Open-to-buy is a control loop, not a quarterly report

Borrow one idea from engineering and open-to-buy suddenly makes sense. A control loop has three parts: a set point, a measurement of where you actually are, and a correction that closes the gap between the two. A thermostat is the cleanest example. You set a temperature, it reads the room, and it turns the heat up or down until the reading matches the target. Take away any one of the three and it stops working.

Open-to-buy is supposed to be exactly this. The set point is your inventory plan. The measurement is live sell-through. The correction is the reallocation of spend toward what is selling and away from what is not. Run properly, OTB keeps your inventory dollars tracking demand the way a thermostat keeps a room at temperature, continuously, with small corrections, all season long.

But most brands do not run it as a loop. They run it as a report. You set the plan, you measure once a quarter, and by the time the reading comes back the season is nearly over. That is not a control loop. That is an open loop, and an open loop, in control terms, cannot control anything. It can only describe, after the fact, how far off you drifted.

Map the three parts onto how OTB actually gets run and the failure is obvious. The set point exists: everyone has an inventory plan and an OTB budget by category. The measurement exists too, technically: sell-through is sitting in the data warehouse. What is missing is the correction, and specifically the correction happening fast enough to matter. The reading comes in monthly or quarterly, the response takes a week of manual rework, and by the time the buy is adjusted the demand that triggered it has moved on.

A thermostat that read the room once a quarter and took a week to respond would let the house freeze in the morning and bake by afternoon, and you would call it broken. We accept exactly that behaviour from open-to-buy and call it the planning calendar.

Control engineers have a name for the delay between a change and the response to it: lag. And they know that lag is not a minor inconvenience, it is the thing that makes a loop unstable. Too much lag and a controller does not just respond late, it over-corrects, because by the time it acts the situation has already moved, so its correction is aimed at a problem that no longer exists. This is exactly what happens with a quarterly OTB. By the time the buyer reacts to a category running hot, the season may have turned, and now they are chasing a runner that is cooling, or dumping a laggard that just found its feet. The lag does not only slow the loop. It makes the corrections wrong.

Here is what an open loop does when demand moves and the correction never comes. The categories running ahead of plan sell down their inventory and stock out at full price, so you lose the sales you would most profitably have made. The categories running behind keep receiving against a plan the market rejected, so stock piles up until you discount it to clear. Both are the same failure: the loop did not close, so the drift became the outcome.

The drift is not random noise you can average away. It is directional and it compounds. Every week the loop stays open, the runners get more starved and the laggards get more bloated, because the buy keeps flowing against a set point the measurement already contradicted. By quarter close, when the report finally lands, the drift is baked into the P&L as lost full-price sales on one side and markdown on the other.

This is the part that makes the open-loop framing more than a metaphor. In an open loop, error does not self-correct, it accumulates, because nothing is feeding the output back to the input. Your inventory plan is the input. Your sell-through is the output. If the output never reaches back to adjust the input, the two are free to drift apart with no limit except the end of the season, when reality forcibly reconciles them at whatever price the clearance rack sets. A closed loop, by contrast, keeps error small by construction: it never lets the gap grow large enough to matter, because it is always nudging the input back toward the target. The difference between the two brands is not effort or talent. It is whether the wire from output back to input is connected.

Fourteen points of full-price sell-through is the visible signature of a closed loop versus an open one. The brand correcting weekly keeps its dollars pointed at demand and clears at full price. The brand reporting quarterly discovers the drift after it is too late to act on, and pays for it at markdown. Same plan, same buyers, different loop.

Closing the loop means two changes. First, the measurement has to be continuous, not periodic: OTB reads live sell-through, returns and channel mix, not a monthly extract. Second, the correction has to be fast and bounded: the system reallocates spend toward the runners and away from the laggards automatically, but only inside the guardrails you set.

Those guardrails are what make an automated correction safe. The working-capital ceiling, the markdown headroom, the MOQ and lead-time floors become the limits the loop operates within, the way a thermostat has a maximum and minimum it will not exceed. Inside those bounds, the agent stages the moves: release spend from what stopped selling, propose reorders on what is pulling, and present the set for the buyer to approve in one pass. The buyer is the operator setting the target and signing off the corrections, not the mechanism grinding through the math by hand.

There is a good reason to keep the human on the loop rather than fully out of it. A control loop is only as good as its set point, and set points are a judgment call: how much markdown headroom to accept, when a strategic bet overrides the numbers, where the brand wants to lean in regardless of last week's sell-through. Those are decisions the buyer should own. What the buyer should not own is the grinding recomputation between decisions, the part that used to eat a week and arrive too late to use. Let the loop handle the measurement and the mechanical correction, keep the human on the set point and the approval, and you get the speed of automation with the judgment of a merchant. That is what a well-designed control loop looks like in a business: the machine holds the target, the person decides what the target should be.

That capital is the drift you stop paying for. When the loop closes, dollars stop getting stranded in the laggards and stop getting starved from the runners, and the difference shows up as cash freed and margin held. You did not buy less. You kept the buy pointed at demand.

A quarterly open-to-buy is a thermostat you read once a season. It does not control the temperature. It just tells you, too late, how cold the room got.

The framing matters because it tells you what to fix. The problem was never the set point; your plan is probably fine. The problem is a loop that is open where it needs to be closed: a measurement that arrives too slowly and a correction that takes too long. Close the loop, keep the guardrails, and open-to-buy goes back to doing the one job it was ever meant to do, which is keeping your inventory dollars on demand while the season is still running.