What We Do

We are tasked to imagine, design and engineer the products that will shape Staples’ future.

Our Vision

Use data and technology to drive eCommerce to the point where every business process is optimized, and where every customer experience is special, personal, “indistinguishable from magic”.

How We Work

We design and build actual products, in small teams, leveraging data science and engineering to transform eCommerce.

Our Products

Personal Offers

Boosting sales by sacrificing margins is easy – try it, it just works. However, that is just a race to the bottom. And so is raising margins and sacrificing a good portion of your existing sales.

The real art is to somewhat manage to increase sales and margin at the same time; and this is exactly what our “Personal Offer” product pulls off.

The core idea is actually quite simple. Imagine that you have a product – say a notebook – whose cost structure is such that: - you still make good profits with a 10% discount - but would go quickly in the red if you were to offer a bigger discount.

One of the traditional approach for increasing sales for a notebook is to offer a 10% discount to every customer, or at least to a significant portion of your customer base.

This approach is far from being the most effective. To understand why, let’s focus on a concrete example. Bob & Alice both want to buy a notebook and come to your online store. The problem is that Bob is just in a hurry, and could care less about price, as long as the price is not completely outrageous. Alice, on the other hand, is a very savvy shopper, with a lot of time on her hands, and it would take at least a 20% discount to get her sale.

If you give them both a 10% discount what happens? - You secure Bob’s sale, but you unnecessarily sacrificed 10% of the cost in terms of margin, because he would have bought anyways. - For Alice, this is even worse. You lose both the sale and the margin, because 10% is not enough. Darn!

So what our “Personal Offer” product does is to: - detect that Bob is not price sensitive, and we do not give him any discount - detect that Alice would buy with a 20% discount, and we secure her sale by giving it to her, since we “saved” a 10% discount on Bob.

In this way we just doubled our sales and margins, all at the same time!

Of course this is easier said than done, because Bob might actually be price sensitive for specific products, specific times of the day or year, when shopping from a specific location, when exhibiting specific browsing patterns, etc. This is where our expertise in data science and machine learning comes in handy.

“Personal Offer” is already running on staples.com and demonstrating its capacity to boost sales & margins at the same time. This is just the start though – we are improving our models every day.

Personal Email

Another of our products running in production is “Personal Email”. The idea is also quite simple, but it is a powerful one.

We inject personalized content and offers on marketing and transactional emails. The personalization takes into account everything we know about the customer: his online and offline shopping patterns, browsing history, etc.

There is a twist though. We do not decide on the actual content when the email is sent. We defer that decision to the last possible moment: when the customer opens the email.

This approach is challenging in that it forces us to flex our engineering muscles: we need to be able to run sophisticated machine learning algorithms in a couple of milliseconds to avoid impacting the customer experience.

However this approach has also proven to be extremely powerful: - Recommendations can take into account customer activity up to the second before the email was open. This maximizes the relevancy and impact of the personalized content. - Even better, we can plug some adaptive algorithms in the mix that can self-adapt in real-time to patterns that would be hard to anticipate but that can be discovered in real-time: weather episodes, local events and celebrations, etc.

Smart Shipping

Another area of expertise for SparX is precise shipping time computations.

Some websites provide shipping estimations in a broad swath, very generic form: something like “1 to 4 business days”. This is a shame, because if you know for a fact that a specific product will arrive the same day for a specific customer – this is a competitive advantage – and you want to communicate it!

When you already have a great shipping infrastructure like Staples does, the great thing about communicating personalized and highly-accurate shipping times is that it has the potential to drive sales without impacting margins, all the while enhancing the user experience (timing is often of essence for our customers). Hard to pass up!

Our shipping models are based on historical data and can predict both shipping and handling times with a high degree of accuracy.

Experimentation Platform

We have a strong culture of fact-based decisions. This means that we measure all our products on a multivariate testing platform, that we designed and built from scratch to match our needs.

continue to improve on the intelligence in this platform whenever you have new findings…

In addition to pure measurement, we also leverage our experimentation platform for safe and progressive roll-out of our new algorithms and product feature.

At any point in time, we are conducting a lot of concurrent experiments, through multiple channels. The key challenge in this setup is to quickly reach statistical significance, but without sacrificing significant traffic nor profit. We keep learning every day as, in real-life, this problem is far from trivial.