Introducing: Data Science at Syrup

Hi there, I’m Mike, Senior Manager of Data Science at Syrup. If it’s not clear why an inventory optimization company would need a team full of data scientists, this is the post for you.

Let's start with the basics. My team and I are challenged to deliver the best forecasting and optimization engine available for brands and retailers, powered by AI.

But data science can’t operate in isolation, at least not if you want to have meaningful product and customer impact. We have purposefully set up our company so that data science forms its backbone, and likewise, AI is a core element of our very identity.

Why? It boils down to three key reasons that all tie back to our company mission: to create a world where commerce is no longer wasteful.

1. It ensures AI is a central element of our product.

It’s all too easy, especially these days, for AI to be a hooky but unoptimized add-on to existing software. In many cases, product vision and AI vision are separate things — the same basic product gets marketed with some flashy AI buzz, but the two strategies remain independent.

At Syrup, that product-AI disconnect is unthinkable. And it all comes down to delivering customer value.

Our models are tailor-made for the apparel and footwear industry, giving us an extra edge in accuracy and insight. This means that every tweak, every improvement we make to our forecasting engine is rooted in the unique challenges and intricacies of the industry.

But model tweaks in isolation won’t mean much for the omnichannel brands that trust us with their precious inventory. Our AI services form the very identity of our platform, and that’s where the large majority of our data science energy is focused. It is, after all, the foundation on which everything else sits.

We also have data scientists working across our engineering and product teams, and our data science chapter helps ensure we’re all marching to the same beat. This close cross-functional approach ensures that our data science efforts are always aligned with our product vision, eliminating any gap between theory and practice.

2. It allows us to push the boundaries of what AI can do — for each unique customer and at scale.

Demand signals available for analysis are exploding…but making use of those signals to accurately predict and act on demand increasingly requires the complex, multivariate approach that only AI models can support.

My team’s mission, in a sense, is to crack the code of how to make use of these signals — and then ensure our customers can simply and quickly make the most of that behind-the-scenes sleuthing.

The starting point is our experimentation roadmap. We’re exploring new methodologies, playing with cutting-edge models, and pushing the boundaries of what AI can do with fashion-specific data.

I’m excited to announce that we’ll be publishing the results of some of our experiments in a series we’re calling Always Innovating.

Beautifully, the design of our core infrastructure makes it easy to share this value across the planning, buying, and allocating workflows that are built on our platform. Each one of our customers benefits from these experiments, regardless of the specific products they’re using.

And while cutting-edge models certainly have their place, our depth and breadth of expertise allow us to carefully select the most impactful models for our customer's unique apparel and footwear business strategy. It's not about adopting the latest fad; it's about leveraging the right tools to drive real, tangible results: reduced stockouts, full-price sell-through, rightsized production, and meaningful margin improvements.

We also spend a significant amount of time approaching the problem from the other direction: what can we do to ensure our AI is solving for each customer’s unique business needs appropriately?

When we onboard customers, we take the time to understand their data so that we can tinker with features, hyperparameters, and other model elements. The end result is that each customer will likely end up with a model that is ever so slightly different from any other customer: uniquely yours.

One customer may have a simpler model with only a few holiday features, while another customer may have a complex model with a lot more features. One customer may have a hierarchical model, while the other customer may have just a straightforward SKU-store-week level model. One customer may have different models for ecommerce vs. retail, the other customer may have a mono-channel model.

You get the picture.

What we’re not doing is approaching this in the “legacy” way — hastily cobbled together models for every customer with no shared insights, no shared testing infrastructure, and ultimately, no real results.

3. It ensures we’re not offering more of the failed promise that the industry has come to expect.

Perhaps most importantly, our dedication to deeply rooted data science fundamentals ensures we're not just offering more of the same old song and dance. AI can seem like a new offering in the fashion context, thanks to the proliferation of design/creative-oriented solutions powered by GenAI.

But industry vets are very familiar with the siren song of AI, especially for inventory demand…and its historically disappointing outcomes.

With advancements in machine learning, increased data accessibility, and beefed-up processing power, we're in a new era of fashion tech — one where AI isn't just a buzzword but a meaningful, helpful tool.

In many ways, we want to make AI boring but indispensable: the calculator you bring to your math exam.

So there you have it. At Syrup, data science isn't just a team or department — it's the beating heart of our operation. From fine-tuning our engine to pushing the boundaries of AI innovation, we're shaping the future of commerce one experiment at a time.