Method
Method
Model version 2.0 · 77 securities computed as of 09/10/2026, 48 of them scored.
Where the data comes from
Three sources, all official. Quotations come from the Tunis Stock Exchange. Financial ratios are computed from the financial statements companies file on the exchange's website: we read the PDF and extract equity, net income, total assets and EBITDA. Quarterly revenue comes likewise from the activity indicators published each quarter. No aggregate is taken from a third-party service: when a figure is missing the cell stays empty rather than being borrowed.
Independence
Mizen holds no position in the securities it covers. The brokerage account attached to the tool that produces these pages (Kakuzu) is locked: no order can be sent from it. Every figure comes from a primary source — Tunis Stock Exchange, CMF, central bank, documents filed by the companies — and links to the document it is taken from; none is copied from an aggregator.
The formulas
| Performance over a period | computed as total return: the dividend is reinvested on its ex-date, otherwise a detachment would read as a fall. Today's total-return index ÷ index N days ago − 1. The ex-date is not published in a usable field, so it is recovered from prices around the payment date. Performance is published only if the history starts before the target date; otherwise it would cover a shorter period than its label says. |
| Year to date | computed from the last close of the previous year, not from 1 January, which is not a trading session. |
| Annualised volatility | standard deviation of daily log returns over one year, multiplied by the square root of 252 sessions. Minimum 30 sessions. Computed on the total-return series: a dividend detachment is not a market move. Every market session counts, the last price being carried over on days the security did not trade. |
| Beta | covariance of the security's returns with TUNINDEX ÷ variance of TUNINDEX, over market sessions (minimum 60), the security's last price carried over on days with no trade. Security returns are total returns. |
| Maximum drawdown | largest fall from a peak to the trough that follows it, over a rolling year, on total return. |
| Market capitalisation | price × number of shares admitted to listing. |
| Free-float capitalisation | market capitalisation × official free float. The float comes from the exchange's index committee notice, rounded to 10%: the share of capital deemed tradable, the rest being held by control shareholders. For 53 of the 63 covered stocks, ranking by size changes between the two measures. |
| Reservation band | bounds the price cannot cross next session: last price ± the threshold of its quotation group (±6% continuous, ±4.5% auction), rounded inward to the 10-millime tick. The group is read from the Official Bulletin. Indicative only: the exchange decides reservations and its notice prevails. |
| Earnings per share | net income for the year ÷ listed shares. |
| P/E | price ÷ earnings per share. Not published when there is a loss, where the ratio is meaningless. |
| Price / book | price ÷ (equity ÷ listed shares). |
| Return on equity | net income ÷ equity. |
| Return on assets | net income ÷ total assets. |
| Enterprise value / EBITDA | (market cap + financial debt − cash) ÷ EBITDA. |
| Daily liquidity | median value traded over the last 120 market sessions; a session in which the security did not trade counts as zero. |
| Dividend per share | amount read from the payment notice filed by the company after its general meeting. The financial year is the one it states; failing that, the year preceding the meeting. |
| Dividend yield | latest dividend per share ÷ today's price. Not published if the last known dividend is more than two financial years old. |
| Payout ratio | (dividend × listed shares) ÷ net income. |
The composite score
Model v2 — first, the universe. Only securities one can actually buy and sell are ranked: the median value traded over the last 120 market sessions, a session with no trade counting as zero, must reach 5,000 TND. A security already ranked only leaves below 3,000 TND, so that one sitting on the threshold does not reshuffle the ranking every other day. The others keep their page and ratios, without a rank, with the reason ("Not ranked: traded in 8 of 120 sessions, median 0 TND"); their volatility, beta and maximum drawdown are withheld, since a price line that does not move is not a zero risk.
Seven factors. Each is converted into a percentile rank from 0 to 100 among ranked securities. Percentile ranking avoids capping ratios arbitrarily: an extreme P/E ends up last without distorting the scale. The weights below add up to 100%: they are declared choices, not a statistical calibration.
| Factor | Weight | Direction |
|---|---|---|
| P/Eprice / earnings per share, from our reading of the financial statements | 20 % | lower is better |
| Return on equitynet income / equity, latest financial year | 20 % | higher is better |
| Revenue growthyear-to-date total against the same total a year earlier | 15 % | higher is better |
| 1 year12-month total return, dividends reinvested | 15 % | higher is better |
| Dividend yieldmin(dividend, earnings per share) / price; no dividend = 0 | 10 % | higher is better |
| Volatilityon the market-session calendar, last price carried over on days with no trade | 10 % | lower is better |
| Daily liquiditymedian value traded over 120 sessions, a session with no trade = 0 | 10 % | higher is better |
Displayed, not scored. Price-to-book and return on assets stay on the security pages but no longer score: the first repeats P/E and return on equity (ln P/E = ln P/B − ln ROE), the second repeats return on equity and penalises banks by construction.
Missing data. A known unfavourable fact takes the last rank: a loss-making company (P/E), negative equity with a loss (return on equity), no dividend or one more than two financial years old (zero yield), a ratio outside its bound on the unfavourable side. An unknown counts as 50, the median. The score is the weighted average of the seven percentiles; it is published only if the factors actually measured carry at least 70% of the model's weight. The yield is scored on the lower of dividend and earnings per share: paying out more than one earns is not sustainable. Revenue growth compares the year-to-date total with the same total a year earlier, at the quarter published by at least two thirds of ranked securities. Outside banks, leasing and insurance, return on equity is withheld when equity is below 10% of total assets.
Investment companies (SICAF, SICAR). PLTU, SPDIT (SICAF) and TINV (SICAR) are scored neither on P/E, nor on return on equity, nor on revenue growth: their profit is made of the dividends they receive. They are therefore not ranked until their revalued net assets are computed.
A tilt to be aware of. As of 09/10/2026, 6 of the top 10 securities are financials (banks, leasing, insurance). Their P/E is structurally low, which also reflects the cost of risk, not just a discount.
The score ranks securities against one another on a given date. It is neither a forecast nor a buy or sell recommendation.
Plausibility bounds
Outside these bounds the indicator is not displayed: a three-digit return ratio signals equity close to zero, and a four-digit P/E a residual profit or a PDF reading error. For the score, the side matters: beyond the unfavourable bound (P/E too high, return or growth too low) the security takes the last rank; beyond the favourable bound it is probably a reading error, and the factor counts as unknown.
| Indicator | Minimum | Maximum |
|---|---|---|
| P/E | 0.50 | 150.00 |
| Price / book value | 0.05 | 30.00 |
| Return on equity | -60.00 | 60.00 |
| Return on assets | -100.00 | 100.00 |
| Enterprise value / EBITDA | 0.50 | 100.00 |
| Dividend yield | 0.00 | 30.00 |
| Payout ratio | 0.00 | 300.00 |
| Revenue growth | -80.00 | 150.00 |
Current coverage
Ratios depend on automatic extraction from financial-statement PDFs, whose layout varies from one company to another. Here is what is actually available today.
| Indicator | Securities covered |
|---|---|
| P/E | 57 / 77 |
| Price / book value | 66 / 77 |
| Return on equity | 63 / 77 |
| Return on assets | 67 / 77 |
| Volatility | 52 / 77 |
| beta | 52 / 77 |
| Market capitalisation | 72 / 77 |
| Revenue growth | 72 / 77 |
| Enterprise value / EBITDA | 31 / 77 |
| Dividend yield | 59 / 77 |
| Payout ratio | 51 / 77 |
One item is not yet produced from exchange filings and therefore appears nowhere: the ownership structure, which will have to be extracted from annual reports and threshold-crossing declarations.
Sector medians
Each company page shows, next to the P/E, price/book, return on equity and dividend yield, the median for that company's sector. It is computed on the same day, from the same indicators, using only the sector constituents for which the indicator could be extracted — never an estimate. A median is published only from 5 constituents carrying the indicator; below that the page states that no median exists for the sector rather than showing one that would represent nothing. The sample size is shown on hover and repeated below.
| Sector | Published medians |
|---|---|
| Banks | Dividend yield 3.71 (n=11) · Price / book value 1.19 (n=12) · P/E 12.18 (n=10) · Return on equity 10.26 (n=12) · Volatility 26.85 (n=9) |
| Construction and materials | Price / book value 1.37 (n=7) · Return on equity 3.08 (n=5) |
| Financial services | Dividend yield 4.78 (n=10) · Price / book value 1.67 (n=9) · P/E 13.11 (n=10) · Return on equity 12.46 (n=9) · Volatility 30.93 (n=7) |
| Household and personal care | Dividend yield 3.26 (n=5) · Price / book value 1.65 (n=5) · Return on equity 13.09 (n=5) |
| Insurance | Dividend yield 3.38 (n=6) · Price / book value 1.88 (n=7) · P/E 17.55 (n=7) · Return on equity 9.79 (n=7) · Volatility 31.12 (n=5) |
| Retail | Dividend yield 3.11 (n=7) · Price / book value 2.76 (n=9) · P/E 15.50 (n=9) · Return on equity 17.81 (n=9) · Volatility 29.64 (n=8) |
The gap to the median is shown by a chevron, never by colour: being above the median is neither good nor bad in itself — a high P/E means the stock is expensive, a high return on equity means it is profitable. The comparison positions, it does not conclude. Sectors missing from the table above do not have enough rated constituents for a median to mean anything.
Numerals and separators
All three editions use Latin numerals (0-9), including the Arabic edition. This is not a translation gap: the ar-TN locale defaults to that numbering system, as do ar-MA and ar-DZ, and the Tunis Stock Exchange, the Central Bank, the official gazette and filed financial statements all publish that way. A site that claims its sources reproduces their form.