Universe ranking — 2026-07-28

How companies are ranked

Each company receives a composite score between 0 and 100. The higher the score, the more attractive on today's numbers. The score is composed of two equally weighted parts: Earnings Quality (how good is the earnings quality really?) and Valuation (do you get fair value at today's price?).

1. Earnings Quality — How good is the earnings quality?

We measure whether earnings are genuine, sustainable and efficient. Eleven components where the main weight is on 5-year averages rather than the latest year — this captures cyclical capital efficiency better and dampens one-off effects. The key question: does the company create value on the capital actually allocated? — this is answered by Incremental ROIC.

¹ Assessed only for manual deep-dives where footnotes have been read. For Börsdata extractions these are missing and the remaining components are reweighted so the sum equals 100%.

2. Valuation — Do you get good value at the price?

Six components measuring what you pay for each unit of earnings, cash and capital value. Three yields (return on your EV investment) and three multiples (what you pay per unit of earnings). A proprietary EP yield captures the unique mttssn metric — capital surplus above WACC.

Companies with losses (negative EBIT/EBITDA) are automatically classified as most expensive in percentile rankings. Implied growth is calculated via reverse-DCF for reference but is not included in scoring.

3. How scores are combined

Each component is ranked in percentiles within the universe — i.e. compared with all other ~650 companies. A component at percentile 80 means the company is better than 80% of the comparison group on that specific dimension.

If data is missing for a component (common for Börsdata extractions that cannot read footnotes, or for young companies without 5-year history) it is excluded — the remaining components are reweighted so the sum equals 100%.

4. Five-step rating

Composite scores are translated into an easy-to-read rating:

5. Three analysis modes — manual, streamlined, Börsdata

You see three "Source" badges in the ranking table:

Both Manual and Streamlined are permanently protected — Börsdata never overwrites them. Streamlined captures the most important mttssn adjustments (R&D-cap, PPA, restructuring, APM-bridge) with a time cost of 15-25 min per company. Manual takes 30-45 min and additionally captures nuances in acquisition detail, tax structures, contingencies and customer concentration.

6. Adjusted ROIC and Economic Profit — the core of the mttssn method

Both measures are shown in the detail view for each company and underpin several of the components above (ROIC 5y in Earnings Quality, EP yield in Valuation).

Adjusted ROIC = Adjusted NOPAT ÷ Adjusted Invested Capital. This is ROIC after removing accounting noise: capitalised R&D is removed from the capital base (it is actually a cost), PPA amortisation is kept in opex (it is genuine intangible wear), restructurings are normalised. For Börsdata extractions, GAAP-EBIT is used directly without these footnote adjustments.

Economic Profit = NOPAT − 8% × Invested Capital. This is the profit above the cost of capital. Positive EP = the company creates value. Negative EP = the company destroys value, even if earnings appear positive in the report.

7. What the rating does not tell you

• Sectors cannot be directly compared. A bank with ROIC 15% is not the same thing as an industrial company with ROIC 15% — they have different capital structures. The percentile ranking mixes them.
• Preference shares are excluded because their numbers do not represent operating activity.
• WACC is assumed to be 8% blended across the entire universe. For companies with a higher cost of capital (cyclical, small, highly leveraged) the capital charge is underestimated, making EP yield look better than it is. Use manual analysis for large positions.
• The 5-year averages require history. Companies listed less than 5 years ago or after spin-offs receive incomplete components and are weighted on the remaining ones — can be misleading for them.
• Manual deep-dives provide nuances that structured data misses. APM transparency and R&D discipline are only scored for manual analyses.

8. How to best use the rating

The ranking is a screening function, not an investment decision. Use it like this:

• Top-30 Attractive → candidates for deep analysis. Click the row, read flags.
• High EQ + low Val → quality company that is expensively priced (classic "quality compounder" trap)
• High Val + medium EQ → cyclical or misunderstood companies — requires sector understanding
• Bottom-30 Very Unattractive → avoid, or deep-analyse for a short thesis

Use the "Manual" badge as a marker for the reliability of the analysis. Companies with a manual rating have undergone full footnote reading and are more reliable than Börsdata-only versions.

9. The Conviction layer & the Value-Momentum option

The screen above ranks the whole universe on today's numbers. The Conviction layer (the landing page) is a separate, higher-bar view built only from names we have actually deep-dived — each with an authored scenario valuation (bull / base / bear with probabilities), a recommendation and a conviction level. The conviction portfolio is sized by risk-adjusted expected return × screen agreement, not by the screen score; the screen acts as a weight haircut and a disagreement flag, never a hard gate.

It applies Pring's principle that the price trend should confirm the fundamental case before you commit capital — the defence against the single biggest value-investing failure mode, the value trap (cheap, and staying cheap). Two stages:

• Stage 1 — Value selects. The conviction layer chooses the eligible universe and the base weight (quality + valuation + scenario). Unchanged.
• Stage 2 — Momentum sizes & times. The spine is 12-1 month momentum (the trailing-year return, skipping the most recent month — the canonical value+momentum factor of Asness–Moskowitz–Pedersen, Value and Momentum Everywhere), z-scored across names to size positions continuously. A fast 40-week (≈200-day) moving-average gate then times entries (the Faber/CTA standard), while the slow 200-week MA and the 52-week high are kept only to flag the over-extended. The earlier design used the 200-week MA as the gate — too slow: it only fires after a name has already round-tripped most of a cycle, late on both the avoid- and buy-side.

Broken-trend value names aren't discarded — they move to a bench ("awaiting trend confirmation") and re-enter the book when momentum turns and the 40-week trend is reclaimed. Signals are computed from Börsdata weekly price history (Nordic and US); young listings with under ~40 weeks use the longest window available (flagged short history), and names with under ~30 weeks honestly show "no data" rather than a fabricated signal. Position size also blends in a second, less crash-prone leg — fundamental momentum: realized ROIC and operating margin vs each name’s own 5-year baseline, plus 5-year revenue growth (Novy-Marx). It modulates size only (never un-benches a value trap), so deteriorating fundamentals shrink a high-price-momentum name while an improving ROIC lifts a quiet one. Price momentum is weighted 65%, fundamentals 35%; the ΔROIC column on the conviction page shows the headline driver.

Survivorship bias caveat

mttssn research · Adjusted Fundamentals · Methodological inspiration from New Constructs (Trainer et al.), Pring "Weight of the Evidence", and the EVA methodology (Stern Stewart). Implemented for Nordic Mid + Large Cap.