Data science is certainly at the top of many business leaders’ minds at the moment. It seems that almost daily, headlines announce things like the increase in the number of firms that are investing in AI and analytics, or about the trillions of dollars in “business value” that experts say will be created in the next decade by these technologies. On the other hand, other experts tell us that a shocking number of data science projects (50-85%, depending on who you ask) never make it to production.
These two threads don’t seem to connect; how can so much business value be created if the vast majority of these efforts never make it to production? Let’s examine some trends in more detail and I’ll share why Atrium created a service offering specifically designed to help you build and deploy solutions that provide impactful results.
We started Atrium four years ago because we saw an opportunity in the market to help our customers transform their CRM systems into systems of intelligence by leveraging data science and analytics techniques. We have learned quite a bit about the best ways to help our customers over the last three years, and I think a few of the lessons we have learned will help explain why projects like these can be difficult without the right support.
Business Leaders are Now IT Buyers
While this observation will likely seem obvious, there are some interesting market forces that make addressing this challenge more complex than it may seem. First, the last several years have been dominated by a tectonic shift in the technology landscape: the transition from on-premise systems to the cloud. In fact, the use of distributed computing resources has become so commonplace that it almost seems trite to call it “the cloud” at all; it’s simply just how technology solutions are delivered in the 2020s. This transition has changed how technology investment decisions are made and, more importantly, who is making those decisions. Business leaders (think sales, marketing or customer service executives) are now IT buyers, thanks to the ubiquity of platforms like Salesforce. In turn, these platforms have advanced to the point that they make previously challenging parts of IT projects easy; security concerns, platform scalability, and ongoing platform support are simply part of the price of admission for these business IT buyers. In addition, they have now been trained that implementing these platforms is relatively easy, and requires little ongoing maintenance.
Data Science Projects are Different
What’s more, these new IT buyers have never had to think about their data specifically. In the past, teams have had to make sure that the system captures the data needed to feed a particular business process, and the kinds of data collected would only change when the underlying process changed. Data science projects demand a different way of thinking about your information.
Here’s a simple example: A picklist of lead sources that included “Other” as a choice would have been perfectly acceptable in the past. However, if half of your team selects that value when entering data into the system, it becomes meaningless from a data science point of view. A model that predicts “Other” as the most important lead source is not helpful for a business user. “Other” simply isn’t actionable.
Data changes as quickly as business conditions change, which is certainly faster than your business processes are updated. If you don’t maintain a predictive model as your data changes, those predictions will be based on old data, which will in turn cause those predictions to be equally outdated and therefore of diminishing value.
Data Science Projects Require These New (and Difficult-to-Find) Skills
At first glance, this is also pretty obvious: data science projects require data scientists, and people with those skills are in high demand. However, the challenge is even greater than that. In order to implement a successful project, you need a team of people with complementary skills that support your data scientists:
There is a significant amount of effort required to extract and manipulate data to prepare it for easy analysis by a data scientist. Having this skill on your team is much more efficient.
The output of a predictive model is often difficult for a person to interpret without help. What does a score of 0.8 mean in a model that predicts customer attrition risk? Is that good or bad? Resources with good analytics and visualization skills can take these outputs and transform them into actionable insights.
Your team must also be able to understand what’s going on in your business. Without this insight, these projects often turn into “science experiments” — models or dashboards that have very little tangible business value or impact.
You’ll also need to have resources that can make necessary changes to the underlying platform to gather more or better data, or to display the right information in the right places to allow your users to take action.
Complicating this challenge even further is the fact that you need all of these resources working together in order to deliver a successful solution. If you lose any one of these capabilities, your ability to support your data science investment will be hampered.
How Can Elevate Help?
All of these challenges can make an investment in data science seem daunting, but these projects are simply different than traditional software projects; you cannot just make a quick-hit investment. Not only do you need to consider your data and data strategy, but you will need a team of people with expensive, difficult-to-find skills in order to support your data science or analytics investment.
We designed our Elevate service offering to address these specific challenges. We are providing our customers ongoing support and enhancements of their data science and analytics solutions with dedicated resources with the essential capabilities. Here’s how it works:
- Continuity. Our Elevate teams are organized into “pods”; each pod contains all of the skill sets required to support our customers: business analysis, data science and analytics, platform skills, and quality assurance. Each pod then supports a subset of our customers, which allows our teams to become deeply familiar with each of the environments they support. This allows us to work efficiently and provides our customers access to a dedicated team.
- Flexibility. The pod structure also allows us to be flexible with work demands. If one customer needs more data science and less analytics in a particular release, we can simply redistribute work to different members of the pod.
- Innovation. Our flexible approach to work enables innovation for our customers. We’re able to work on what’s most important and impactful to our customers at any given time. Our engagement leads work with our customers throughout the engagement to identify and prioritize the efforts and resources needed to deliver against our customer’s highest priorities.
Let Us Solve Your AI and Analytics Adoption Challenges
If you’d like to learn more about Elevate and how our customers are working with us to solve their AI and data science challenges, take a look at our success stories studies on Robert Half or Staples.
You can also contact us, and we would be happy to discuss your AI and analytics adoption challenges.