Which SKUs should we keep, add or remove?
Compare contribution, demand and strategic role instead of using raw sales rank as the decision rule.
Assortment Optimisation · Retail & CPG Decision Intelligence
Which products should each store, customer or channel actually carry?
Use demand, substitution, cannibalisation, margin and operational constraints to recommend the assortment that creates the most value instead of ranking products on sales alone.
Questions this model answers
Turn the range into a decision problem.
What happens when a SKU is removed?
Estimate whether demand transfers to another product, disappears, or is at risk of leaving the portfolio.
Should the range differ by store, channel or region?
Use local demand and operating context to avoid forcing one assortment across materially different locations or customers.
Where is range complexity costing more than it contributes?
Balance incremental value against shelf, inventory, service and operational complexity.
The substitution problem
A low-selling SKU can still be strategically valuable.
SKU A sells little, so removing it looks obvious. But if its buyers switch to a competitor rather than another product in your portfolio, the decision can destroy more value than the SKU's sales suggest.
That is why the model needs substitution, cannibalisation and constraints, not only product rankings.
Decision logic
Demand + product economics + substitution + constraints
The output is a feasible assortment recommendation with an expected commercial consequence, not a popularity list.
From data to recommended range
Optimise the assortment under the rules the business actually has.
Demand and product behaviour
Transactions, products, stores or customers, availability, price, promotions, margin and product attributes.
Substitution and cannibalisation
Estimate how demand can move across products when the assortment changes.
Commercial and operational rules
Shelf space, minimum range, strategic brands, category coverage, availability, pack requirements and other constraints.
Keep, add or remove
Return the recommended range by store, customer, channel or region together with expected impact and scenario evidence.
Where this sits in the decision stack
Forecasting estimates demand. Optimisation chooses the range.
A demand forecast can be an input to assortment decisions, but it does not itself decide what should be carried. See forecasting vs optimisation.
For retailers, assortment optimisation is one of the recurring decisions in the GN Retail Decision Pack. For manufacturers and brands, it can also support customer, channel and portfolio decisions in the GN CPG & Beverages Decision Pack.
Define the economic outcome
Revenue, margin, availability, category coverage or another measurable objective.
Encode what cannot be violated
Space, range, service, strategic and operational rules.
Return a feasible range
Recommend the assortment and expose the expected trade-offs.
Direct answers
Questions buyers ask about assortment optimisation.
What is assortment optimisation?
Assortment optimisation decides which products should be carried in a store, channel, customer group or market by combining demand, product economics, substitution and operating constraints.
Why not simply remove the lowest-selling SKUs?
Because low sales do not reveal what happens after removal. Some demand transfers to another product in the portfolio, while some can disappear or move to a competitor. The substitution response is part of the decision.
Can assortments differ by store, customer or region?
Yes. The model can work at the level where local demand, economics and operational constraints are available and reliable enough to support a different recommendation.
Can we include commercial rules and strategic products?
Yes. Shelf space, minimum range, strategic brands, category coverage, availability, minimum margin, pack requirements and other business rules can be encoded as constraints.
What does Graphite Note return?
Typical outputs include keep, add and remove recommendations, expected revenue or margin impact, substitution effects, scenario comparisons and the constraints that shaped the recommendation.
Decision IntelligenceFrom signal and prediction to a governed business action.AutoMLAutomate repetitive model-building work without losing business context.Causal Machine LearningSeparate drivers from correlations and estimate intervention effects.Predictive AnalyticsForecast demand, score customers and anticipate business outcomes.Decision OptimisationCompare feasible actions and rank them by expected business impact.Forecasting vs OptimisationSee where prediction ends and the business decision begins.Decision APIDeliver recommendations, expected impact and evidence into workflows.Embedded Decision IntelligenceAdd a decision layer to ERP, CRM, BI, workflow and vertical software.Build vs BuyA practical framework for deciding whether to build, buy or embed.
Bring the range decision and the constraints.
We will structure the optimisation around the products, locations, customers and economics that determine whether the recommendation can actually be executed.