Quality

The Quality Factor: Screening for Businesses, Not Charts

Saral Team · 1 August 2026

Most retail screening in India happens on price. Quality investing is the argument that the more durable signal is in the annual report, and that it takes years rather than weeks to play out.

What “quality” actually means

The word is doing a lot of work, so it is worth pinning down. In the factor literature quality is not a vague judgment about management or moats. It is a small set of accounting ratios, standardised across a universe, that together describe how much profit a business generates per unit of capital and how fragile its balance sheet is.

Robert Novy-Marx’s 2013 paper on gross profitability is the cleanest statement of the core result: profitability predicts cross-sectional returns roughly as strongly as book-to-market does, and — importantly — it does so in the opposite direction to value. Profitable firms tend to be expensive, cheap firms tend to be unprofitable, so the two factors hedge each other. A portfolio holding both is more stable than either alone.

Fama and French added profitability and investment to their three-factor model in 2015 for the same reason: the three-factor model could not explain the returns of profitable firms, and adding a profitability leg fixed a real pricing failure rather than fitting noise.

Why it works

There are two competing explanations and they have different implications.

The risk story says profitable firms are riskier in some way not captured by beta, and the excess return is compensation. This is hard to square with the data — profitable firms have lower earnings volatility and lower default risk, not higher.

The mispricing story is more persuasive. Investors form expectations by extrapolating recent outcomes, and they extrapolate price more readily than they extrapolate profitability, because price is on the screen and profitability is in a PDF. Profitability is also unusually persistent: a company earning a high return on equity this year is very likely to earn one next year, far more reliably than a company with high returns this year will have high returns next year. When a durable characteristic is underweighted by the market, the result is a premium that persists until enough capital notices.

The practical consequence of the mispricing story is that the premium should be larger where attention is scarcer — smaller companies, thinner analyst coverage — and should decay as a factor becomes crowded. Both are consistent with what has been observed.

Building it from Indian filings

The catalog stores raw line items, not ratios, which is deliberate: a ratio is an opinion about which denominator matters, and different investors want different ones. Return on equity, for instance, is built rather than read:

roe             = net_profit_annual_consolidated / total_equity_annual_consolidated
debt_to_equity  = total_debt_annual_consolidated / total_equity_annual_consolidated

Three choices in those two lines are worth defending.

Consolidated, not standalone. Indian companies file both. Standalone accounts exclude subsidiaries, which for any group structure means excluding most of the business. A holding company’s standalone profit can be almost entirely dividend income from subsidiaries; its consolidated profit is the actual operating result. If you screen on standalone numbers across a universe containing both simple and group structures, you are comparing different things.

Annual, not quarterly. Quarterly figures are noisier, more seasonal, and in India frequently unaudited. For a factor that is supposed to capture a durable characteristic, the annual number is the right frequency, and the loss of timeliness costs little because the characteristic itself moves slowly.

Publication lag matters. A company’s FY2024 results are not knowable on 31 March 2024; they are filed weeks or months later. A backtest that reads them from the fiscal year end is reading the future. Any credible fundamentals backtest has to shift the effective date forward to when the filing actually became public, and if your tool does not do this you are measuring a forecast, not a strategy.

Blending several factors into one score

With three raw metrics on different scales — a ratio around 0.15, a ratio around 0.8, and a rupee figure in the hundreds of crores — you cannot simply add them. The standard fix is cross-sectional standardisation: convert each metric to a z-score across the universe, orient it so higher is always better, then average with weights.

Two refinements matter in practice.

Winsorization. A single company with near-zero equity produces an ROE of 4,000%, and one such row will dominate a z-score for the entire universe. Clipping the extreme couple of percent of each factor before scoring removes this without discarding the observation.

Direction. Leverage is scored lower-is-better while profitability is higher-is-better. Getting one of these backwards produces a screen that looks like it is working — it will still select thirty stocks and produce an equity curve — while systematically selecting the wrong ones. This is the most common silent bug in factor construction.

NSE’s Nifty200 Quality 30 uses exactly this construction on three inputs: return on equity, debt-to-equity, and earnings-growth variability, z-scored and averaged, with a 5% single-stock cap.

Why leverage is in there twice

Low debt-to-equity earns its place in a quality composite for a reason distinct from profitability. Profitability is about expected return. Leverage is about the shape of the distribution.

Highly indebted companies do not underperform gradually. They perform adequately for years and then fail discretely, which means a factor portfolio that ignores leverage carries a fat left tail that average returns conceal. In India this is not hypothetical — the 2018 IL&FS default and the subsequent NBFC credit freeze repriced leveraged balance sheets across several sectors within months, and no amount of profitability protected the names that needed to roll debt into that market.

Rebalance slowly

Quality metrics update when companies file, which is four times a year at most and meaningfully once a year. Rebalancing a quality screen monthly generates turnover — and therefore cost — against a signal that has not changed. Quarterly is the natural cadence; annual is defensible.

This is a general principle worth stating plainly: match rebalance frequency to signal decay, not to how often you feel like checking. A monthly rebalance on an annual signal is a cost programme with a strategy attached.

Try it

The Quality 30 template screens the Nifty 200, builds ROE and debt-to-equity from consolidated annual filings, blends them with operating profit into a winsorized z-scored composite, and sizes positions proportional to score. Quarterly rebalance.

Run it alongside the Value 40 template over the same window. The two will hold almost no names in common, and their drawdowns will not line up — which is the practical argument for combining them rather than picking one.