Seven Ways to Drive Size Optimization Precision
Over the last three years, fashion retailers heavily invested in tightening assortments and improving their inventory efficiency. While these investments were needed, all the benefits come undone if a customer cannot find their size. Worse, tightening assortments exacerbates the size stock-out problem, damages customer satisfaction, and creates significant financial pain unless size allocation is accurate. A 20% size misallocation drops margin dollars by 50% if markdown pricing is uniform across all sizes.
What is a retailer saying when they show you incredible outfits at a great price, but you can't find your size?
They're saying, "This is fabulous, just not for you."
About 2 years ago, I dove into the reasons behind this financial loss, the challenges merchants have been experiencing, and the shortcomings of traditional size optimization tools. Since the issue still lingers, I thought it was time to revisit the topic.
So, let’s talk about seven key areas where you can achieve precision size optimization, despite all your challenges and constraints:
1. The Foundational Forecast
Essential with any size optimization is the foundational forecast of consumer demand. This forecast must understand the differences between regions, stores, online, and even the fulfillment type.
State-of-the-art forecasts serve as the backbone for optimization and planning technologies such as allocation, fulfillment, assortment, and pricing decisions. This approach connects operations across different decision points to accomplish common goals.
2. Solving Data Sparsity Challenges
Data sparsity is the most crucial obstacle for precise size optimization. Even if you have perfect data history, data cleanliness, and data access – which, let’s admit, you probably do not – there remains the issue of the one-time, seasonal nature of fashion items. There is no history, especially if a size range changes from S/M/L to 2,4,6,8,10 or if you decide to add new fringe sizes. How do you manage the demand transference between these?
To solve these data sparsity problems, you begin by…
- building profiles at every level of a custom sizing hierarchy that includes product levels and attributes.
- analyzing profiles at each hierarchy level for data sufficiency, and then assigning a weighting based on the sparsity of data at each level for the profile.
- creating a composite size profile, if needed, by combining the multiple hierarchy profiles according to weight.
This process provides detailed level profiles where selling volumes are sufficient. Where they aren’t, profiles are inherited from higher levels.
This sparsity-weighted aggregate methodology allows the system to provide accurate profile values for new sizes compared to what has been purchased before. If a product is ordered in a new size, the relative demand impact of that size is borrowed from higher-level profiles and applied to the new size level profile.
Finally, the weighting approach provides a much better basis for sales imputation. Understanding where data sparsity can lead to flawed sales imputation assumptions will avoid over or under allocation of core and fringe sizes.
3. Providing Coverage Minimums
There are multiple methods to achieve coverage minimums, from a very complex calculation to a straightforward approach. We've found that the straightforward approach is a better solution through many retailer evaluations.
By establishing a threshold value, by store, for what defines a core size and then defining the unit coverage minimum for the core sizes, you can apply coverage minimum responsive to store/product selling without the complexity of detailed rules sets.
4. Establishing User Control
Size optimization will provide no benefit if your team does not trust the system, regardless of how good it is. The most critical components to building trust are visibility and control. Therefore, applications must provide a transparent analysis of current profiles with an easy way for users to modify profile details if required. The right solution will offer the best of both worlds by encompassing a high degree of modeling automation and profile generation alongside powerful tools for user review and entry, if needed, of profiles.
5. Omnichannel
Omnichannel must be natively built-in with the rest of the application, not as an additional add-on or +/- logic external to all decisions. Profiles should reflect store and online sales independently as size demand differs significantly based on channel. Hence, solutions should not aggregate or intermix channel demand; distinctive size profiles must be preserved.
Size profiling demand should also allow for ship-from-store (SFS) aspects of location inventory. This profiling enables any given location size profile to represent two demand sources: native sales and online sales fulfilled from store stock. This aspect becomes even more crucial as retailers increase their stores for online fulfillment.
6. Integration
Size profiling is an integral part of the merchant ordering process. Therefore, the solution must seamlessly integrate with any order management system (OMS). High-performance optimization allows quick API-driven size order quantities available as users write orders. Any value system should provide intelligence within the current OMS infrastructure to enable you to enjoy the advantage of advanced analytic determination of order quantities without replacing expensive ordering systems.
7. Taking that Last Step
Size optimization remains an unprioritized solution in many retailers' portfolios – possibly even yours. It’s often an afterthought because it remains at the end of an exhausting planning process, performed right before cutting the order. But what's worse:
- Putting in that last 5% to ensure all your prior work pays off, or
- skipping the last part to get it done, but then seeing all your earlier hard work not coming to fruition?
If you’ve been struggling with size optimization, let’s talk.
###
You May Also Be Interested in These Resources:
David Barach
David Barach is currently a Senior Director for Retail Software Solution Strategy supporting the overall product content, positioning and adoption for antuit.ai’s suite of merchandising solutions for retailers. antuit.ai was recently acquired by Zebra Technologies.
Previously, he has worked for 20 years in the retail solutions space in Pricing, Planning and Inventory Optimization solutions for 84.51 (Kroger Co,), SAS Institute, DemandTec (now IBM), ProfitLogic (now Oracle) and Spotlight Solutions. Prior to the solution space, he worked for 15 years in store, planning and pricing roles for Target Corp, Mercantile Stores and Macy’s Department Stores. He has an MBA from Memphis State University and BA from Vanderbilt University.
