Two supplier trials, four weeks each: what the data must show

Both pads passed. That was the conclusion of a four-week trial that had cost two people most of their Mondays, and it was useless. Nothing had failed, nothing had been measured properly, and the decision went back to price. A trial that cannot separate two products is a trial that should not have been run.

The design is not complicated. It is four variables held still and five numbers written down every day.

Fix the four variables

Product, pack format, position and interval. Same species and cut in both arms. Same tray size and film. Same shelf height in the same chiller. Same change interval, written down and enforced, even if one pad could last longer. If any of those move between arms, the comparison is between two conditions rather than two pads.

Position catches people out. The top shelf of a chiller runs two degrees warmer than the bottom, and the difference is enough to change drip by a fifth. Alternate the arms between shelves every week, or run both at once on the same shelf.

Interval is the variable that gets quietly abandoned. Week one runs to plan. By week three the busy Saturday has pushed the change time out by an hour, and the pad in the second arm looks worse for a reason that has nothing to do with the pad. Put the interval on the line board and check it twice a week.

The five numbers to record daily

One sheet of paper per day, five fields:

• Packs built and pads used, which gives sheets per pack.
• Weight of ten spent pads at change, which gives grams absorbed.
• Packs with free liquid at the end of the interval, counted, not estimated.
• Changes per shift, including unscheduled ones.
• Any complaint, markdown or rejection traceable to a wet pack.

Grams absorbed is the one that separates grades. Sheets per pack is the one that predicts cost. The other three tell you whether the difference matters on the shelf.

Weigh the same ten pads every day, not ten pads chosen at the end of a shift. A selection bias creeps in without anyone intending it: the person weighing will pick the pads that look worst, because those are the interesting ones, and the average will shift by 20 percent across a month.

Four weeks, not four days

A week-long trial measures the pads. Four weeks measures the operation, including the delivery cycle, the weekend peak, the staff who work only Saturdays, and the one chiller that runs warm. Seasonality is the trap: a trial in a quiet February compared against a trial in a busy December will produce a confident wrong answer.

Run both arms at the same time if you can, in different aisles of the same department, and swap them at the halfway point. Two weeks each way cancels the shelf and the week. Note the swap date on the same sheet as the daily numbers, or the change will be forgotten by the person doing the weighing.

Deciding, and what a tie means

If one pad wins on grams absorbed and sheets per pack and ties on complaints, the decision is easy. If they tie on everything, the tie is the finding, and the next question is commercial: lead time, carton size, minimum order, and whether the supplier will hold stock for you.

Do not let a trial run into a second month because the result is unclear. Two four-week trials in one year is a reasonable pace. A trial that rolls on indefinitely costs more in management attention than the pads are worth, and it usually ends with the incumbent staying on price anyway.

The trial above ended in a tie because nobody weighed the spent pads. Both products looked the same wet and both went in the bin at the same time. Ten pads on a kitchen scale each morning would have settled it in a week. A trial record template and the measurement method are set out in the absorbent pad user guide.