Should I Outsource All Data Science Functions to Just One Vendor?

You may now be trying, or have tried in the past, to build a core competency in data science. Perhaps you are considering bringing just a few data science techniques, such as predictive analytics, into your company to start with.

No matter where you currently are, deciding whether to outsource is the first step. Your second step will be to decide whether to outsource just predictive analytics or all the components of data science: data management, business intelligence and predictive analytics. The good news for you is that there now are outsourced solutions that can provide expertise in all three areas. But, there are tradeoffs.

Use case: expanding from business intelligence to predictive analytics

Meet Connie. She oversees customer value maximization for a small but growing retailer. In the course of her job, Connie has implemented a few business intelligence practices that give her insight into what is happening with her customers. For example, she is using a rudimentary data lake along with ad hoc reporting. This is helpful, but only to a point, as the insight Connie receives is strictly ‘post mortem’ – or after the fact. This means that all Connie can do is to look at a situation after it has happened and attempt to understand past causes and results.

Still, this basic set up provides Connie with tremendous, data-based insights that allow her to quickly discover and correct problems. Naturally Connie would like to do even more if she could. What Connie really wants is to be able to see into the future and forecast which customers are not likely to renew their orders. In addition, she’d like to predict what offers her company could extend to these customers to entice them to stay instead of leaving.

There are other important areas where Connie’s company would like to use business intelligence and predictive analytics:

  1. Fraud detection
  2. Optimal marketing mix of products
  3. Forecasting sales and inventory

As the person in charge of making such decisions, Connie recognizes several strong arguments for outsourcing business intelligence and predictive analytics:

  1.  Quicker and easier to implement than in-house solutions
  2.  No long-term investment in employees or recruiting needed
  3. The ability to rapidly assimilate best practices

Without any current data scientists on staff, Connie is not quite sure how to get started. Furthermore, she wonders which functions (data management, business intelligence, or predictive analytics) should be outsourced or if outsourcing all three is best.

What are the benefits of outsourcing data management, business intelligence and predictive analytics to a single provider?

After deciding to outsource, you’ll need to decide if you are just outsourcing your predictive analytics or all of your data science competencies. This is, unfortunately, not a simple decision. What you should consider asking when deciding is, what do we stand to gain from outsourcing everything as opposed to just our predictive analytics? Specifically, you gain a vendor who already understands and works with your data to do more than just predictive analytics.

From the 30,000-foot view, if different vendors manage different aspects of the solution then each team needs to understand the data independently. Which is a replication of effort and can lead to mistakes when communication and interpretations of the data occurs across different teams. With just one team handling all data issues, synergies could be leveraged and opportunities for mistakes could be limited.

Having all these teams working and sharing knowledge together provides several advantages. Many companies extend extra effort and resources to make sure that the data science team understands business issues and that the business units understand a little about data science. In a recent survey from MIT Sloan and SAS (“Minding the Analytics Gap”), 34% of organizations in the survey reported training their data science professionals in the business issues of the company. This can pay such tremendous dividends down the road that participation in this kind of training should really be closer to 100%. Best practices would therefore be to do this in-house as well as to find an outsourced partner that takes this kind of training seriously and to budget for this training as you plan.

Are there challenges in understanding the business model when you outsource to another country?

As with any business undertaking, challenges certainly exist. However, most of them may be minimized by adhering to well-developed, industry-wide best practices.

Some of the advantages of outsourcing are also its risks. By nature, outsourcing makes it easy to find and recruit great talent and get up and running quickly. However, once you have chosen your partner, you will still want to continuously challenge them to maintain high levels of quality and performance. If they have all of your data and it is expensive to move it to another vendor this may prove difficult. As such, avoid any situation where your data could potentially be ‘held hostage’ by an outsourced vendor.

A simple contract requiring regular (e.g. weekly or monthly) full and functional database dumps of all data, predictive models, and metadata can prevent this. You may also consider hiring a second vendor to work on some of your data as a precautionary measure as well. This is a bit redundant and inefficient in the short term (and a bit of an administrative headache), but it remains a smart practice in the long term because it validates the quality and performance of your primary vendor. (See also my previous blog on predictive analytics mediators.)

What are current trends in outsourcing predictive analytics?

Here are some of the dominant trends in outsourcing complete predictive analytics solutions:

  1. Predictive analytics is becoming a requirement. As continuing advancements in data size and quality have now made predictive analytics possible, it is becoming more and more often a requirement in order to secure competitive advantage. 
  2. Business intelligence and predictive analytics are being integrated. While business intelligence and predictive analytics were developed separately, they are beginning to converge into seamless, more fully-integrated offerings.
  3. Business intelligence tools are becoming more self-service. The transition that started nearly a decade ago from old-style OLAP (Online Analytical Processing) to self-service business intelligence (BI) tools will continue.
  4. The cloud. Everything is moving to the cloud. Computer and software maintenance has become a commodity and, as a result, much cheaper and more robust. Technology maintenance has moved out of corporations’ core competencies, allowing them to instead focus on building core competencies in predictive analytics and business intelligence.
  5. It will be Software as a Service (SaaS) all the time now. Just as hardware maintenance has transitioned to a SaaS model, business intelligence and predictive analytics will both increasingly be utilized as a service.

How should my business plan for the next four years?

Remember Connie? She should, as should anyone with her decision-making power for their own company, strongly consider outsourcing predictive analytics, business intelligence, and data management functions to a full-service vendor. Just as companies today outsource billing and payroll, they will outsource data functions more and more frequently over the next four years. The key to success here will be the development of industry best practices that allow companies to evaluate and measure the performance of their vendors in a convenient and transparent way.

Experts

Many thanks to Gurpreet Singh who provided expert feedback on this article. Mr. Singh is founder and CEO of DEFTeam Solutions.

Are you an expert in applying data science to business?

I’d like to hear your thoughts. Drop me an email: [email protected]

Stephen J. Smith

Stephen Smith is a well-respected expert in the fields of data science, predictive analytics and their application in the education, pharmaceutical, healthcare, telecom and finance...

More About Stephen J. Smith

Books by Our Experts