From a universe of 12 stocks with book to price, market value and a month's return, build a value factor long the top third and short the bottom third, show that it is mostly a small cap bet, then rebuild it as a size-neutral two by three sort.

The naive value factor returns 5.25%, but about 2.00 points of that is a size bet, and the size-neutral factor returns 3.25%. Three of the four value stocks are small and three of the four growth stocks are big, so in a month small caps rallied the naive factor looks better than value really did. Sorting on book to price inside small and big separately, then averaging, gives a value return of 3.25% and a separate size return of 5.08%.

Rank all twelve by book to price, buy the top four and sell the bottom four. The value leg earns 6.125% and the growth leg 0.875%, so the naive value factor returns 5.25% for the month. It looks like a strong month for value, and that is the claim the rest of the case tests.

Look at who is in each leg. 3 of the 4 value stocks are small and 3 of the 4 growth stocks are big: the value leg's average market value is Rs 4,625 crore against Rs 44,500 crore for the growth leg. Cheap stocks tend to be small, so a pure value sort quietly buys small and sells big. It is like judging whether a coaching method works by comparing its students with another centre's, when its students were also younger: two differences are tangled together.

Split the universe at the median market value into small and big, then within each half sort on book to price into value, neutral and growth, two stocks each. The value factor is the average of the two value cells minus the average of the two growth cells, so every rupee long value in small stocks is matched by a rupee short growth in small stocks, and the same in big. Small value beats small growth by 4.50 points and big value beats big growth by 2.00 points, so the size-neutral value factor is 3.25%. The same grid gives a size factor, small minus big across the three columns, of 5.08%.

Four, each with its reason. Weight by market value inside each cell, not equally, so the factor can be traded without piling into the smallest, least liquid names. Set the size breakpoint using larger stocks only, so a flood of tiny listings does not move it. Use accounting data with a lag of several months, so the factor never uses a book value that was not yet published. And consider sorting within industries, so value is not just a bet on whichever sectors are cheap. This construction follows the approach Fama and French set out in 1993; its limit is that one characteristic, book to price, is a blunt measure of cheapness for asset-light businesses.