The optimal organisational model for data and analytics requires balancing centralised and decentralised resources that collaborate with lines of business to deliver against business objectives.
Traditionally, organisations have based their data and analytics efforts around a centralised, IT-centric team. These efforts emphasise governance and consistency across the organisation.
Today, however, an increasing proportion of data and analytics initiatives originate from within departments and lines of business. In order to craft the right organisational model for data and analytics, the tension between ‘centralised control’ and ‘decentralised agility’ must be balanced:
- Centralised – Control, consistency, consensus, share best practice
- Decentralised – Freedom, autonomy, agility, innovation
The question is, how can we deliver the best of both to ensure that each project delivers its desired outcome in an efficient manner?
It is possible to create a model that strikes a balance between these two approaches; a distributed hybrid organisational model consists of a core centralised team — an analytics centre of excellence (COE) — that coordinates with multiple decentralised teams in a cross-functional approach.
Empowering Local Teams
Much of the trend toward self-service data and analytics over the past decade has been focused on empowering individuals. However, more and more frequently organisations are coming to the realisation that data and analytics is a team sport. That’s because, as the data and analytics field grows and evolves, it’s virtually impossible to find one individual with all the necessary skills to deliver end-to-end. As such, in order to ensure success with your data and analytics approach, aim to establish and empower cross-functional teams across the organisation with a mix of the following skills:
- Technical: Data engineering skills for building, managing and operationalising data pipelines
- Quantitative: Data science and data analysis
- Domain understanding: Including customer-facing teams (for the external point of view) and business-facing teams for the internal point of view
The traditional analytics operating model was inherited from big ERP initiatives that took the approach of establishing a single
Although an organisation may have a strong centralised data and analytics capability, the proliferation of data and analytics use cases means decentralised analysis also occurs at various levels throughout the organisation.
Business domains have justified, more specific analytical needs and generally use their own self-service analytical tools. Unfortunately, this approach has its own issues when it comes to sharing data across the organisation. Decentralised teams placed in business areas are essential to delivering data and analytics with the domain expertise and responsiveness necessary to meet demand for a given business area. However, most organisations do not adequately distribute responsibility for the full set of shared work and competencies.
The data and analytics operating model should make it clear who will do what as part of the allocation of work between the distributed team and the centre, and this should be clearly articulated for each project. The model should be specific about whether or not the decentralised teams are empowered to create prototypes, pilots or even full-production analytic content:
When decentralised teams develop their prototypes, the central data and analytics team often views this as a risk. The autonomy of business units to build their own, local “contextual version of the truth” may lead to potential conflict with the “single version of the truth” promoted by the central team. This can result in hesitation to move the prototype into production, and innovation remains untapped.
To avoid this challenge, organisations should allow autonomy for the decentralised teams, with each user persona granted the appropriate permissions, access rights and privileges to develop their own analytical prototypes.
For pilots, they can shift toward a more adaptive form of governance, where resources from the central team support content creation based on best practices and common requirements. After the success of the pilot program has been established, it can be readied for production with support of a centralised team if required before being handed over to end users.
The centralised elements should be used primarily to set standards and control governance across the whole organisation (to be used for any decentralised analytics activities). Some examples that drive standards and processes are:
- Documentation of processes, for example:
- Requirements gathering
- Testing – e.g. Test scripts
- QA (Quality Assurance)
- UAT (User Acceptance Testing)
- Roll out/ go-live/ production
- Templates for data visualisation including corporate design standards
- Approved training and learning materials
- Certified data sets
While decentralised, cross-functional data and analytics teams can use business domain expertise to drive innovation and results, it’s still imperative to maintain enterprise-wide standards of governance and best practice. As such, coupling these decentralised units into a hybrid organisational structure with a centralised data and analytics team is the best way to ensure consistency and stability and, ultimately, success in your data and analytics strategy.
This article is part of our Data-Driven SMB series. For more information, advice and resources on how to accelerate your organisation’s data and analytics maturity, click here or contact us today.
Senior Consultant With over 5 years’ experience in providing analytics business solutions for clients, primarily across Public Sector, Healthcare and FMCG organisations. Ciaran holds a MEng in Mechanical Engineering and specialises in analytics enablement using platforms such as Tableau, Power BI and Alteryx.