Data Science and Organizational Structure
When a company wants to make data one of the prime focuses of the business, often the first problem that arises is how to fit a Data Science team into the current organizational structure. Similar problems have risen in the past with Business Intelligence/Business Analytics, and other disciplines, but the reason that Data Science presents a different problem is because it is a superset of previous areas in terms of technical and business scope. Since a Data Science group is concerned with so many areas of the business, and many levels of technical and product detail, it’s not always straightforward to arrange a Data Science team within the existing organizational structure.
In order to find the best place for the group, the first priority is to understand what it will actually do in the organization. What is the specific mandate from the board? From the CEO? Questions on overall scope and company strategy are an absolute prerequisite. Assuming the company has an overall strategy to use data more effectively and not just some short-term product goals, the Data Science team should be focused on concerns across departments and hierarchies. This could include things like building forecasting or customer churn models for Sales, helping HR optimize recruiting, helping Marketing or Sales better target potential customers, or building new recommendation systems into the product.
Within the scope of individual product or service offerings, a Data Science team is concerned with many levels of detail, from the presentation of data and user interface (note: user interfaces are not always technical) all the way down to the efficiency with which the data is stored in the back-end systems.
This broad scope, and deep reach, has the potential to create massive benefits for a company, and also the potential for a lot of internal politics and conflict. If the group is not well-positioned and the mandate of the board members absolutely clear, the result can be increased politics and lack of effectiveness in the broader organization. In order to be successful, a Data Science group needs independence, support, and wide authority to operate where and how needed.
A Data Science group is ideally a neutral and top-level part of the organization. This positioning will reduce structural alliances or preferences to any other group and maximise the changes for success. All parts of the company should be options for improvement as the company begins to develop a culture of using data properly for all aspects of the business. This means improving current products and services and also finding new revenue sources from existing data. In order to be effective, a Data Science team needs access to essentially all data the company has and make any suggestions about how to act, and this requires a mandate and support from the absolute highest levels of the company.
As with the creation of any new department or group within an organization, at any level, there is potential for causing the current political climate of the organization to deteriorate. This varies greatly by company, but is often compounded by the fact that a Data Science group may become not only the source of data products but also the best source of information about the company. Some people could seek to capitalize and exploit this, and such motives are something that the board and top executive should keep in mind. From an operational perspective, this data access and analysis capability can also devolve into a stream of neverending ad-hoc data requests, which has the potential to distract the team and keep it focused on daily operational work instead of more strategic data goals. This sort of reporting work is better accomplished by a BI department, or possibly a Finance and Controlling group.
The best solution is to have a top-level Chief Data Officer position, that perhaps reports directly to a senior board member. Without this, neutrality is automatically compromised since having the group under any other department implicitly makes it a tool of that department. This can be intentional due to politics and management, or unintentional and due to the simple fact that being embedded in one department makes people more aware of the problems faced by that particular department. Having a top-level and neutral role avoids intentional or unintentional favoritism and allows for the desired broad benefits to the company to materialize. The Chief Data Officer role is not a new one. It has existed for over a decade at places like Facebook, Yahoo, CitiGroup, and even in government as cities in the US like San Francisco, Chicago, Philadelphia, Baltimore, and New York all have Chief Data Officers. Even the US Army has a Chief Data Officer. Private companies, strangely enough, are lagging behind government entities in creating the role.
Aside from reporting directly to the Board, the next best option is for the Chief Data Officer to report directly to the CEO (in cases where other executives report to the CEO) or to the CTO/CIO (in cases where all executives are equivalent and report to a higher board). This still preserves some measure of neutrality, and allows the person to provide the organization with broad advice and results. There can be problems that arise in the case that there are separate executives for the Product and Technology groups. Since a Data Science team will need to work unhindered in both areas, there is potential for conflict.
It is possible for a Data Science group to be part of the Engineering or Product departments, as the group will naturally work very closely with both of those areas. Engineering regarding implementation of data products in production systems and Product in developing and prioritizing new data products, improving their knowledge of the customer, and collaborating on how data products fit into the overall product roadmap. This has the potential for the conflicts mentioned previously, but it’s still better than some other options.
This is the least-desirable situation. In this scenario the Data Science group is separated from the Product and Engineering teams, meaning they are too far away from the entire product development process. Additionally, it is often the case that the reason the Data Science group ends up inside of Sales or Marketing is because the head of Sales or Marketing is the biggest data champion in the company due to the fact they see they have a lot to gain from better knowledge and usage of their data. Things like identifying customers who are likely to churn, improving target lists for new sales, and so on are certainly within the realm of what a Data Science team should be doing, but those aren’t the only things they should be doing. The risk of having a Data Science team inside the Sales or Marketing group is that they become servants of only that group and the company loses out on the broader benefits of having a skilled Data Science team, like improvements to products, internal processes across departments, and so on.
Having a Data Science group is of course a great benefit to many companies, but where to place the group to maximize the benefits and chances of success is another matter. The most important requirement is of course that the group is as neutral as possible, and has as much freedom to operate as possible. This means ideally positioning the group directly under the Board of Directors, with additional options being directly under the sole executive (e.g., CEO) or another executive department (CIO/CTO). An acceptable but limiting arrangement could be positioning inside the Engineering or Product departments, with the least desirable outcome being positioning in an area like Sales or Marketing.