What makes a successful AI assortment solution?
Fri Jan 10 2025
Author
Alok Joshi
For retailers, the volume of data becoming available to analyse throughout their entire value chain, from supply chain and logistics, to customer interactions in-store and digitally, has massively increased the opportunity to leverage AI to optimise these interactions. Given this data, AI can greatly improve customer satisfaction by providing a more tailored experience, starting with the range of products that are made available to them - assortment planning.
The specific assortment planning challenges vary by market niche, but success typically hinges on the answers to the two most common and crucial questions: “What do our customers actually want?” and “How can we lower the cost of value-adding and selling?”
Only by excelling in those, businesses can secure their stake in the game for the long run. AI-powered assortment plays a critical role in answering these questions.
The Future of Assortment Planning: Why AI is a Must
The end goal of assortment planning is to range the right products, in the right locations, at the right times of the year. Objectives however can differ greatly depending on the needs of the business. For example, maybe the business strategy is about market penetration so increasing sales is the objective, perhaps it is about minimising operating costs so profitability becomes the key driver, or perhaps there is a need to balance both.
There are so many factors that must be considered when trying to achieve an optimised assortment to meet your objective. Think about the products to protect, to delete, % coverage you want to maintain, how many facings are available at each location, the level of diversity and the impact to cannibalisation, what days cover you want to maintain, what cost-to-serve threshold needs to be adhered to, lead times to fill, capacity to stay within, what level of change to the overall assortment will be acceptable by your customers, what products conflict with each other, what makes good bundles, what would happen if you increase or decrease the shelf capacity, change the safety stock or replenishment frequency etc.
As you can see, there could potentially be billions of permutations and combinations of settings and product assortments to trial, all depending on the objective you want to set. This is where the power of AI comes in. AI allows for seamless computations of predicted outcomes when certain variables are changed and an outcome is simulated. This ability to simulate is why AI is a must to stay ahead in today’s competitive retail landscape. It’s a way to have smarter AI-driven decisions that can help you be proactive to meet customer needs.
Crafting an effective AI-driven system requires more than algorithms—it demands thoughtful design, integration with business processes, and a deep understanding of retailer needs.
Components of a Good AI Assortment Solution
A well-constructed assortment solution requires more than just data-driven demand forecasting and optimisation algorithms to make an accurate prediction that answers these 2 questions. Before even getting to the models, AI must have an established platform to enable scalable growth and learning. Learning requires quality and complete data to enable the “engine” (i.e. the algorithms) to provide quality recommendations and predictions. And no matter how good these algorithms are, frameworks to land and build trust with AI are needed to realise the benefits of optimisation. Below are the core components that form the recipe for delivering successful value.
Laying the Foundations:
- Modelling (ML) Platform: A robust, scalable, and UX/DX friendly Modelling (ML) platform that supports the entire lifecycle of forecasting models and optimisation algorithms, from build, experimentation, to production and ongoing maintenance. A crucial component to get right in order to keep ongoing costs low.
- Data Pipeline and Engineering: Historical and forward-looking data should be thoroughly cleansed and include alerts for potential issues. This data must be engineered to accurately reflect the patterns and relationships between assortment and demand. The better the data fed to the models, the more accurate the recommendations.
Core Solution:
- Demand Forecasting: Accurate and holistic demand forecasting solution that captures the relationship between assortment and demand, and with API support for scenario planning for assortment planning.
- Assortment Optimisation: Modern and scientific algorithms to help decipher potentially billions of range and planogram combinations and select the best / optimal assortment for the objective that's right for your business.
- Cost to Serve: Consideration for operational costs such as transport and labour associated to execute the assortment so optimisation recommendations are still profitable. This helps to achieve an end-to-end view of assortment changes needed.
- Holistic Architecture: A flexible and extensible model architecture that integrates different components above, and ensuring assortment is not a siloed process but is aligned with other activities in the value chain to create optimised assortment plans.
- Scenario Planning: Ability to leveraging demand forecasting and the assortment optimisation. engine to take any input scenario, predict outcomes, and determine the best actions to take through simulation of multiple parameters that could change the outcome of the assortment’s level of success.
- Friendly UI: A comprehensive view and streamlined workflows in an intuitive UI, providing insights into the performance of assortment allocation and enabling users to manage, collaborate, and adjust the plan seamlessly. An assortment is not a set and forget activity, it needs to be actively monitored to respond to changes in customer behaviour.
Landing the Solution:
- MLOps + Production: Ongoing production support to maintain the assortment solution needs to be made easy to diagnose and remediate issues quickly so trust is not degraded in the solution.
- Human in the Loop: A way for users to interact with, provide feedback and governance on the allocations. As much as we want all decisions to be automated as much as possible with AI, this will be a crucial part of any AI solution to ensure there is no reputational or legal damage.
- Delivery: Strong cadence of iterative delivery of the right priority features to make sure the product evolves with the changing needs of customers.
- Change Management: Landing the assortment planning and building trust needs well thought out change management frameworks that address training, communication, perception and overall endorsement of the solution.
- Measurement and Improvement: A way to realise and track the value delivered by the assortment planning, not just for accounting (although that is nice!) but also to learn and improve the entire process
How Does Your Assortment Solution Stack up? Take the Self-Assessment
We’ve put together some assessment criteria with a rating system to help you determine where you might need to improve your assortment solution. This will get you to think about the maturity level of the components/capability you have in the current state. Once completed, may even be the trigger to develop a strategy to address any shortcomings to get your AI-driven assortment to the level desired.
Figure 1. Assortment Solution Self-Assessment Criteria
Got a question on assortment or need assistance with the assessment? Reach out to us to have a chat at info@jahan.ai. We’d love to walk you through what good looks like, assess your current state and help develop strategies to address the gap. If you are interested to learn more about assortment, why not check out our SaaS offering or download our Retail Assortment Handbook - full of advice and best practices to set you up for success.
#assortment #range #assortmentplanning #retail #AI #AIsolutions #machinelearning #jahanai #saas
At jahan.ai, we build end-to-end AI twins for retail and supply chain businesses. We take the concept of a digital twin, a virtual replica of business processes, further with advanced AI, which not only mirrors these processes but also automatically optimises them. In this way, we support businesses in driving efficiency, going beyond their potential.
Reach out to jahan.ai to explore how we achieve this or how we can help optimise your business.