The subscription-commerce model achieved, during the period from 2020 through 2022, a rate of category penetration that the trade literature of the time treated as the early stage of a permanent realignment of retail commerce. The literature’s expectation was that subscription would, over the subsequent five-year horizon, become the dominant fulfilment pattern for a substantial fraction of categories that had previously operated on the discrete-purchase model; the categories named most frequently included household consumables, beauty and personal care, certain food categories, certain apparel-replacement categories such as socks and underwear, and a long tail of speciality categories from coffee to vitamin supplements to pet food.
The expectation has, in the period since, been substantially disconfirmed by the retention-curve data the subscription operators have, with varying degrees of disclosure, made available. The pattern that emerges from the data is approximately the following: the categories in which subscription was viable as a permanent business model are a smaller subset of the categories the literature predicted, the rate of churn the unsuited categories produce makes the model unsustainable for those categories regardless of acquisition spend, and the retailers who built operations on the assumption of broader applicability are, in 2026, in various stages of unwinding the subscription components of their business in favour of the discrete-purchase model the subscription was intended to replace.
This post is an account of the retention-curve patterns the data exhibits, the categories in which the model continues to function, and what the implication is for retailers contemplating subscription as a strategic option in 2026.
The retention-curve patterns
The subscription-commerce retention curves, on the data I have collected from clients with subscription operations, exhibit three broadly distinguishable patterns.
The first pattern, observed in categories where subscription has proven durable, is the asymptotic-retention curve. The customer base churns at a relatively high rate during the first three to six months of the subscription, with the rate declining as the subscription matures; by approximately the twelve-month mark, the retention curve has stabilised at a sustainable level (typically in the range of forty to sixty per cent of the original cohort), and the surviving customers exhibit retention behaviour that is broadly stable on subsequent monthly intervals. The categories that produce this pattern are, on the data, those whose products satisfy a recurrent need that the customer would, in the absence of the subscription, fulfil through a series of discrete purchases anyway — household consumables that the customer reorders monthly, pet food, coffee, certain vitamin supplements, certain personal-care products. The subscription is, for these categories, a convenience improvement over the discrete-purchase pattern rather than a fundamental change in the customer’s relationship with the product.
The second pattern, observed in categories where subscription has not proven durable, is the continuing-decline curve. The customer base churns at a high rate throughout the subscription’s lifecycle, with the rate not stabilising at any sustainable level. By the twelve-month mark, the retention is, on most of the deployments I have observed, below twenty per cent of the original cohort; by twenty-four months, retention is in single digits. The categories that produce this pattern are, on the data, those whose products do not satisfy a recurrent need that the customer would otherwise fulfil through repeated purchases — speciality apparel, novelty subscription boxes, curated-selection services in categories that are not, in the customer’s existing behaviour, recurrent-purchase categories. The subscription is, for these categories, an artificial structure imposed on a non-recurrent purchase context, and the customer’s churn behaviour reflects the structural mismatch.
The third pattern, observed in a smaller set of categories, is the seasonal-stabilisation curve. The customer base churns through the first year, exhibits a partial recovery during a category-specific seasonal window, and stabilises at a level that is sustainable but considerably lower than the first-pattern asymptotic level. The categories include certain beauty and skincare subscriptions whose customer cohort responds to seasonal product variation, certain food-category subscriptions where the seasonal cycle of the product range supports a level of cohort renewal that the steady-state retention does not. The subscriptions in this pattern are operationally viable but require sophisticated lifecycle management to sustain; the operators who have invested in the lifecycle management are sustaining the operations, while those who have not are unwinding them.
The categories that survived
The subscription-commerce operations that remain commercially viable in 2026 are, in approximate order of stability, the following.
The household-consumable category — laundry detergent, dishwasher tablets, paper goods, basic cleaning supplies — is, on the retention data, the most stable subscription category and the category where the model approximates the original literature’s prediction. The customer’s purchase behaviour in this category is genuinely recurrent and the subscription’s convenience benefit is correspondingly real; the retention curves stabilise at sustainable levels, and the operations have, with very few exceptions, sustained the growth they achieved during the pandemic period.
The pet-food category is similarly stable, for the same underlying reason: the customer’s purchase behaviour is recurrent, the subscription is a convenience improvement over the discrete-purchase pattern, and the retention curves reflect the stability of the underlying behaviour. The category has, on the public-disclosure data of the major operators, sustained both the retention rates and the absolute customer-base size from the pandemic period.
The coffee category occupies a position between stability and decline. The retention curves are, on average, slightly less stable than household consumables, with the additional sensitivity that the category is more competitive and the retention improvements from convenience are partially offset by the customer’s exposure to alternative coffee suppliers via the broader retail landscape. The category remains commercially viable but with margins that are, in 2026, considerably tighter than the pandemic-period peak suggested.
The vitamin-supplement category has bifurcated. The subscription operations whose product offering corresponds to medically-recommended ongoing supplementation have sustained their retention; the operations whose offering corresponds to non-recommended supplementation (which is, in the broader category, a substantial fraction) have followed the continuing-decline pattern, with the customer’s churn correlating with the customer’s eventual recognition that the supplementation is not producing the outcomes the marketing claimed.
The categories that did not survive
The subscription operations that have, in the period since the pandemic, substantially regressed include the curated-selection apparel subscriptions, the speciality novelty-box subscriptions, and the broader range of subscription operations whose product range did not correspond to a recurrent purchase need. The pattern of regression is, on the public-disclosure data of the affected operators, substantially as the continuing-decline curve predicted; the operations have, in essentially every case, either pivoted to a discrete-purchase model, accepted a substantially smaller customer base than the pandemic-period peak, or discontinued operations entirely.
The specific operations affected are well-documented in the trade-publication coverage of the period; I will not enumerate them here. The pattern is, in summary, that the operations whose initial growth was driven by pandemic-era convenience-substitution behaviour failed to sustain the growth once the pandemic-era constraints relaxed and the customer’s behavioural alternatives reasserted themselves.
The implication for retailers in 2026
The implication for retailers contemplating subscription as a strategic option is, in summary, that the appropriate question is not whether the subscription model is generally viable but whether the specific category and product range under consideration satisfies the recurrent-purchase property that the durable subscription operations all satisfy. Categories that do satisfy the property remain commercially attractive subscription opportunities; categories that do not are, on the data, not commercially viable for the model regardless of the marketing investment available to support the launch.
The recommendation that has emerged from the analysis is that retailers contemplating a subscription launch first examine the existing retail behaviour of the target customer in the product category — specifically, whether the customer is, in the discrete-purchase pattern, exhibiting recurrent-purchase behaviour against the same brand. If the customer is, the subscription is offering a convenience improvement to behaviour the customer is already engaged in, and the retention curves can be expected to follow the asymptotic pattern. If the customer is not, the subscription is attempting to construct a behavioural change in the customer’s relationship with the category, and the retention curves can be expected to follow the continuing-decline pattern that the unsuited categories exhibit.
An advisory close
The subscription-commerce model is, in 2026, a viable but considerably narrower commercial option than the trade literature of 2021 implied. The retailers who have sustained their subscription operations through the post-pandemic period are those whose categories satisfy the underlying recurrent-purchase property; the retailers who have unwound their subscription operations are those whose categories did not. The procurement decisions for new subscription launches in 2026 should, accordingly, be considerably more selective than the broader literature continues to suggest; the data on which categories work, and on the magnitude of the difference between working and not-working, is by now sufficiently mature that the selection can be made with reasonable confidence.
It is recommended that retailers contemplating subscription launches do so with explicit reference to the existing retail behaviour in the category, and with explicit recognition that the subscription model is appropriate for some categories and is not appropriate for others. The categories where it is not appropriate are, on the data, the categories where the procurement investment is most likely to produce operations that do not, in the post-launch reality, sustain themselves.