Enterprise Intelligence Group
Assortment optimization: what to test before buying or changing the software
Evaluate assortment planning software using product mix, demand transfer, NPI, private label, space and data requirements. Includes a practical review template.
Assortment optimization evaluates which products to offer for a category and location, subject to commercial objectives and operating constraints. The useful question is not simply which products sell most. It is what changes when products are added, removed or given a different role in the range.
This guide sets out a practical way to evaluate assortment planning software or review an existing process. It starts with product and category decisions, then connects them to data, space and implementation.
Separate planning, intelligence and optimization
Assortment planning defines the intended range, its structure and the process for maintaining it. Assortment intelligence, as used here, means the evidence and analysis supporting those decisions; it is not the name of an EIG product. Optimization evaluates alternatives against objectives and constraints. These activities can overlap in a software offering, but a buyer should ask which decisions each capability actually supports.
Do not begin a comparison with a feature count. Establish whether the priority is a category review, a new-product introduction, localization, range simplification or implementation of a new operating process.
Follow one range change
Consider an illustrative proposal to introduce a private-label item into a category with limited shelf capacity. The item may attract new demand, replace sales of an existing own-brand item, draw demand from a national brand or influence adjacent products. None of those outcomes is established by the new item’s sales alone.
Ask the evaluator to show the baseline range, the proposed change, the assumed sources of demand and the uncertainty around those assumptions. Inspect what happens when the store has a different shopper mission, a smaller fixture or an availability problem.
PROPOSED RANGE CHANGE
Substitution within category
New demand
Demand leaving the category
Adjacent-product effects
Hypotheses to test, not measured demand flows.
See an illustrative private-label demand-transfer review
Product attributes and consumer decision trees
Product attributes help describe which items shoppers may consider comparable: pack, size, flavour, format, brand or other relevant characteristics. A consumer decision tree can help frame how those attributes relate to choice. It should not become an unquestioned explanation of every shopper or location.
Ask how the attributes and tree were produced, which observations support them and how they are maintained. Expert judgment can be useful, but it should be distinguishable from observed data and model inference. AI-proposed attributes need review before they become trusted inputs to commercial decisions.
New products and cannibalization
A new product has limited or no sales history in the intended setting. The evaluation should make analogues, launch assumptions and constraints visible. Ask what would cause the team to revise its initial view and how actual performance will be reviewed after introduction.
For cannibalization and demand transfer, request an explanation of the proposed relationships, not only an uplift number. Ask whether the analysis considers unavailable items, promotions, price differences and movement outside the category. A model’s prediction is not proof of a causal effect.
Make location and space constraints explicit
An acceptable national range may be impractical in a smaller store. Test how the system represents location differences, store clusters, mandatory products, facings and capacity. Examine both the proposed assortment and the conditions required to execute it.
Clustering should support an actual planning decision. A statistically neat set of stores is not sufficient if its range cannot be implemented or its commercial meaning cannot be explained.
Explore space planning and store clustering support
A range decision is also a space and execution decision
Consider an illustrative private-label introduction. The proposal adds an item to a category whose smaller stores already have limited shelf capacity. The category team expects demand to transfer from existing products, but that expectation has not yet been validated.
There are several decisions here. Category management establishes the proposed product mix and its role. Micro space planning tests how products and facings fit the fixtures. Macro space planning considers the category footprint and its relationship to other parts of the store. Store clustering helps define where a shared plan is appropriate and where an exception is needed.
Record these decisions together. A model’s recommended range is not an executable store plan until the constraints and responsibilities have been addressed.
Illustrative range-to-space decision record
| Decision | Evidence or assumption to record | Required validation |
|---|---|---|
| Why introduce the item? | Category objective, shopper need and proposed role | Category owner agrees the objective before reviewing results |
| Where may demand come from? | Existing private label, national brands, new demand or adjacent products | Keep each proposed source as a hypothesis unless supported by evidence |
| What changes in the range? | Additions, removals, retained mandatory items and alternatives | Evaluate lost demand and substitution, not only the new item’s sales |
| Does the plan fit? | Dimensions, facings, fixture capacity and category footprint | Micro and macro owners resolve infeasible combinations |
| Which stores can use it? | Cluster rationale and known local constraints | Validate exceptions; an average store is not proof for every location |
| Can it be replenished? | Pack sizes, supply constraints and available shelf capacity | Replenishment owner confirms the operating assumptions |
| Was it implemented? | Plan version, effective date and observed store exceptions | Separate planned from implemented locations in the review |
| What happened afterwards? | Category outcomes, availability, price/promotion changes and competing explanations | Review the agreed objective without attributing all changes to the new item |
A nearby product’s decline does not by itself prove cannibalization. Availability, promotion timing and changes in execution may also explain the result. Similarly, a successful planogram publication does not prove the store implemented that version.
EIG can help define a scoped assortment and demand-transfer review, connect the decision to space and store constraints, and establish the evidence required for implementation and post-change evaluation. The result is a decision record with accountable actions, not an unsupported promise of category uplift.
Explore merchandising and assortment consulting
Review micro and macro space-planning requirements
Use the same evidence test for each option
Assortment evaluation worksheet
For each row record the evidence reference, unresolved question, accountable owner and next validation step. Mark “not assessed” where evidence is absent. Do not turn missing information into a positive score.
| Decision area | Test to perform |
|---|---|
| Category objective | Explain the objective and how conflicting goals are resolved |
| Product mix | Compare the current range with a proposed addition or removal |
| Attributes and choice | Show attribute provenance, decision-tree assumptions and exceptions |
| New-product introduction | Identify analogues, uncertainty and the post-launch review method |
| Demand transfer | Explain potential substitution, lost demand and adjacent-product effects |
| Location and space | Demonstrate a constrained store or cluster, not only an average location |
| Data readiness | Identify required history, hierarchies, availability and missing inputs |
| Human decisions | Show approval, override reasons and a retained decision record |
| Delivery | Demonstrate how an approved range reaches downstream planning and execution |
| Value review | Define baseline, observation period, operating measures and interpretation limits |
Use one permitted sample and consistent scenarios across vendors. Treat mandatory constraints separately from scored preferences. A polished demonstration does not remove an unresolved data or execution dependency.
What implementation needs to carry forward
The buying decision should leave a usable record of requirements, assumptions, data ownership, acceptance tests and operational responsibilities. Those become inputs to implementation, not documents abandoned after selection.
Plan how category teams will use the outputs, who will resolve exceptions and when the first range review will take place. Software capability and organizational adoption need separate acceptance evidence.
Apply the review to your category decision
EIG supports assortment requirements, product analysis, data readiness and implementation. A scoped engagement can examine a defined range decision or software-evaluation process and produce a practical set of findings, questions and next actions.
Discuss merchandising and assortment support
If your immediate need is a vendor shortlist or RFP, explore technology selection support. If a product-data issue is blocking the decision, explore retail data and integration.
Use the planning-data readiness checklist
Discuss your requirements or vendor selection
The EIG Library · Insights · AI Maturity Assessment · Contact