High-SKU ecommerce creates a paradox. The catalogue becomes more valuable because it can satisfy more particular needs, yet it becomes progressively harder for a marketing team to decide which products deserve attention. The usual response is to build more segments, more automated flows and more campaigns. That increases activity, but it does not solve the underlying decision.
The real question is not which campaign should go to a group. It is which product is most pertinent to one identifiable customer now, whether the evidence is strong enough to act, and when that decision should be reconsidered.
For a retailer with thousands, tens of thousands or millions of eligible products, that distinction changes the operating model. Human campaign construction can remain useful for launches, editorial moments and genuine brand events. It is no longer an adequate method for continuously matching the breadth of the active catalogue to the changing interests of every customer.
Catalogue scale changes the problem
A small catalogue can be understood through a modest number of commercial stories. A team may know its hero products, seasonal priorities, replenishment cycles and obvious complements. As the number of products grows, the number of plausible customer-product relationships grows much faster than the headcount available to interpret them.
More choice is not automatically more useful choice. Research into online shopping found that larger online choice sets can increase choice difficulty and deferral, while the desirability of the assortment and the shopper’s vigilance alter the effect.[1] A broader high-SKU ecommerce catalogue therefore creates value only when the retailer can help each person encounter the part that is relevant to them.
The wrong response is to shrink every customer’s world to the same bestseller list. A large assortment exists because different people want different things. Field evidence from an online platform found that assortment expansion can attract new customers while reducing consumption frequency among customers who remain, and that targeted reductions in the choice set can improve revenue for existing customers.[4] The commercial opportunity is therefore to make the catalogue feel smaller to the individual without making it smaller in reality.
That is a selection problem. It cannot be completed merely by writing another email, adding another branch to a workflow or dividing the database into a few more groups.
A segment describes a group, not a person’s next decision
Segments are useful for reporting, eligibility, governance and broad commercial planning. They can identify shared characteristics: recent purchasers, high-value customers, footwear buyers, lapsed customers or people who entered through a particular category.
The limitation appears when a descriptive group is treated as the final product decision. Two people can belong to the same segment while differing in their recent sequence of views, depth of interest, purchase intervals, exclusions, price sensitivity and changing relationship with the catalogue. Their shared label does not establish that they should receive the same products at the same time.
Segmentation also freezes a boundary for operational convenience. A person is placed into a group because some evidence crossed a rule or clustering threshold. Their next behaviour may immediately make that boundary less useful. If the campaign remains attached to the old classification, the retailer is acting on a summary of the past rather than a current individual decision.
This is why personalisation and individualisation are materially different. A segment can remain an input or a safeguard. It should not be mistaken for the person.
Flows automate a route that somebody has already designed
Automated flows are valuable when the route is genuinely known. Order confirmations, delivery information, consent records and other deterministic communications should be reliable and repeatable. A flow is also suitable when a retailer has consciously chosen a fixed commercial treatment for a defined event.
But a flow does not become autonomous because it contains many branches. Each branch still represents an answer anticipated in advance: if this happens, send that; if the customer belongs here, choose this content; if a product was abandoned, repeat it or select from a predefined set.
As the catalogue and customer base expand, the decision tree becomes an accumulating record of yesterday’s assumptions. Teams must add conditions, resolve collisions, maintain exclusions, refresh creative and decide which flow has priority. Complexity moves into the workflow rather than disappearing.
The central limitation remains. A human has narrowed the available answer before the individual’s current evidence is evaluated. The software executes the chosen route efficiently, but it does not independently determine the best product across the eligible catalogue and revisit that decision as circumstances change.
The distinction between repeating a process and revising a decision is explained further in how autonomous systems differ from automated systems.
Campaign production creates a high-SKU ecommerce bottleneck
Campaigns impose an editorial hierarchy on the assortment. Merchandisers and marketers select a manageable collection, create a message and distribute it to an audience. That method is necessary for genuine announcements and brand narratives, but in high-SKU ecommerce it is structurally biased towards products that a team has time and confidence to choose.
Hero products, new ranges and proven sellers repeatedly receive attention. Thousands of legitimate products remain commercially quiet, not because customers would reject them, but because no campaign slot was available and no team could construct a separate case for every possible person-product pairing.
Research using a randomised field experiment across 82,290 products and more than one million users found that recommender design can materially affect sales diversity. Traditional collaborative filters helped individuals explore products, yet aggregate purchasing could still become more concentrated because similar users were directed towards similar kinds of items.[2] The lesson is not that recommendation is undesirable. It is that catalogue coverage and individual relevance must be observed rather than assumed.
High-SKU retailers should therefore ask more demanding questions:
- What proportion of the active catalogue receives meaningful consideration?
- How much selection is predetermined by campaigns or collections before individual evidence is evaluated?
- Do different customers repeatedly converge on the same popular products?
- Can a product outside the bestseller group earn exposure when it is the strongest candidate for a particular person?
- Does the decision change when that person views, ignores, buys, returns or develops a different interest?
These questions examine the selection system rather than the volume of marketing output.
The alternative is not indiscriminate automation
Moving beyond campaigns does not mean removing commercial control. A high-SKU ecommerce retailer must still govern stock, price, legal eligibility, consent, brand presentation, product exclusions and operational constraints. Autonomy should operate inside those boundaries, not override them.
The change concerns who makes the continuing product-and-timing decision. Instead of a marketer repeatedly choosing a collection and building an audience around it, software evaluates the eligible catalogue for each identifiable customer, ranks plausible next purchases and recalculates as new evidence arrives.
That decision should remain probabilistic. The leading product is not a declaration of certainty; it is the strongest available candidate at that moment. A person can surprise the model. New behaviour should be allowed to change the answer rather than being forced back into an old journey.
Research on choice overload similarly cautions against universal rules. A meta-analysis found that choice-set complexity, decision difficulty, preference uncertainty and the consumer’s goal all influence when assortment size becomes burdensome.[3] Individual context matters. The same catalogue can be liberating for one person and exhausting for another.
High-SKU economics reward individual relevance
The purpose of a large high-SKU ecommerce catalogue is not to make every product equally visible. It is to make a much wider range of needs commercially serviceable. That creates three connected opportunities.
First, the retailer can reduce dependence on the narrow group of products that always win campaign space. Relevant long-tail products can be surfaced when individual evidence supports them, rather than being promoted indiscriminately.
Second, relevance can reduce the impulse to make price the primary reason to act. Blanket discounts compensate for weak selection by making a broad offer more attractive to everybody. A better individual decision creates a route to full-price consideration without claiming that discounts will never be appropriate. The commercial discipline is described in how autonomous product selection protects gross margin.
Third, the retailer can separate activity from causation. More sends, more attributed orders and more elaborate journeys do not prove incremental value. A stable holdout can test whether autonomous individual decisions create additional net revenue or gross profit compared with the counterfactual. The method is set out in measuring incremental ecommerce revenue.
These gains must be measured together. Broader catalogue exposure is not valuable if returns increase sharply. A higher conversion rate may be misleading if the same customers would have purchased anyway. More full-price orders may still shift demand from another channel. The correct evaluation combines individual relevance, catalogue coverage and controlled commercial outcomes.
What autonomous individualisation changes
SwiftERM is an autonomous individualisation platform for ecommerce. It analyses first-party behavioural, product and transaction evidence to calculate what each identifiable customer is most likely to buy next and when. It then supplies that individual product selection through the retailer’s established environment.
The retailer’s ecommerce platform, email infrastructure, catalogue governance and brand controls continue doing the jobs for which they were selected. In high-SKU ecommerce, SwiftERM changes the decision layer: product and timing are recalculated for the individual rather than repeatedly fixed by segment, flow or campaign.
That distinction is especially important in high-SKU ecommerce. The greater the catalogue breadth, the less plausible it becomes that a team can manually anticipate every commercially useful relationship. Adding people, rules and campaigns may increase throughput, but it cannot make manual selection continuous at individual scale.
A practical test for high-SKU retailers
A high-SKU ecommerce retailer can test whether its current operation has outgrown campaign logic without replacing anything immediately. Take a representative period and record:
- the percentage of active products that received meaningful customer exposure;
- the percentage of selections predetermined by a campaign, segment or fixed flow;
- the concentration of exposure and sales among the leading products;
- the staff and agency time required to maintain journeys and content variants;
- the delay between a meaningful behavioural change and a changed product decision;
- the incremental net revenue or gross profit demonstrated against a valid holdout.
The result reveals whether the operation is using catalogue breadth or merely carrying it. A high-SKU retailer should not have to choose between overwhelming customers with variety and hiding most of the assortment behind a handful of campaigns. The more valuable route is to preserve the breadth, govern it carefully and make it individually navigable.
Segments can still describe. Flows can still execute known processes. Campaigns can still express real editorial intent. None of them should be required to make every customer-product decision.
High-SKU ecommerce has outgrown the idea that more marketing machinery is the answer. The next stage is a continuously revised individual decision—made across the eligible catalogue, constrained by the retailer and judged by incremental commercial evidence.
high-SKU ecommerce: the current operational framework
high-SKU ecommerce can now be considered through SwiftERM’s autonomous individualisation. Product relevance is recalculated for each identifiable customer from that customer’s own behavioural and transactional evidence, rather than being assigned by a fixed segment, campaign or predetermined journey. This distinction matters because an individual’s interests can change between visits, purchases and periods of inactivity. A current decision therefore needs current evidence.
SwiftERM evaluates the products that are becoming most relevant to each person and applies that decision autonomously. The operating model does not depend on a marketer continually creating segments, selecting products or scheduling another campaign. It is designed to use first-party behavioural and transactional signals already associated with an identifiable customer, so the decision remains specific to that customer rather than a broad cohort.
For high-SKU ecommerce, the practical test is not whether a retailer can create more rules. It is whether the system can recognise a change in an individual’s buying intention, reassess the available products and act without waiting for a manual campaign cycle. The output must remain traceable to the customer evidence and the ecommerce catalogue used to make the decision.
Implementation also needs disciplined data handling. Product identifiers, availability and customer events must remain consistent enough for relevance to be recalculated. Existing ecommerce operations should retain control of catalogue truth, order truth and customer permissions. Autonomous individualisation then operates within those established boundaries instead of inventing a parallel customer record or an ungoverned product feed.
The commercial value of high-SKU ecommerce should be assessed through observed customer response, orders and turnover, not through the volume of campaigns produced. That keeps attention on whether product selection becomes more useful to the individual and whether the retailer captures demand that would otherwise remain unrecognised. It also separates autonomous decision-making from conventional automation that simply executes a previously written rule faster.
This current framework preserves the original article’s subject while bringing its explanation into line with SwiftERM’s present capability: one identifiable customer, individually recalculated product relevance, and an autonomous decision made from that person’s own changing evidence. The same principle applies as catalogue choice, customer behaviour and commercial conditions continue to change over time.
In operational terms, the retailer retains the commercial context while the software performs the individual calculation. That division is important: the catalogue and customer record remain governed business assets, while product selection responds continuously to evidence rather than to a static assumption made when a campaign was first configured.
Relevant independent context is available in “Choice Frictions in Large Assortments” from knowledge.uchicago.edu.
Relevant independent context is available in purchases made as a result of receiving a marketing message, email has the highest conversion rate (66%).
Relevant independent context is available in Email still developers over 70% more sales than every other comparison.
References and accreditations
- Wang, F., Wang, M., Zheng, Y., Jin, J. and Pan, Y. (2021), “Consumer Vigilance and Choice Overload in Online Shopping”, International Journal of Electronic Commerce, 25(3), 364–390.
- Lee, D. and Hosanagar, K. (2019), “How Do Recommender Systems Affect Sales Diversity? A Cross-Category Investigation via Randomized Field Experiment”, Information Systems Research, 30(1), 239–259.
- Chernev, A., Böckenholt, U. and Goodman, J. (2015), “Choice Overload: A Conceptual Review and Meta-Analysis”, Journal of Consumer Psychology, 25(2), 333–358.
- Natan, O. R. (2021), “Choice Frictions in Large Assortments”, PhD dissertation, University of Chicago, doi:10.6082/uchicago.2977.