Read a hundred substitutions with the store manager.

Pull 100 substitution decisions from one store’s online picks in the last month. For each, the item ordered, the item picked instead, the picker, the time of day, and the shopper’s response where one exists: accepted, refused at the door, refunded, or nothing. Print them. Then sit with that store’s manager for two hours, in the store, and read them one at a time. For each decision the manager says whether they would have made the same swap, and why. For the first fifty you say nothing. For the second fifty the two of you write the reason down in one sentence.

What a shopper does when the item is missing.

The stockout study the industry still cites is from 2002. Gruen, Corsten and Bharadwaj, funded by a supplier and commissioned with the retailers’ association, measured what a shopper did in the aisle when the item was not there. 45% took an alternative, 26% a different brand and 19% a different size of the same brand. 31% went to a different store. 15% waited, and 9% did not buy. The study is old, it measured the aisle and never saw an online order, and the report uses it for the shape of the risk and nothing more. The 45% chose the alternative themselves, with the pack in their hand and tonight’s dinner in mind. In a picked order an associate makes that choice for a household they have never met.

The most delicate judgment, handed to the least-briefed person.

The argument is the report’s own, since nothing published measures substitution acceptance by fulfillment method. A swap is harder to get right in a picked order than in a self-served cart, because the picker does not know the recipe, the allergy or the child who will refuse the other brand. What the picker has is a handheld, a time target per order and a screen that says substitute if unavailable. In this release 46% of the 200 grocers tell the shopper a substitution may happen and how the rule works. 25% let the shopper accept or refuse it online before the order leaves, and 8% do that in real time while the pick is happening. 41% allow a refusal only at pickup, after the item has been picked, bagged and driven, and 34% give no say at all. Picker-to-shopper messaging, the channel an approval would travel on, is at Advanced or better at 12%. So at 75% of grocers, the picker’s guess is the last word.

The cold items in the hundred.

Many of the hundred will be fresh or frozen. FMI’s State of Fresh Foods puts fresh at 42% of grocery sales and finds that 9.3% of fresh sales are made online. In this release 31% of grocers send refrigerated and frozen items in dedicated insulated packaging. 49% bag them separately and no more, 20% do not separate them at all, and 2%, four banners, monitor the temperature and can say the order stayed cold. So note, for each cold substitution, whether the item picked instead went in the cold bag or an ordinary one. A swap the shopper would have accepted still comes back as a refund if it arrived warm.

How to read a hundred decisions.

Sort them three ways. By reason for the swap: the item was out at the pick, the item was out in the store all week, or the picker could not find it. By kind of swap: the same brand in a different size, a different brand in the same size, a different product, or nothing. By response: accepted, refused, refunded, silent. The silent rows will be most of the hundred at most grocers, and silence is not acceptance. The manager will see the patterns before you do. The same item swapped every Tuesday because the delivery lands on Wednesday. The picker who swaps organic for conventional to save the household money. The pack size nobody should ever change. Each pattern is a rule.

What the ten rules look like.

The store manager writes them, in the words the handheld uses. Never substitute infant formula. A different size of the same brand before a different brand. Organic stays organic. A refrigerated item is replaced with a refrigerated item and goes in the cold bag. If this household refused this swap before, do not offer it again. On a fresh meat cut, nothing rather than a guess. Ten of them and no more, because a picker on a Saturday morning will keep ten in their head and will not open a document. The manager writes them because the manager will have to defend them to the pickers, and because a rule written by the digital team reads as a policy while a rule written by the manager reads as an instruction.

The chapter’s decision is that in every store that picks, the shopper approves or refuses each substitution online while the pick is happening, with the picker’s alternative shown. The ten rules are what the picker does until that exists, and what the shopper is approving once it does. Three measures tell you whether the rules are working: the substitution acceptance rate, refunds for unwanted substitutions per 1,000 orders, and whether the household orders again in the 30 days after an order with a swap in it. Then pull the next hundred with the same manager, and see how many of the ten survive.