More companies are creating data science capabilities to enable competitive advantages. Because data science talent is rare and the demand for such talent is high, organizations often work with outsourced partners to fill important skill gaps. Here are a few reasons to consider outsourcing. What can go right and wrong along the way?
A great number of companies are investing in data science, but the results they’re getting are mixed. Building internal capabilities can be time-consuming and expensive, especially since the limited pool of data scientists is in high demand. Outsourcing can speed an organization’s path to developing a data science capability, but there are better and worse ways to approach the problem.
“The decision to outsource is always about what the core competency of your business is, and where you need the speed,” said Tony Fross, VP and North American practice leader for digital advisory services at Capgemini Consulting. “If you don’t have the resources or the ability to focus on it, sometimes outsourcing is a faster way to stand up a capability.”
A recent survey by Forbes Insights and Ernst & Young (EY) revealed that most of the 564 executive respondents from large global enterprises still do not have an effective business strategy for competing in a digital, analytics-enabled world.
“Roughly 70% said that data science and advanced analytics are in the early stages of development in their organization,” said John Hite, director, analytic architect, and go to market leader for the Global Analytics Center of Excellence at EY. “They said they had critical talent shortfalls, inconsistent skills and expertise across the organization.”
Unfortunately, data science projects and initiatives can fail simply because organizational leaders don’t think hard enough about what the business is trying to accomplish. They also need to consider what resources, if any, are already in place, and how the project or initiative will affect people, processes, technology, and decision-making.
Have a Goal in Mind
Businesses are building their data science capabilities with the goal of driving positive business outcomes. However, success must be defined more specifically, and the results of the effort must be measurable.
“A lot of times, the client feels like the faster they launch a project, the faster they’ll achieve the outcome without defining first what needs to be achieved,” said Ali Zaidi, research manager at IDC.
Goal-setting, particularly at a departmental level, business unit level, or for a one-off project may actually work against a company’s best interests, especially when the strategic goals of the organization have not been contemplated.
“The first conversation [shouldn’t] just focus on the fire that needs to be put out, but the key challenges faced at the top level of the organization,” said Eric Druker, a principal in the strategic innovation group at Booz Allen Hamilton. “You also need to understand how analysis is currently done, in stove pipes, or whether data is being shared across the organization. You also need a coherent strategy for linking subject matter experts to data scientists.”
Even if the business problem is well-defined, the data science team, whether wholly or partially outsourced, needs to work backward from the goal to understand how the planned change will impact end-users, business processes, and decision-making processes.
For example, an EY client built a customer churn model that was capable of identifying which customers would defect in two weeks. Unfortunately, the marketing and sales teams needed 4- to 6-week lead times to take appropriate action, so the model had to be re-tuned.
“Starting with the end-user and how the [business] process is going to change can sometimes be overlooked,” said EY’s Hite. “Even if you get that right, do the end-users have the skills required, and are they incented to take the action you want?”
One company built a predictive model capable of identifying the customers who were likely to pay late. However, the customer service representatives tasked with sending payment reminders to those customers were compensated on customer satisfaction levels, not whether customers were paying their bills, Hite said.
Choose Your Partner Wisely
The growing demand for data science and data scientists is creating a market ripe for consultant organizations that now include the big consulting firms, systems integrators, traditional tech vendors, boutique firms, startups, and firms focused on specific vertical industries. One option is extending the relationship with a current service provider, assuming that provider actually has the level of expertise the organization requires.
“[If] you have a trusted partner relationship, you have everything contract-wise you need. Speed is paramount,” said Capgemini’s Fross. “You also need to consider who will give you the best resources immediately.”
Different parts of an organization may be outsourcing different data science projects or initiatives to different parties to achieve different goals. Sometimes the lack of orchestration among the various operating units can have an adverse effect on the enterprise.
“Data science is a cultural change in the way we make decisions. Firms that come in to solve an ad hoc problem miss all these great opportunities to understand the context for decision-making and how decision-makers use data,” said Booz Allen Hamilton’s Druker. “[If you’re working on an ad hoc basis,] it creates an impression that progress is being made. But because it’s firefighting, it may inhibit the movement on a data science capability down the road.”
Some organizations choose to work with outsourcing partners who specialize in a particular industry or who have consultants with specialized business domain knowledge. Others are looking for expertise that that is best found outside one’s own industry.
“People in your own industry will be laggards in the same way you are,” said Capgemini’s Fross. “If you want to understand customer context, you want to consider someone from a retailer, because they know context better than anyone. If you’re a pharma company and you’re trying to get your act together around data and MDM (master data management), you probably want to look at something from financial services.”
Regardless of which types of firms are on the short list, companies should put more effort into due diligence than they often do.
“As I’m talking to the vendors, I’m asking them about recurring business,” said Jennifer Bellisent, principal analyst at Forrester Research, in an interview. “How many of these projects are one-offs? And how often are you engaged on a retainer basis, so you’re not just doing a project, but you’re available to answer questions [and] be an extension of [the] strategy organization.”
Get Ready for Change
Outsourced data science projects and initiatives may be strategic or tactical, depending on the nature of the work and the mindset of the people hiring the help. A strategic engagement will involve an assessment of the business and what it is trying to achieve; an understanding of where the company is now, and what needs to happen when; and a concept of how the changes will affect the organization, its people, and processes.
Tactical engagements tend to address a specific problem, sometimes in isolation. Either way, the project will likely effect some level of change that should be comprehended and managed.
“The senior leaders need to understand the value that data scientists and analytics can provide, but we also need to have the broader community and managers at all levels understand the value, see the benefit, and [leverage] training for the take-up and use of the capabilities,” said Martin Fleming, VP, chief analytics officer, and chief economist at IBM, in an interview.
Only 10% of the executives responding to the Forbes/EY survey recognize data analytics as one of their core competencies. Those companies share three traits:
They’re using data analytics in decision-making most of the time or all of the time.
They’ve seen a significant shift in their organization’s ability to meet competitive challenges.
They consider themselves “advanced” or “leading” in their ability to apply data analytics to business issues and opportunities.
“Many companies have reached a reasonable point in terms of being able to produce analytics insights, but they’re having trouble driving that into business processes,” said EY’s Hite.
Companies are also having trouble retaining the data science capabilities they have in house, either because the roles are not adequately supported or because the individuals hired are not challenged with work they consider interesting enough.
“We often find data scientists are not part of a larger team. They’re sort of sole practitioners,” said IBM’s Fleming. “They either don’t have the level of support they need, or they don’t have the functionality that’s necessary, so they struggle with effectiveness and career development.”
Similarly, players on the outsourced team may leave. Even if they don’t, a lot of knowledge that could have been transferred doesn’t get transferred because knowledge transfer wasn’t part of the engagement.
“At some point, if customers realize their data science needs are increasing, they need to start hiring some of the skill sets internally,” said IDC’s Zaidi. “There has to be a stream of knowledge transfer, because when the project is done, the customer will have some of that knowledge inside.”
Continuing to outsource can get very expensive and so can hiring underutilized specialists in-house. This is another reason why companies need to understand their goals, their own ability to address those goals, and the resources they require to meet those goals — all of which are dynamic factors.
Stop Looking for a Silver Bullet
The savviest outsourcing resources have difficulty making a positive impact when the client organization is change-resistant and unclear about its goals. Tools can help facilitate good data science, but they’re no replacement for the human side of data science, at least not yet.
“You need to have a combination of human capital and software. If you just use human capital, it will take much longer to deliver data science solutions. If you just use tools, there’s no tool that is so cognitive it can provide business insight on its own, so you need human capital,” said Zaidi.
Some of the large consulting firms are building platforms and other software capabilities that supplement their service offerings. For example, EY recently launched its Synapse analytics-as-a-service platform, which expands the company’s managed services capabilities.
“Most IT organizations are struggling with traditional BI warehousing. Now we’re throwing big data constructs at them, and the way data science wants to leverage the technology is different from BI/OLAP environments and processes,” said EY’s Hite. “Finding the right mix of skills is important, but you also need the constructs to make it [work].”
There is no shortage of open source and commercial tools available. The constant stream of innovation is making it difficult to keep up. Outsourcing partners are often brought in to assess the environment and to make technology recommendations based on a client’s business objectives.
Tools are making it easier to operationalize data science, but the underlying data science must be sound in the first place.
Outsourcing partners are available in all shapes and sizes, but when it comes to data science, not all of them can solve the same problem equally well. Many outsourcing relationships are less successful than they could be because the client failed to consider its own objectives or the client organization resisted change.