A factor is not its name. “Quality” is not a factor; it is a label that several competing constructions wear. One firm defines quality as gross profitability scaled by assets; another as the change in operating profit scaled by book equity; a third as a composite of accruals and investment. Each of these is a defensible answer to a different question, and the questions do not collapse into one another. Before any discussion of whether a factor “works”, the construction has already decided what is being tested.

The first decision is the signal variable — the raw quantity that carries the economic intuition. Momentum, for example, is usually the past twelve-month return excluding the most recent month, but the exclusion window, the lookback and the treatment of non-trading days all change the series. A reader who compares a “momentum” result across two papers without checking these definitions is comparing two different signals.

The second decision is sorting and weighting. An equal-weighted long-short sort on the top and bottom decile is not the same animal as a continuous z-score weighting across the whole cross-section. The decile construction concentrates exposure in the tails and is sensitive to small samples; the z-score construction spreads exposure and is usually closer to a portfolio that could actually be held. A factor that looks strong under decile sorting can become a thin, noisy spread under continuous weighting.

The third decision is neutralisation. A raw factor return is rarely what a portfolio manager can trade, because it carries market beta, sector tilts and size exposure that may be unwanted or already held elsewhere. Neutralising against the market, against industry groups, or against a size factor changes both the average return and the volatility of the residual. Two papers that report “the quality factor” with different neutralisation baselines are not in disagreement; they are measuring different residual quantities.

The label is the last thing to check. The construction is the first.

What this means in practice is that a factor result is only as comparable as its construction notes are complete. A paper that reports a factor’s average return, Sharpe and turnover without specifying the signal variable, the weighting and the neutralisation has not reported a result that can be reproduced; it has reported a number. The honest reading of a factor, then, begins not with its label or its headline statistic but with the construction pipeline that produced it — because that pipeline is the factor.

For investors who use third-party factor data, the same caution applies in reverse. A vendor’s “value” factor is a specific construction with specific choices baked in, and those choices interact with the user’s own portfolio in ways the label does not reveal. Treating the label as the factor is the most common, and the most expensive, reading error in factor investing.