© 2021 CMC-Global Directory, MIchiko I. Wolcott, CMC®
I’ve been on the service provider side of analytics for my entire professional life. I’ve also been in leadership positions in which I’ve had to evaluate the pros and the cons of getting external analytical resources for one reason or another. Having been on both sides of the matter has given me a solid understanding and appreciation for what makes a successful partnership.
Outsourcing for analytics (including data science, AI, machine learning, or whatever your favorite buzz term) has been a topic of discussion for some time. Is it necessary? Does it ever work? Does it make sense? Why would any organization want to outsource capabilities that are so intimately connected with its own data, full of nuances specific to its business and proprietary to the organization?
There are different interpretations of what “analytics outsourcing” means. For the sake of this discussion, and in the interest of inclusiveness, let’s take the broadest view here: enlisting a third party to execute some (any) analytical activity. The third party in question could be an organization or an individual. While there usually is some financial compensation involved, that does not have to be the case.
The reasons for “outsourcing” analytics, then, basically fall into one of two broad categories: for the capacity (bandwidth and resources) you don’t have, or for the capability (skills, expertise) you don’t have. Sometimes it is the mix of the two, but one is usually the driving factor over the other.
Outsourcing for capacity is conceptually easier and more straightforward. You may not want the headcount, it may be a temporary need and flex resourcing is more appropriate, or the third party has access to resources that are much more cost effective than what you have. “Resources” do not have to be limited to analytical manpower; they include computing resources and even administrative resources related to, for example, regulatory compliance. The bottom line is that the organization has the skills to do it but chooses to enlist external help. Perhaps for that reason, organizations that (intentionally) outsource for capacity tend to be more analytically mature (whatever that means). However, the flex resourcing idea is obviously very attractive and effective for early-stage organizations like tech startups as long as they know what they are doing.
The simplest case of outsourcing for capability is when the organization does not have any analytical capabilities whatsoever, and they get someone else to do some analysis for them. The lack of capabilities is common with not only small organizations and startups but also organizations of any size at the beginning stages of the data-and-analytics journey.
Besides the obvious skill sets and expertise reasons, there are some curious cases with roots in other factors. For example, you may need to incorporate an important external data source to your own data set, and that external data source may have restrictions on how it could leave the premises of the third party, if at all. Here, the external data source is essentially a “capability” that you do not have, and the third party may be in a better position to provide the analytical services. U.S. lenders engaging analytical services of credit reporting agencies (i.e., credit bureaus) on the lenders’ own customers due to regulatory constraints is one such example.
I have also seen organizations opt to outsource analytics for brand recognition. The organization may have internal analytical capabilities but partners with a brand-name analytical services provider for recognition. However, I find that these organizations do not always have an “exit strategy” for when the market recognition is established. As a result, some end up “brand hostage” to the external analytical services provider with subsequent difficulty building their own capabilities.
The biggest problems I’ve seen with outsourcing generally have to do with how it is outsourced not reflecting the true need to outsource: outsourcing for capacity as if you are outsourcing for capability and vice versa, or worse, naïvely outsourcing for capacity expecting capability in return. With that said, here are three tips or considerations for outsourcing in analytics.
Do you seek capability or capacity through outsourcing? The distinction is not trivial and a lot more complex than you might realize.
This determines what you should outsource and how you should outsource. For that, it is immensely helpful to know the difference between a technology architect, an analytical architect, a technology developer, and an analytical developer. If you don’t know, get help from somebody who does. Don’t hire an analytical developer when you need an analytical architect. Don’t hire a technology architect when you need an analytical architect. And don’t hire a technology developer when you need an analytical developer. Which one of these you do and don’t need depends on why you are outsourcing.
Do not assume an analytical developer knows how to design an analytical solution. As little as this distinction is recognized in the analytics world today, it is a critical one that can make or break your outsourcing efforts.
By far, requirements that are not sufficiently detailed are the primary reasons for cost and time overrun in analytical outsourcing. In contrast, the main factor in successful analytical outsourcing is an appropriate and agreed-upon level of detail in the requirements as well as in the acceptance criteria (see next).
Outsourcing for capability often depends on the expertise of the third party to design or provide guidance on analytical matters. This means the requirements tend to have more conceptual components and therefore are harder to articulate. Outsourcing for capacity, on the other hand, to get the return you need, requires a lot more detailed (and appropriate) technical requirements that have as little room as possible for interpretation. For that reason, it is much better suited for well-defined projects and activities.
If you’re outsourcing for capability, make sure that the business requirements are clearly articulated for the external analytical architect. A good amount of work is required to set business expectations—much more effort than people expect in my experience. It’s not true that you can just tell your vendor what you need, and the solution magically appears. That does NOT work. Your service provider needs to digest and translate business needs into analytical design. Don’t leave the analytical architect to infer or interpolate the details of your needs prior to translating. Definitely don’t expect a simple hand-off; too many unsuccessful and unhappy experiences result from just that.
If you’re outsourcing for capacity, make sure that you clearly state the technical requirements for the analytical developer. In this case, business requirements alone are NOT sufficient.
I cannot stress this enough. Many organizations don’t even know what acceptance criteria should look like for analytics. On the other hand, where else do you ever outsource without having acceptance criteria and SLAs and be successful all the time? Somehow, analytics has become the untouchable that defies all common sense. No wonder we are all disillusioned.
Make sure you clearly state the requirements for the deliverables and make them binding. This is not just for the analytic itself, but also for any executables, documentation, etc., that will help you understand and carry out your business that happens to depend on the analytics delivered. In some cases, penalties for failure to meet the acceptance criteria on time may be called for.
That it is data and analytics and therefore something you may or may not fully understand does not absolve of the need for you to articulate your expectations and hold your providers accountable. Outsourcing for analytics is not any different from outsourcing for anything else. Expectations, requirements, and terms that both parties can clearly understand are critical, independently of whether you understand the technical details behind their work.
Originally published at https://www.datadriveninvestor.com on April 22, 2021.