“How do we start an analytics practise?” – I have often been asked this question. While there is no one size fit all solution to it and various organizations have achieved success through various models. We have come up with a comprehensive 5 step framework to seamlessly get started with an analytics function, that mitigates risks and exemplify analytics output. These steps could be taken up internally by businesses, though we recommend employing external experts that bring in a third eye perspective to your business.
Even before you acquire skillsets and set aside your yearly budget for analytics, know what is needed. Get an assessment of your current data landscape and key use cases that are helpful for your business success. This is essential to identify appropriate sponsorship and acquire right skills.
Select projects that are not only interesting and valuable but also small enough to quickly capture value from analytics. Your early use cases need not be those bulky implementations that needs collaboration from various functions in your organization, or are too complex to be understood by business.
Most data science initiatives fail not because of lack of skillsets or use cases, but due to lack of right management focus on it. This is primarily confronted when analytics insights fail to move from lab to actual implementation.
Over 100,000 people subscribe to our newsletter.
See stories of Analytics and AI in your inbox.
Most of the times, business managers lack right focus and knowledge to incorporate data science insights into their operational strategies. Implementation of analytics is mostly seen as a recruiting bunch of data scientists that churn out interesting insights and visualizations. Yet, success in analytics is more than that. Most importantly, it is the knitting of analytics in the cultural fabric of the organizations, where there is a seamless exchange of ideas and insights.
On-board Talent & Tools
This seems too obvious, but it’s difficult in practice. Data science is a rare skill to find in market currently. This is mostly due to fact that it’s a combination of multiple interdisciplinary subskills, each of which contributes equally to the overall skillset – knowledge of business processes, data, statistics, visualization and design thinking, scientific and research methodologies, software programming and above all a consulting bend of mind.
Organizations for long have looked at these multiple sub-skills discreetly and devised their analytics hiring strategy around each of these. So, organizations hired visualization experts separate from programmers or number crunchers. This works well to quickly ramp up large teams but in long run creates operational bottlenecks and increase investments on key deliverables.
A much contemporary style of hiring data scientists is to look at professionals that loosely touch upon all (or combinations) of these sub-skills and then train them to specific needs of the organization.
Formalize Governance Process
Continuous success out of analytics practise involves putting up a centralized governance group. Most organizations are too focussed on deliverables, leading to gradual degradation of vision and quality over time.
Governance involve setting up an internal standards and controls policy. Call it internal regulation, which tracks quality of deliverables and their transformation to business value.
Besides, governance also involves implementing accountability within data science teams. Most data science projects are result of complex collaborations between different stakeholders, leading to a confusion on who actually owns which part of the asset.
Continuous & Inclusive Training
Investment into training and upgrading overall skillsets within the team is well recognized by management. Yet more importantly, the two key themes here are ‘Continuous’ and ‘Inclusive’.
Analytics training should start as soon as you decide to embark upon analytics journey. Training should not be seen as a one-time activity that is taken up once every quarter. The mode of training (internal or external) is non-essential when you start. What is of essence is that a continuous focus from management is required to upgrade the data science skillsets with the organization.