Category Management

Breadth × Depth × Role

The strategic tier of retail merchandising. Before deciding which SKUs go on which shelf, a retailer chooses how many sub-categories to carry per category (breadth), how many tiers within each sub-category (depth), the private-label share, and the category role — traffic-driver, profit-generator, image-builder, or convenience. Foundational refs: Dhar, Hoch & Kumar (2001); Ailawadi & Harlam (2004); Broniarczyk, Hoyer & McAlister (1998).

Why it matters

Industry adoption · documented margin lift · structural share

~80%
Share of grocery retailers in mature markets that use a formal category-management framework, typically based on the ECR 8-step model.
Source: ECR Europe / IGD industry surveys.
+5–15%
Documented category-margin lift from data-driven category management vs. ad-hoc SKU-level decisions in US grocery panels.
Source: Ailawadi & Harlam (2004), Marketing Science 23(1).
~25%
Share of US grocery dollar sales attributable to private-label SKUs — a strategic share-of-shelf decision made at the category level.
Source: PLMA (2023) Private Label Yearbook.
role-aware
Assigning each category an explicit role (traffic-driver vs profit-generator etc.) raises both store loyalty and category profitability vs. uniform treatment.
Source: Dhar, Hoch & Kumar (2001), J Retailing 77(2).

Where the decision sits

Strategic tier · sets the frame for shelf and assortment

Category management is the strategic tier of merchandising. It defines what each category is for — does “Wine” pull weekly traffic, signal a premium image, or quietly generate margin? — before anyone counts SKUs or measures shelf-feet. The category-role decision propagates downward: a traffic-driver justifies wider breadth and deeper national-brand depth; a profit-generator tolerates narrower breadth with higher private-label share. Operational tools (shelf-space allocation, assortment planning) execute against the category structure set here.

Define category rolestraffic / profit / image / convenience
Set breadth/depth\(b_c\), \(d_{c,k}\), \(pl_c\)
Allocate shelfshelf-space
Executeassortment planning

Problem & formulation

Hierarchical: strategic category structure feeds operational shelf and assortment

OR family
Hierarchical MINLP
Decision tier
Strategic (annual / semi-annual)
Solver realism
★★ Educational
Reference
Dhar-Hoch-Kumar (2001)

Sets and indices

SymbolMeaningDomain
\(c \in C\)Category (e.g., Grocery Staples, Wine)discrete
\(k \in K_c\)Sub-category within category \(c\)discrete
\(i \in I_{c,k}\)SKU within sub-category \(k\) of category \(c\)discrete

Parameters

SymbolMeaningUnit
\(w_c\)Category role weight (traffic-driver = high, convenience = low)\(\in [0,1]\)
\(m_c\)Average per-unit margin in category \(c\) (national brands)$ / unit
\(T_c\)Per-category traffic (visits attributable to \(c\))visits / period
\(S\)Total available shelf space across all categoriesunits
\(B_{\max}\)Maximum allowed breadth per categorysub-cats
\(pl_{\max}\)Maximum allowed private-label share\(\in [0,1]\)

Decision variables

SymbolMeaningDomain
\(b_c\)Breadth: number of sub-categories carried in \(c\)\(\in \{1, \ldots, B_{\max}\}\)
\(d_{c,k}\)Depth: number of SKU tiers per sub-category\(\in \{1, \ldots, D_{\max}\}\)
\(pl_c\)Private-label share within category \(c\)\(\in [0, pl_{\max}]\)

Sales response (per-SKU MNL)

Within a category, per-SKU sales follow a Multinomial Logit choice model — the same demand primitive used in assortment optimisation. At the category level we collapse this into a concave bilinear lift on breadth and depth:

$$\text{Sales}_c \;=\; w_c \cdot T_c \cdot f(b_c) \cdot g(d_c) \cdot \bigl(1 + h(pl_c)\bigr), \quad f, g \text{ concave}, \; h \text{ small bonus}$$

\(f(b) = \log(1+b)\), \(g(d) = \log(1+d)\) capture diminishing returns; \(h(pl) = \beta \cdot pl\) gives a modest margin bonus from PL.

Objective

$$\max_{b, d, pl} \;\; \sum_{c \in C} m_c \cdot T_c \cdot f(b_c) \cdot g(d_c) \cdot \bigl(1 + h(pl_c)\bigr) \;-\; \lambda_S \cdot \text{shelf}(b, d) \;-\; \lambda_R \cdot \text{plRisk}(pl)$$

Margin contribution net of shelf-space cost and a private-label brand-risk penalty (high PL share weakens national-brand co-op funding and customer image of the chain).

Constraints

$$\sum_{c} b_c \cdot d_c \;\leq\; S \qquad b_c \;\leq\; B_{\max} \qquad pl_c \;\in\; [0, pl_{\max}]$$

Total shelf footprint capped; per-category breadth and PL-share within strategic limits.

Why it’s hierarchical

Category-level decisions \((b_c, d_{c,k}, pl_c)\) are made before the operational models run. Once the category structure is set, downstream models inherit it: the shelf-space allocation problem distributes \(S\) across categories using \(b_c, d_c\) as inputs; the assortment optimisation picks specific SKUs within the depth budget \(d_{c,k}\) honouring the PL-share \(pl_c\). Re-optimising the strategic tier weekly would destroy planogram stability — CM cycles are typically annual or semi-annual.

Interactive solver

Strategic CM allocator over 4 categories · concave breadth/depth response

CM strategic allocator
4 categories · breadth (1-8) · depth (1-6) · PL share (0-50%)
★★ Educational
Sum of \(b_c \cdot d_c\)
$ per shelf-unit
Penalty on \(pl_c^2\)
\(h(pl) = \beta \cdot pl\)
—
Total margin ($)
—
Shelf used / budget
—
Private-label revenue %
—
Total SKUs
Margin contribution (left bars) Breadth \(b_c\) (right stack) Depth \(d_c\) (right stack) Private-label share \(pl_c\) (right stack)

Under the hood

For each category we scan the discrete grid \((b_c, d_c) \in \{1, \ldots, 8\} \times \{1, \ldots, 6\}\) and a coarse PL grid \(pl_c \in \{0, 0.1, 0.2, 0.3, 0.4, 0.5\} \cap [0, pl_{\max}]\), score each triple by the per-category contribution \(m_c \cdot T_c \cdot f(b_c) \cdot g(d_c) \cdot (1 + \beta \cdot pl_c) - \lambda_R \cdot pl_c^2\), then select the highest-scoring per-category triples whose total \(\sum b_c \cdot d_c\) fits the shelf budget via a simple greedy — sort categories by marginal-margin-per-shelf-unit and add until full. ~288 evaluations per category × 4 categories = sub-millisecond in the browser. This is an educational approximation; production CM tools combine MNL demand fits with joint shelf-space MIPs.

Reading the solution

What a category manager actually does with the output

Three patterns to watch for

  • Role weight dominates breadth. Categories with high \(w_c\) (traffic-drivers like Grocery Staples) earn wider breadth: the diminishing-returns curve \(f(b)\) is still steep at moderate \(b\) when \(w_c \cdot T_c\) is large. Convenience categories collapse to \(b = 1, d = 1\).
  • PL share is bounded by risk, not opportunity. Without the \(\lambda_R\) penalty, the optimiser pushes every category to \(pl_{\max}\). With the penalty, PL share clusters around the role-weighted optimum — deeper PL in convenience and image-builders, lighter in traffic-drivers (where national brands draw shoppers in).
  • Shelf reallocation, not blanket cuts. Tightening \(S\) does not shrink all categories proportionally — the greedy reallocator drops depth in low-margin/low-traffic categories first, preserving the high-role categories. This matches Dhar-Hoch-Kumar’s empirical finding.

Sensitivity questions the model answers instantly

  • Lift the role weight on Wine from 0.4 to 0.7? — breadth shifts toward Wine; Snacks gives up depth.
  • Drop \(\lambda_R\) (corporate accepts higher PL risk)? — PL shares rise across categories; total margin lifts modestly.
  • Cut \(S\) by 25%? — convenience categories go to minimum; traffic-drivers retain breadth at the expense of depth.

Model extensions

From the strategic baseline to the wider CM landscape

Hierarchical CM (3 tiers)

Strategic role → tactical breadth/depth → operational planogram. The full Dhar-Hoch-Kumar pyramid; this page covers the top two tiers.

Private-label sourcing

Joint sourcing/branding decision: in-house manufacture vs co-pack vs national-brand-equivalent. Adds supplier-side cost terms.

Category-role transitions

Roles drift over time — Wine in 2010 vs 2025. Multi-period model with role-transition costs (planogram redesign, supplier renegotiation).

Omnichannel category roles

The role of “Grocery Staples” differs in-store vs e-commerce. Channel-specific \(w_c\) and \(T_c\) with a shared inventory pool.

Joint with assortment

Solve breadth/depth/PL and SKU selection simultaneously via a large MNL+MIP. Computationally heavier but captures cross-tier interaction.

Assortment →
Joint with shelf-space

Plug the strategic \((b_c, d_c)\) directly into the shelf-allocation MIP rather than passing through aggregated category footprints.

Shelf →
Store clustering for localised CM

One CM plan per store cluster (urban small-format vs suburban big-box), driven by demographic and basket-mix similarity.

AI-driven CM

Replace the hand-coded \(f, g, h\) lift functions with ML demand models (gradient boosting, deep MNL) fit on POS panels. The optimisation skeleton is unchanged.

Key references

Foundational category-management and private-label literature

Dhar, S. K., Hoch, S. J. & Kumar, N. (2001).
Effective category management depends on the role of the category.
Journal of Retailing 77(2): 165–184. doi:10.1016/S0022-4359(01)00040-7
Ailawadi, K. L. & Harlam, B. (2004).
An empirical analysis of the determinants of retail margins: The role of store-brand share.
Marketing Science 23(1): 147–166. doi:10.1287/mksc.1030.0035
Broniarczyk, S. M., Hoyer, W. D. & McAlister, L. (1998).
Consumers’ perceptions of the assortment offered in a grocery category: The impact of item reduction.
Journal of Marketing Research 35(2): 166–176. doi:10.1177/002224379803500204
Hübner, A. H. & Kuhn, H. (2012).
Retail category management: State-of-the-art review of quantitative research and software applications in assortment and shelf-space management.
Omega 40(2): 199–209. doi:10.1016/j.omega.2011.05.008
Kök, A. G., Fisher, M. L. & Vaidyanathan, R. (2014).
Assortment planning: Review of literature and industry practice.
In: Retail Supply Chain Management, 2nd ed., Springer, ch. 6.
ECR Europe (1997).
Category Management Best Practices Report.
Efficient Consumer Response Europe, Brussels.
Pauwels, K. & Srinivasan, S. (2009).
Pricing of national brands versus store brands: Market power components, findings and research opportunities.
Marketing Science 28(6): 1066–1078. doi:10.1287/mksc.1080.0432
Geyskens, I., Gielens, K. & Gijsbrechts, E. (2010).
Proliferating private-label portfolios: How introducing economy and premium private labels influences brand choice.
Journal of Marketing Research 47(5): 791–807. doi:10.1509/jmkr.47.5.791

Back to the retail domain

Category management sits at the strategic top of the merchandising stack — the category-role choice that frames every shelf and assortment decision below it.

Open Retail Landing
Educational solver · concave bilinear lift and quadratic PL-risk penalty · production CM combines fitted MNL demand with joint shelf-space MIPs and store-cluster heterogeneity.