Logo

The Data Daily

5 Reasons Why Most Data Science Projects Fail To Get Adopted

Last updated: 03-24-2020

Read original article here

5 Reasons Why Most Data Science Projects Fail To Get Adopted

Some years ago, at Gramener, we built a customer churn modeling solution for one of the largest global telecom operators. The machine learning solution predicted which of their customers would leave, one month before they stopped usage. In test pilots, the solution helped reduce customer churn by more than 56 percent compared to the earlier process.

We were amazed at the impressive results and stellar accuracy. But the celebrations were a bit premature, for the solution was never used. Despite a sound solution and successful pilot, months of our painstaking effort went down the drain.

Sound familiar? Unfortunately, this is far too common with technology transformations and data science initiatives. What causes these failures in adoption? How can you successfully navigate the necessary change management?

Here are the top five challenges your analytics projects will face, and how to tackle them.

A Gartner report says that 80 percent of data science projects will fail. Most initiatives don’t deliver business benefits because they solve the wrong problem. The problem with these pilots is that most of them are too technology-focused, quite like science fair projects. They are not driven by the challenges organizations face, and hence end up not being useful for the business.

Never start data analytics initiatives by talking about data or analytics. Brainstorm to identify all the roadblocks for your business objectives. Prioritize them based on three factors: business impact, urgency, and feasibility. Pick the top-ranking business problems from this list and carve out a solution using data science.

Your data science journey is like a 30-hour marathon hike. Pick the wrong target on the map, and all your heroics will be in vain.

Teams often get over-ambitious and build a solution for everyone. If you aspire to satisfy everyone’s wants, you will meet no one’s needs. Another common mistake is to build on behalf of the end users without talking to them. If you miss understanding the user’s context or their natural workflow, your solution will stick out like a sore thumb.

Start by defining who the users are — and who they aren’t. Conduct interviews, build personas, and understand the scenarios of usage. Be ruthless in prioritization to trim the asks. Design the solution into their natural workflow. Ensure that the model results are explainable and talk to the user’s needs. Yes, you’ve heard this before, but it’s missed too often in practice.

Take the case of Grammarly, the popular grammar-checking tool. While the model suggestions are accurate, its biggest win is in designing a solution that reduces friction for the user. The spell-checks work in-line, across apps, and they continuously learn from user feedback. Now, imagine if you must copy and paste the text into a separate interface to do your spell checks. Would you use it?

If your project addresses a burning need for defined users, then they need to know about it. Even the best of solutions need a sales push. People often think that if you build it, the users will come. But projects die a slow death without carefully planned marketing efforts. Organizations make the mistake of leaving the internal marketing efforts to teams that build the solutions.

Define your go-to-market approach for every internal project. Plan launches by the executives, organize roadshows, and run internal campaigns. Engage users through gamification and hand out cool giveaways to spread the message. Accept the help of professional marketers. Track your adoption metrics and celebrate small wins. Quantify return on investment to show the value and help secure future budgets.

A good example is the launch of a TV audience analytics solution by our client, one of the largest media companies in the world. Starting with the senior leadership rollout, a national roadshow was organized with champions covering every regional team. This led to stellar adoption with continued usage.

Organizations often get too fixated on the first version and don’t plan enough for scaling. Even the most successful products need several iterations and constant tweaking to get the solution right. Most projects don’t plan this runway to let their solutions take off. They don’t set aside a budget to improve on continual user feedback. They miss setting this expectation with users.

While building your first version, plan for a broader vision and roadmap. Explicitly set aside a budget, resources, and expectations for rapid revisions. Avoid letting teams getting too wedded to the initial solution and keep them willing to make major changes. Plan periodic upgrades to keep the user interest levels high and avoid a loss of momentum.

Ed Catmull, co-founder of Pixar Studios, famously said our most daring ideas are ugly babies: “They are truly ugly: awkward and unformed, vulnerable and incomplete. They need nurturing — in the form of time and patience — in order to grow.” Having a strong vision at the start helps, too.

Even when all of the above are taken care of, initiatives will fall flat if they lack an executive mandate. Change is not easy and the natural human tendency is to resist it. In addition, organizations often have conflicting priorities. This makes new initiatives highly vulnerable in their early stages. If not nurtured carefully, transformation projects stand very little chance of success.

Innovations must be led from the top to see the true benefits. Executives must present a vision for the future and rally people towards it. You need to push firmly to abandon old habits, at times, with unpopular calls to avoid a relapse. Make sure to onboard leaders at the next levels who can champion the initiative and act as change agents.

According to a survey by Deloitte, executive sponsorship is vital to organizational change. Companies with the CEO as the lead champion are 77 percent more likely to exceed their business goals significantly.

At Gramener, we learned our lessons from the project with the telco churn model that failed to get adopted. When the next opportunity arose, this time with a global conglomerate, we started the data science initiative differently.

We found that their top business challenge was to make better decisions in commodity purchases and sales. Working with their target users, we prioritized the features based on what was impactful, urgent, and feasible. We decided to forecast commodity prices, and we built a minimum viable prototype that used explainable time-series models for forecasting.

Early feedback showed that what users really wanted was the direction of movement, not the exact price forecast. We changed the solution by picking a different class of simple models. It was actively marketed to all user segments, with iterative improvements.

We channeled executive firepower to build momentum for the initiative and promote usage. In production, the solution led to a savings potential of $22 million for one of their largest commodities. This time, we didn’t have to watch our efforts go down the drain, and the celebrations were shared by all.


Read the rest of this article here