Your own order book, and what it is allowed to prove

The only demand data you own

Marrash cannot see what Ovenden sells in its shops. It can see every order Ovenden has ever placed with it: every quantity, every size ratio, every colour, every date. That record is small. It is biased. It arrives late. It is also the only demand data the factory owns outright, so it is worth reading properly.

Reading it properly means two jobs at once. Count what is there. Then be strict about what the count is allowed to prove.

Repeats, and how late they arrive

Marrash went through six seasons of shipping records. For each buyer it counted how many styles it shipped, and how many of those came back as a second order.

The median is the middle value. Half the repeats came sooner than that, half came later.

Compare Stilbrook with Ovenden. Stilbrook repeats just over twice as often, and it repeats 2.5 times faster. Marrash's first thought was that Stilbrook is the better customer. That is not what the table says.

Stilbrook sells online only. Between a shopper deciding to buy and Stilbrook placing a repeat order there is one warehouse, and no allocation out to 340 shops. Ovenden's repeat has to get through a store-by-store allocation, a regional trading review, and an open-to-buy line that somebody else has already spent. Open-to-buy is the money a retail buyer still has left to spend in a given period. So the repeat rate measures how far the buyer sits from its own shopper. It does not measure how good the product is. Two buyers can sell the same garment at the same speed and still repeat completely differently. Marrash would read that as a difference between the garments.

That is the first rule for reading your own book. A number about a buyer is not a number about demand until you have taken the buyer's own machinery out of it.

What a repeat proves, and what an absence proves

A repeat is real evidence, but narrow evidence. It proves one thing: at that buyer, the style hit whatever internal trigger makes that buyer reorder. It does not tell you how fast the style sold. You cannot see their sales, their markdowns, or how many shops ever got the style at all.

The absence of a repeat proves almost nothing. This is where factories mislead themselves, week after week. Marrash listed the reasons a good style does not come back:

• the buyer had already spent its open-to-buy for that period;
• the season ended before the reorder window opened;
• the buyer moved the style to another supplier, on price or on capacity;
• the buyer rebuilt the range and the style no longer had a slot;
• it did not sell.

Five reasons, and only one of them is about demand. So a missing repeat is one-fifth of an argument at best. Treat it as a verdict on the product and you kill a style that was working. If you cannot tell which of the five it was, write unknown in the ledger. An honest blank beats a confident wrong reading, and it costs nothing to leave.

The size curve is the sharpest thing in the book

A quantity tells you how much a buyer committed. A size curve tells you something about who is buying. A size curve is how an order splits across the sizes. It changes from one order to the next, which makes it the most useful field on a purchase order.

Ovenden's first order for OV-2140 and its repeat carried different ratios:

The curve moved up. Marrash's merchandiser read that as the shopper being bigger than the first buy assumed. Then he stopped, because the reading depends on a question the purchase order does not answer.

If the repeat is a replenishment, the curve is built from what has sold in the shops, and moving up means the bigger sizes are selling faster. If the repeat is a second buy for new shops, the curve is built from what Ovenden's planner now believes. That may be the same thing, or it may be a correction of a curve that was wrong the first time. The two readings point at different actions, and the document looks identical either way.

So the merchandiser sent one line of email: is this a replenishment against sales, or a second buy? It cost nothing, and it is the difference between data and a guess.

The correction nobody applies

Stilbrook's curve carries a fault that Ovenden's does not. Marrash found it because it works for both of them.

A size curve is normally built from units sold. Stilbrook sells online, and its own measured return rate — Stilbrook's figure, given to Marrash — is 34.1% on its end sizes and 21.8% on the middle three. Applied across its own gross curve, those two rates blend to 24.4% overall. Gross means before returns are taken off.

Returns do not spread evenly. They are heaviest at the ends of the curve, because that is where a shopper orders two sizes to keep one. So a curve built on gross sales asks for too much of exactly the sizes that come back.

810 pieces are cut in the two sizes most likely to be sent back, and the middle of the range is short by the same 810. At Stilbrook's own stated cost of USD 3.15 to handle a return, the handling alone on one style is USD 2,551.50. And that is before the middle sizes run out early and the style dies with stock still in the warehouse.

That cost is Stilbrook's, not the factory's, which is exactly why it matters. Marrash is the only party in the chain that sees size curves from a high-street chain, a wholesale brand and an online-only brand side by side. Stilbrook sees only its own. The signal you can hand back is worth more than the signal you receive. It is the one thing a factory can offer a buyer that a cheaper factory cannot copy, and it costs a spreadsheet.

Act as a factory merchandising manager who is strict about what data is allowed to prove. Read my own order book as a demand signal, and keep the reading separate from the caveats. Here is what I have. For each buyer: [BUYER NAME], [CHANNEL - HIGH STREET, WHOLESALE, ONLINE, DEPARTMENT STORE], [NUMBER OF DOORS OR NONE]. For each style shipped in the last [NUMBER] seasons: [STYLE CODE], [FIRST ORDER QUANTITY], [FIRST SHIP DATE], [REPEAT ORDER QUANTITY OR NONE], [REPEAT ORDER DATE], [FIRST ORDER SIZE RATIO], [REPEAT SIZE RATIO], [COLOURS ORDERED FIRST], [COLOURS REORDERED]. Do the following. First, give me a repeat rate and a median days-to-repeat for each buyer. Say plainly how much of the gap between buyers comes from how far each buyer sits from its own customer, rather than from the product. Second, for every style that did NOT repeat, list the possible reasons - open-to-buy already committed, season closed, moved to another vendor, range rebuilt, did not sell - and tell me the one question to the buyer that would separate them. Do not rank the styles by repeat rate as though a non-repeat were a verdict. Third, take each size-curve shift. Tell me what it would mean if the repeat was a replenishment against sales, and what it would mean if it was a second buy. Mark the orders I cannot tell apart. Fourth, for any online buyer, ask me for their measured return rate by size band. If I have it, rebuild the curve on sales net of returns, and show me the piece difference by size on my next order quantity. Fifth, write the ledger itself as a table with three columns - what was observed, whose demand it is about, how old it was when it reached me - and fill it in from what I gave you. Sixth, list every conclusion you could NOT draw, and say what data would have let you draw it. Where the sample is too small to carry a number, say so and give me the count instead of the percentage.