The low stock badge is a promise. Breaking it costs more than never making it

2026-08-26

The low stock message is one of the few things in ecommerce that reliably produces a number in the right direction on the first test. Add "only 2 left" to a product page, watch add-to-cart rise, ship it to 100% of traffic, move on. I have done this. Most people reading this have done this.

Ecommerce
Customer Experience
UX
Retail

The badge that works until it works

The test that very few people carry out is the one that has commercial importance, namely what occurs to the customers who decided to buy the product on the basis of the badge only to discover that it had disappeared. Such customers are no minor issue. When the badge is honest, they are a direct result of having shown it, since you are intentionally directing demand towards the final few units of the product; and when the badge is not honest, they are also a direct result of that, although for different and more serious reasons.

The article concerns this second group of customers and the back-in-stock notification, even though most retailers regard it as merely an email plumbing feature when in fact it is the mechanism that determines how much a stockout costs you.

Scarcity works, but the source of the scarcity decides how much

Begin with what the evidence really supports. The meta-analysis on the effects of product scarcity, which was published in the Journal of Retailing in 2022, shows that scarcity cues increase purchase intentions and separates out the effect according to its source. Scarcity caused by limited supply, scarcity caused by high demand, and scarcity caused by a time limit are not interchangeable and do not have the same effect. It is when all three are treated as one lever, known as urgency, that many implementations go wrong.

More recent studies reinforce this point. A study carried out this year in the fields of Psychology and Marketing on scarcity in digital environments explains why cues relating to product supply are generally perceived as credible indications of actual availability, whereas social and temporal cues usually are not. For example, a pop-up stating that twelve people are looking at this item or a countdown which resets each time the page is refreshed leads to a completely different kind of interpretation.

That different processing has a name. The persuasion knowledge model holds that when a consumer recognises something as a persuasion attempt, they stop evaluating the message and start evaluating the messenger. The cue does not just lose its effect, it becomes evidence about you. Reduced perceived credibility is how this shows up in the data, and it does not stay confined to the cue that triggered it.

The effect for a trading team is that your stock badge is drawing credibility away from your inventory system. In the case where the badge is based on actual unit counts and is updated each time those counts change, it constitutes a factual disclosure and therefore receives an appropriate response. But if the badge triggers at a fixed threshold, or even worse if it activates according to a rule having nothing to do with stock levels, you are using up trust in order to secure conversions, and you will be charged for this at a rate which has not been modelled.

The disclosure makes the failure worse, not better

Here is the finding that changed how I think about this. Peinkofer, Esper and Howlett, writing in the Journal of Business Logistics in 2016, examined what happens when an online retailer discloses limited inventory availability and the item subsequently turns out to be unavailable to that customer. The counterintuitive result is that pre-stockout disclosure of low availability produced greater consumer dissatisfaction than no disclosure, in the situation where the stockout then occurred.

The idea put forward is that consumers compete with each other. A stockout that happens without any prior notice is simply a shortage, and shortages are annoying. However, if you had been informed that the stock was low and then the item was out of stock, that situation is equivalent to having entered a competition and lost to other shoppers. By changing the way the situation is viewed in this way, the emotional tone is completely altered, and it is completely up to you because it was you who provided that way of looking at it.

I want to be careful about the age of this. Ten years is a long time in ecommerce, and I do not think a 2016 experimental study should carry an argument on its own in 2026. So the honest question is whether more recent work confirms the direction or contradicts it.

It confirms it, and extends it. A 2025 paper in the Journal of Retailing, running four experiments and a field study, found that when the limited quantities in an online promotion come in below what the consumer expected, perceived retailer sincerity falls and purchase likelihood falls with it. The authors term this the scarce-insincere inference: the consumer makes a leap from a judgement about the product to a judgement about whether the retailer is genuine. The 2016 finding is about the emotional cost of a broken scarcity promise. The 2025 finding is about the attribution the customer makes afterwards, which is the more expensive of the two because it persists.

What the same research says about fixing it

The 2025 paper does not stop at the problem, and this is the part I would put in front of a board. The negative effect on perceived sincerity was mitigated by two specific interventions. The first is a timely availability guarantee. The second is external attribution framing, meaning the shortage is explained by a cause outside the retailer's control rather than left for the customer to interpret.

If you look at the two mitigations together, you will see that they outline what a proper back-in-stock notification should be like. A committed statement regarding availability constitutes a guarantee, while a sincere explanation of why the stock sold out is an example of external attribution. The purpose of the notification is not to be a mere courtesy email but rather to act as the means by which a broken promise is turned into one that is kept, and the research shows that the difference lies in purchase likelihood and not just in satisfaction scores.

There is another mechanism that deserves to be mentioned since it accounts for the behaviour you observe in session recordings. According to Brehm, psychological reactance refers to the reaction that occurs when a person's freedom of action is taken away—they then act in order to regain it. A product page that states the item is out of stock, provides no options, and leaves the customer with no way forward is an example of this. Here, the restoration of freedom comes about by allowing the customer to open a competitor's website in a new tab. Simply offering the customer the option to wait, with a specific date given, is not just a more pleasant option; it maintains the customer's sense of agency, and it is this sense of agency that prevents the reactance response.

Building the notification so it does the work

The version that retailers send is a grey button with the message 'Notify Me', which collects email addresses and displays a general message when stock becomes available. Although it does capture some customer demand and is better than offering nothing, it still leaves most of the potential value in the box.

A date is more valuable than a status. "Back in stock" informs the customer only that an event has taken place, something which they can see for themselves. "Expected 4 September", on the other hand, is a commitment and commitments can be counted on, planned around and assessed. It is also the version that corresponds to the availability guarantee mentioned in the research. The obvious counter-argument is that supply chains are prone to delays and it would be better not to make a commitment. That objection deserves to be taken seriously, and the solution is to provide a date range together with a level of confidence that you are willing to back, rather than eliminate dates altogether.

It is important to give attribution whenever it is true and only in those cases. For example, if a line of product sells faster than expected because of some coverage, you should state that. If the shipment is delayed at a port, you should mention that too. The thing you must never do is to make up an external cause, since a fabricated attribution is exactly the kind of manipulative signal that the literature on persuasion knowledge identifies, and the cost of being caught will be greater than the cost of the shortage.

On the day you mentioned, send the email to all those who signed up before putting the stock back on general sale. That final point about the sequence is the one where I've noticed most of these programmes failing silently. The stock arrives, it is made available on the site, and then four hours later the notifications are sent out, so the customers who had waited find that the units have gone to people who didn't. That is the original broken promise being repeated with an additional insult attached.

When you cannot avoid substituting

Grocery and consumables have a harder version of this problem, because the stockout can happen after the order is placed. The product was available at checkout and is unavailable at pick. The customer has already committed.

Research published in the Journal of Retailing in 2022, using computer-simulated purchase experiments across more than 3,000 households and five product categories, examined which substitution policies mitigate the negative response. The useful finding is that matching the substitute on the category's dominant attribute increases acceptance, and that the dominant attribute is not the same across categories. For horizontally differentiated categories, such as cereals or crisps, the dominant attribute is flavour. For vertically differentiated categories, such as margarine or ketchup, it is brand. A substitution engine that applies one rule across the catalogue will be right in half the categories and wrong in the other half.

The previous lack of a quantified cost associated with getting the entire sequence wrong is no longer the case. A study published in the Journal of Retailing in 2025, which made use of data from the online operations of a major omnichannel grocery retailer, discovered that order fulfilment failures on average delayed the customer's following order by 7.22%, the delay being more significant in the case of non-perishable products and spending reductions particularly noticeable when promoted items did not ship. The same study found that although substitutions do recover the immediate sale, they often increase the delay and add to customer dissatisfaction. Since substituting is a revenue recovery measure that carries a customer retention cost, both aspects of that trade-off must be included in the model.

What to measure

Three numbers make this a commercial conversation rather than a UX conversation.

The first is the recovered revenue rate on notifications: of the customers who signed up, what proportion bought within a defined window after the email, and what did they spend. That number on its own overstates the value, because some of those customers would have come back anyway, so the comparison you want is against sessions that hit the same out-of-stock page and did not sign up. It is an imperfect control, and it is far better than the alternative of counting all notification revenue as incremental, which is what most reporting does.

The second is repeat purchase rate at 90 days, split by whether the customer's last interaction included an availability failure. This is the number the 2025 fulfilment study was measuring in a different form, and it is the one that tells you whether your stockouts are costing you transactions or costing you customers. Those are very different problems with very different fixes.

The third case is the honest one—how frequently a low stock badge is followed by a real stockout for a customer who has seen it. In this situation, if the rate is high, then the badge is functioning precisely as it was intended and is therefore causing the dissatisfaction mentioned in the 2016 study; but if the rate is very close to zero, it is likely that the badge is not linked to actual inventory and you are using the manipulative version even though you haven't deliberately chosen to do so.

On the evidence

It should be made clear how old two of the sources are: Peinkofer, Esper and Howlett's is from 2016 and Brehm's study on reactance dates from 1966. I have chosen the first one since its finding is specific, counterintuitive and directly relevant, and because the 2025 Journal of Retailing study on the scarce-insincere inference supports rather than opposes its conclusion. I have selected the second one since reactance is a basic construct rather than a disputed empirical claim and because I am not relying on it to provide a magnitude.

The 2022 scarcity meta-analysis, the 2022 substitution study, the 2025 sincerity paper and the 2025 fulfilment failure study are all recent, have all been peer-reviewed, and all indicate the same conclusion, a finding which is more reassuring than any one of the individual results.

What I have deliberately not used is the large body of circulating statistics about what stockouts cost retailers globally. Figures in the hundreds of billions and above appear constantly in vendor content, and I could not trace them to a method or a sample I would be willing to defend. So they are not here. The 7.22% delay figure is in because the study states what it measured and where the data came from.