheet (can it survive stress), Earnings Quality (are profits cash-backed — the accruals lens few investors use), Capital Allocation (value creation versus dilution), and Momentum (does the market confirm the story). A metric with no data is dropped and the remaining weights renormalize — we never impute a neutral value.
Weight Sets, Versioned
One weight set cannot fit a semiconductor cyclical and a software business at once, so the composite applies one of three profiles by company type: general (quality and growth led), software/agentic infrastructure (growth 25%, quality 20%, cash quality 15%), and semis/cyclicals (momentum and balance-sheet led). The weights are judgment, which is why they are printed next to every score and versioned — currently v1. When they change, the change is documented, never silent.
What We Deliberately Do Not Do
We do not turn the score into a recommendation — Closelooknet is an investment diary and research publication, and the score is descriptive context, not advice. We do not hide the inputs behind the composite. And we do not average away warning flags: the Beneish M-Score is shown but excluded from the composite, because a manipulation probe is a prompt to read the filings, not a quantity to blend.