Key factors in deciding where to place your data organisation

Background 

The rise of data and analytics within corporate investment plans brings with it a number of key decisions which need to be made, such as how it should be funded, what measures of its success to take and which skills sets are required. Only rarely do companies start with the question, where should it sit in the organisation?

Yet there are good reasons for considering this issue sooner, rather than later in the development of D&A within a business. According to research carried out by E&Y, 81% of firms agree that data should be at the heart of the decision-making process. It is this idea which drives the adoption of D&A in the first place. 

But unless this resource sits in the right place, Its impact is likely to be lessened. As E&Y state in “Becoming an analytics-driven organisation to create value,” written in collaboration with Nimbus Ninety: “Without the right organisational structures, processes and governance frameworks in place, it is impossible to collect and analyse data from across the enterprise and deliver insight where it is most needed. This results in a siloed approach to big data deployment that limits a company’s ability to find, measure, create and protect value across diverse operational areas.”

Figure 1 shows the findings from this research and the limited extent to which organisational structure has been recognised as an important aspect of an investment project. Adjusting the business to embed D&A and putting in place an over-arching data strategy significantly lag behind the ambition to become data-driven. 

This article considers evidence for the impact which different organisational structures may have, the benefit of introducing a chief data officer, and how to mature a data and analytics function once it is up and running, regardless of where it sits.

Choose a starting point

Given the nature of technology and the need for data management in almost all industry sectors over the last few decades, it is rare for any organisation to lack some level of data and analytics capability completely. Even where data is recorded in Excel and simple reports are generated, super-users may represent pockets of skill around which D&A maturity could be developed. But they may also be the very reason why a new focus on data should not be influenced by existing practitioner.

Best practice, rather than existing practices, should inform decisions on the operating model for          the unit. In tandem with a new focus on empowering decision-makers with data should come two      key strategic plays:

1. Establishing clear data governance policies – as the DataIQ Leaders CARBON™ assessment identifies, a Level 5 (Optimised) organisation will have in place clear governance policies across all areas and be constantly assessing them. By contrast, at Level 1 (Aware), no such policies exist.
2. Establishing a unified data strategy – similarly, CARBON™ identifies a single, over-arching data strategy with common standards as being a characteristic of Level 5 (Optimised) and their absences as typical of Level 1 (Aware).

For these reasons, an organisation should establish an information council (or similar) if none already exists before undertaking a data and analytics investment. If one is already in place, then it should be used as a lever for the D&A evolution, not least as a key enabler of the new function.If an information council is outside of the scope of what the organisation can manage, then at a minimum data governance policies and a data strategy need to be drawn up, agreed at the top level of the business and a specific individual given ownership of them.

Introducing the chief data officer (CDO)

Evidence from across DataIQ Leaders members and the broader data and analytics community addressed by DataIQ shows that the introduction of a CDO is often the decisive first step on the road towards D&A maturity. Having a CDO is significant because they are able to act as a “door opener” for the use of data in the organisation and can look across functions. It is an essential role for any organisation considering artificial intelligence and machine learning. Such an appointment typically indicates several key things:

1. recognition by the organisation of the growing importance of data;
2. commitment to expanding the use of data and analytics;
3. acknowledgement of the need to apply standards, controls and governance over the use of data.

As has been well-documented, there are three “flavours” or stages of CDO which reflect the way this role develops:

• CDO v1.0 – all the data management responsibilities, including governance; fundamental to introduction of D&A.
• CDO v2.0 – supporting data applied to analytics; engaged with the business; potentially leading a centre of excellence.
• CDO v3.0 – combined data management and analytics; acts as data governor, enabler, filter and gatekeeper.

A handful of very advanced organisations are verging on having a CDO v4.0 who creates business out of data, for example by finding entirely new business opportunities or operating models through the application of data science. Achieving this requires a rare combination of skills, resources and especially a culture of innovation which is capable to absorbing outputs of this type.

There are risk factors associated with each of these. CDO v1.0 may arrive with a strong mandate around data governance, but it is hard to sustain investment into this role on this alone – without building out the value-creating activities of a CDO v2.0, the organisation may retreat from its data initiative. 

Equally, for a CDO v2.0, the risk is largely a political one when it comes to taking ownership of key data sets and MI processes, especially where these are viewed by other CxOs as part of their power base. 

If this is overcome and a CDO v3.0 emerges, the biggest risk is typically that of a change of CEO or superior CxO who decides to decentralise the data organisation.

Deciding on the CDO’s line of reporting

Who the CDO reports to sits at the very heart of choices about where data fits into the organisation. Just as appointing a CDO indicates a commitment to D&A, their line of reporting reveals just how deep that commitment goes.

Advocates for the creation of a CDO role in an organisation will argue that the line of reporting should be to the CEO. Evidence from DataIQ Leaders members, however, shows that not only is this rarely the case, but that CDOs have widely varying places in the organisation. Almost every key CxO position can be seen in Table 1. More than anything, this reflects the low state of maturity of data and analytics as a whole with no clear acceptance that the CDO should be an independent, high-ranking position. 

Within the CARBON™ assessment, the presence of a CDO who sits on the board is considered to be Level 5 (Optimised). However, it should be noted that no DataIQ Leader members have achieved this to date. A small number have a CDO who sits one level below the C-suite, but more typical are organisations where the CDO is two levels down (Level 3 – Defined).

Creating a data organisation: local, federated or central?

Just as the line of reporting for CDOs varies widely, so does the position of data and analytics teams within the organisational structure. In research among DataIQ Leaders members, all models of organisational arrangement were found, from fully-centralised centres of excellence to decentralised function-based operations and hybrids of the two where shared activities have been centralised, but line of business-specific services remain within each LOB.

For smaller organisations, local D&A may be a more feasible option, at least initially, as sponsorship and funding are easier to gain from a function which will experience direct benefits. CRM and marketing are the commonest places to find local teams, although finance and business intelligence-based examples have also been found.

The benefits of centralisation typically flow from the ability to share best practice and peer review activities before outputs are given to internal customers, as well as building an effective culture and political base. Risks can be from the visibility of a large overhead which may get targeted during financial streamlining or exposure to plans to decentralise, which is a cyclical initiative experienced by many organisation. One DataIQ Leaders member has created three DACE organisations within different companies across three different sectors, only to see two of them broken down into functional-level units. 

The benefits of a federated model are being able to draw on multiple sponsors and sources of funding, as well as the ability to disperse specialist practitioners into lines of business when demand is high or centralise them when demand is inconsistent. Risks are that a federated D&A operation may lack a sense of coherence or identity, causing problems with recruitment and retention.

The benefits of a local model are the direct impact which it will have on the specific function it supports, which may in turn become a champion for the value of D&A for the organisation (which may ultimately see it federated or centralised). Risks are around maintaining commitment within a small group of practitioners, as well as keeping skills and knowledge up-to-date.

Although focused specifically on the creation of a data science team and whether it should standalone or be embedded, this short video offers a useful discussion of the issue

Moving the data organisation up (or down) the maturity curve

Data and analytics are so new within organisations that their place in the structure is far from fixed. Unlike well-established functions like IT, HR, marketing and sales which can be found in virtually all types and scales of organisation, D&A is more likely to be absent than present. And where the function has been introduced, it may operate as a sub-unit within a master department (as seen above with the reporting line of the CDO).

As the use of this resource matures, the organisation may decide to formalise it as a standalone function. This does not automatically mean full centralisation – a federated model can reflect this decision if it includes promoting the CDO up one level.

During organisational transformations in the near future, it is entirely possible that a centralised DACE will be broken up and dispersed, often as a result of a desire to empower key functions with more localised support (although the decision can be as much political as practical). Historically, this trend can be seen in the rise and fall of business intelligence which used to be clearly identifiable as a function in its own right, but in many cases has been supplanted by self-service reporting created by D&A. In the future, automation of processes may have the same consequences for this function in turn.