4 ways to improve your data science ROI

4 ways to improve your data science roi
Nathan Trousdell
Nathan Trousdell

Director of Data Science & Strategy

For many companies, “data science” represents some kind of alchemy. Put a coin in the machine, pull the lever, and watch as your organization magically transforms into the next Netflix or Amazon.

Unfortunately, it’s not that simple. And despite costly investments into hiring expensive data scientists, consultants and machine learning engineers, most companies aren’t actually seeing any ROI.

This is primarily because there’s no clear business purpose to their data science activities. When the music stops and the purse strings tighten, the resources that are funding these initiatives will dry up – and the finger-pointing will begin. It’s then that data science will be at risk of being accused of being nothing more than an overhyped and expensive fad.

So how do you avoid the pitfalls of buying into the buzzwords, and actually turn data science into a source of value? Here are four key areas you should focus on.

1. Have a clear data strategy

To begin, your company must have a clear data strategy. This is tied directly to, and serves, the overall business strategy.

A good data strategy is comprised of many parts, and should be executed from both top down and bottom up. The image below illustrates what I mean:

4 ways to improve your data science roi

Source: Global Data Strategy Ltd.

As you can see, it involves multiple components, but these are two essentials that any company has to get right:

Creating a clear and simple data architecture. This is the blueprint for how your enterprise will access, manage and utilize data in a scalable manner. You’re going to start out with a landscape full of diverse, disparate data sources. Through mapping this into a centralized, well-governed environment—or what’s commonly called a data lake—you will be able to begin leveraging on your data effectively.

Putting such an architecture into practice creates the opportunity to get cleaned, defined data that is reusable, curated and accessible with a relatively small and focused team. Over time, your data improves in quality and accuracy, and you’ll avoid inaccuracies and frustration due to conflicting information.

Having involvement, alignment and buy-in from all stakeholders. If you’re not good at traditional business intelligence, you’re not ready for artificial intelligence. That’s why it’s crucial that your data strategy is, from the very start, developed together with people who can align it with the overarching business needs. If they don’t see the benefit in it, then you’ll get blocked before you can start.

2. Create and retain talented teams

Often, after deciding to become “data-driven”, the first hire made by most companies is a data scientist. This is a mistake.

You also need business experts with knowledge and challenges to solve. Data scientists or engineers without business context will be too academic to create real value for the business from their work.

Finally, the team should also be able to independently deploy their solutions as freely as possible. You’ll need to make sure that they have core IT involvement in the planning, as well as DevOps to implement solutions, instead of having their projects stuck in backlogs for months.

3. Integrate data science into the business from the start

Unless you’ve done this before, the sheer difficulty of trying to operationalize data science can come as a surprise. And as a result, data science initiatives often end up being left as R&D, which not only frustrates the talented people doing the work, but also makes it much harder to produce meaningful benefits. Data can be the difference maker for your enterprise, but only if you take it seriously from the start.

Data science models and artificial intelligence must be built into processes, applications and dashboards so that the benefits flow to end users who can deal with business problems at the source. To do this, you need to dedicate proper resources and planning to the initiative.

Only then will people see the benefit of data science and start believing in its potential.

4. Overcome cultural resistance

In order for data science to be an integral part of your business processes, your organization needs to embrace it as part of the culture. And changing culture is hard–but many data science groups make it hard on themselves.

One thing we did at Payvision to combat this was changing the name of our department, from “Data Science” to “Data Services”. Internally, there was a perception that we were just a group of people sitting in an ivory tower playing with algorithms. That’s really damaging to our cause, because we really do believe in being a group that serves the needs of the business.

What else do we do to overcome cultural resistance? Actions speak louder than words, so we conduct workshops with different groups throughout the company to understand their problems. When we develop a proof of concept, we also make sure to involve them in the process and then share the results with them. We want our data scientists and engineers to be as embedded in the business as possible.

We hope that this will begin to spark people’s imaginations, resulting in more ideas on how we can leverage on to deliver value – to our colleagues, and ultimately, our customers.

At a recent dinner hosted by the Ecommerce Foundation, I gave a presentation together with Crobox, an Amsterdam-based company that uses machine learning to boost ecommerce conversion. We talked about how organizations can successfully implement data science to grow their business, which led to some very interesting conversations with the audience.

If you’re interested in finding out more about how to begin your data science initiative, or if you’d like to share some valuable tips for doing so, I’d love to talk to you. Connect with me on LinkedIn and let’s chat!