70% of data science projects fail, and therefore generate zero business value. Most of the time, the reasons for failure are not technical.
Lack of strategic alignment with business, non-fluid links between business and technical departments, immature industrialization processes, lack of skills or loss of key data scientists, unclear data governance rules, poor comprehension of regulatory rules, or inefficient project steering are among those reasons. An agile organization, combined with efficient governance, is required to solve those issues.
There is no one size fits all organizational design: whether a business utilizes centralized data teams or data skills are distributed across the business highly depends on where they stand on their data journey.
The north star must be how to most efficiently help every single business operation within the organization.
Our approach is to start with a systematic maturity assessment, covering three pillars: business operations, the data and technology stack, and people.
We distinguish between two different kinds of business operations.
Firstly, those which already utilize data as a business asset such as BI (Business Intelligence) reporting, ROI measurement, and operations which leverage data for some business outcome or have already implemented data-driven decision making processes. For these mature functions, our approach is to build efficient processes focused on two areas: business acceleration, and industrialisation.
On the other hand, operations with lower maturity require more framing, new methodologies, and experienced data scientists to initiate new workstreams. For these operations, our proposed structure ensures the right priorities are set from the start, avoiding losses from endless proofs of concept.
A maturity assessment will also reveal whether data science capabilities need to be industrialized and, if so, whether a centralized team should be in charge of paving the way to turn data into a service.
This process may include: designing and implementing the architecture of data systems (databases & datalake(s) and data computation tools), maintaining processes and rules to homogenize datasets and the way they are populated, managing rights and accesses, or ensuring compliance and privacy. These functions require a clear understanding of the type of data a business needs to manipulate, and what it will be used for.
We install a holistic governance approach that creates operational fluidity between those central functions and the business use cases stakeholders, with principles such as sharing rituals and transversal KPIs upheld thanks to a data-oriented collaborative tool. For example, YOOI, an affiliate of Ekimetrics, developed such a tool which allowed them to break down silos and deliver their data roadmap.
Regarding people, we can create the roadmap to build a 3-to-5 year data talent management program, centered around the principle of a data community. In time, this data community will spread across the entire organization. We suggest managing this as a “virtual team” through a gaming tool which makes everyone’s expertise and skills tangible, clear and accessible to others.
This structure will allow long term and transversal career development paths that assign “grades of contribution” within the community, giving people increasingly exciting challenges within the same company. Dedicated mobility programs will leverage the company’s diversity, it’s training capabilities, and it’s openness to and engagement with the external data science community.
Setting up an efficient organization with Ekimetrics will give you four distinct takeaways: the organizational structure, the roles and rules, the tools, the recruitment & adoption plans.
Practically, shaping organizational structure requires clearly defining the functions required to run a data science based program, and assigning them as much as possible to existing functions in the organization. Dedicated functions will be created only if needed: data must not be seen as something additional, but a value-adding structure integrated into existing ways of working.
Crucial functions include: strategic alignment of business use cases, use case sponsorship, business adoption, data transformation program steering, data capabilities roadmap management, data governance management, regulatory, security and legal compliancy, data people career development, and data ecosystem openness.
There will be several types of rules and roles: data management (which data is collected, cleaned, converted in respect of regulatory and technical constraints, and stored, who/what can access it), business processes (how the processes are re-written with data insights, how to manage exceptions and dysfunction, etc.), investment and ROI steering (decision making processes, arbitrations KPIs), capabilities development (how to identify strategic assets, dedicate budget for their development, etc.) and career development processes (mobility rules, community management, awards, etc.).
Once the organizational chart is set, the roles and rules defined, and job descriptions written, a recruitment plan is to be set up, with a specific skill set and background profile articulated for each function, that can be made available to internal applicants. Ekimetrics can also help with sourcing people through our network.
Finally, a change and adoption plan will follow, to explain to/train/endorse targeted people, and ensure people work smoothly together. Even the best organization design will not work without people support. The adoption plan is broken down into a timeline of individual, collective or global sessions which develop gradually from awareness raising to specific training.
Thank you for your interest in Ekimetrics. Please fill out the form to ask your question.