You want to make your data projects count. There’s no point wasting time and resources when there are simple ways to make the most of the process. But before we jump into what you shouldn’t do when outsourcing a data project, first let’s take a step back and look at a project that ran like a dream for one of our clients.
We were invited to visit the client’s offices and we stayed there for a week, simply talking to people, listening to them and learning what their needs were. Our board level sponsors gave us interview access to every member of the company and, over those days, we learned so much about the business that we were later told, “You probably understand our technical operations better than anyone else here!”
At that point, we were ready to deliver the best data science work we’d ever been a part of. We got beneath the surface of what they thought they wanted and we could create what they actually needed instead.
This anecdote demonstrates how we approach every data project: as a process of first understanding people and their needs, and only then conjuring up creative ways to use data science and AI to help them.
Unfortunately, many companies outsource their data projects in far less effective ways, making common mistakes that lead to a project stalling, crashing or falling short of the desired destination. Here’s a roundup of what not to do.
1) Stopping short of collaboration
To outsource a data project well, you need much more than a one-click process. That might work for ordering your office supplies but not your office’s virtual architecture.
Data projects require intentionality. To get a good outcome you’ll want to be heavily involved. If you simply sling your requirements over the fence, you’ll get something thrown back - but it won’t be fully formed and it’s unlikely to achieve the results you’re looking for.
2) Putting your suppliers on rails
On the other end of the scale, you don’t want to get so involved in the process that it becomes your own process. Suppliers will inevitably have their own practices that are proven to get results. To make the most of their expertise, you’ll need to trust their ways and means.
3) Forgetting it’s a people project
Data projects might seem like they’re made up of numbers and code but they’re actually made up of people. When you’re thinking about which data experts to put on the case, pay attention to how much they prioritise relationships.
You’re going to want people who can be honest with you, who can tell you the hard truths you need to hear. Data projects often evolve in their scope, requiring rethinking and iteration.
In those moments, you want a supplier who values the relationship enough to push back. If they’re committed to collaboration, they’ll steer you clear of potholes and help you to achieve more than you ask for.
4) Losing track of what’s essential
For a moment, forget about data projects and think about the list that once described the romantic partner you wanted. If you are in a committed relationship now, you’ll be able to appreciate how many discrepancies there are between the list and the person you actually wanted to be with.
It’s very similar with data projects. (Although granted, the initial meetings aren’t quite as awkward and fun as first dates.) Typically you will have a list of needs and outcomes that you wish to communicate to the supplier, but a good consultant will bring some fresh, new perspectives to the conversation. The best outcomes will be achieved only when both parties can be open-minded about the nature of the problems, challenges, and potential solutions.
Some of your ideas will inevitably be a little to the left of what you need or want. For your data project, you’ll need to reflect on what’s actually essential. You’ll also want a supplier who can jump into the details, find when two ideas are in conflict, and help you decide which one is the deal breaker.
5) Assuming all data projects are the same
There’s more than one kind of data project and each requires a different process.
For instance, a data engineering project is mostly about task management. It’s about breaking one goal into pieces, breaking those down into smaller sections, and stringing together a long list of micro tasks that engineers can work on systematically.
But a data science project is very different. In many ways it shares more in common with product design than data engineering. The medium is still data but at the heart of it is a creative process, one that requires much more ideation and collaboration.
So four weeks into a 20 week data science project, you’re not going to be 20% of the way there. It’s not an efficiency equation but a creative one and it’s much harder to turn a project like that into something with a start and end point.
Sometimes people outsource their data projects by saying, “Here’s my data, do something with it.” But that’s a little like handing sheets of plastic and MDF to an architect. In this scenario, creating anything that’s fit for purpose is a matter of guesswork.
For a supplier to make a data science project work they need to take the time to understand you, your business and to get under the skin of what you’re trying to achieve. You might not invite them onto your site to get familiar with the needs of all your staff. But the more runway you give them, the better the take off will be, and the further you’ll be able to fly.
To find out more about how you can make the most of your data project, don’t hesitate to get in touch.