Questions to ask when evaluating AI for merchandise planning

AI is creating a ton of buzz, but you may still be wondering if it’s the right solution for your merchandise planning challenges. Here are four questions that can help point you in the right direction as you begin exploring AI for merchandising.

Does the problem I’m trying to solve map to tasks that AI does well?

AI is exciting in part because of how many things it seems like it can do. But the reality is that it can’t do everything…yet.

Spend time identifying if there’s a good match between “the problem you’re solving” and “what AI is good at.” One way to do this is to start with AI capabilities — reasoning, knowledge representation, natural language processing, etc. This provides a relatively broad way to orient your thinking.

Here is a non-exhaustive list of common fashion activities that align with tasks AI is good at completing.

“Forecasting?” you may be asking yourself. “What does that have to do with AI?” Learn all about AI and the biggest impacts to demand forecasting in our blog.

Does the outcome I’m looking for map to AI’s value?

In the same way that AI has certain tasks it’s better at completing, there are also benefits that it’s better at delivering. To ensure you’re getting everything you want out of your investment, it’s helpful to start with an understanding of typical AI value.

AI for Planning: The Four Pillars of Value

I. Informed decision making

II. Diversity of inputs

III. Granularity of outputs

IV. Speed to insight

I. Informed decision making

AI is great at helping us see more — it consolidates large pools of information so we can more easily examine it, and it also finds patterns that we can’t easily spot. If you’ve been stuck operating on gut feeling, AI can help move towards a more informed decision-making process.

II. Diversity of inputs

Consumer behavior, supply chain logistics, macroeconomic trends…the list of relevant considerations for fashion teams is nearly endless. But analyzing such a wide set of inputs is incredibly challenging to do alone. AI at its best identifies previously untapped data sources and makes it easier to investigate them.

III. Granularity of outputs

Simple tools deliver simple outputs. For the same reason that AI is so great at broadening what we can analyze, it also increases the specificity of what we learn. If precision matters, AI may be the right fit.

IV. Speed to insight

We use calculators to make math speedier. We use assembly lines to make production faster. And now, we can use AI to accelerate a huge array of tasks — including many common merchandise planning to-dos. Spend less time analyzing and more time acting.

Am I set up internally to make the most of AI?

There are three internal components that frequently block the successful adoption of AI at fashion brands: funding, data structure, and a willingness to change.

I. Funding

Funding is not a challenge that is unique to AI tools, but it is common. Merchandise planning teams may have very limited technology budgets, or they may be reliant on technology or innovation teams to green-light purchases.

The key here is to articulate the problem you want to solve + define how success would be measured. AI merchandise planning tools typically support metrics like margin or full-price sell-through — that needs to be something the buying team values.

II. Data

A great place to start is by asking “Do I have access to structured data?” There are AI approaches, like neural networks, that can learn from unstructured data. But some amount of structured data is usually a requirement.

Spend time identifying what data you would need and how it is currently stored at your brand. Common requirements include unaggregated historical transactions, stock receipt records, product attribute models, and purchase/transfer orders.

III. Willingness to change

Not everyone is excited about AI. Misconceptions about how it works or how it impacts existing roles can cause concern — or even active rejection of calls to add AI. For merchandise planning teams in particular, a common objection is “It’s working in Excel, why would I change?”

Before investing too much time in researching solutions, it can be helpful to first gauge how knowledgeable the team is. What are their concerns about AI? Do they know what AI is, actually? From there, building out a case for adopting AI, mapped to challenges the team surfaces, is a great way to build internal alignment.

Can I trust the provider or solution to deliver quality results?

AI is a highly complex technology. Most of us are not AI experts. How can we ensure the tech lives up to the promise? Trust in whoever is providing the AI is paramount.

Whether you hire a data scientist, engage a consultant, or onboard a technology provider, the criteria for good AI support remains surprisingly consistent. A few things to look out for include:

• A clear data philosophy
• An ability to describe why model types are used
• Knowledge of the fashion space — and ideally models developed specifically for fashion

But ultimately, the best sign that you’ll get quality results is when the value promised — the ROI — maps closely to the outcomes you value highly.

‍