They say that wisdom comes with experience, and I’d have to agree with that. Having spent the past two years building an analytics team with the primary aim of having impact in the business, it’s only now that I feel I can say something about how to do that.
One of the reasons why it has taken me some time to form a point of view is that we have experimented with a few different models and structures in order to find out which ones work most effectively. Another reason is that no single individual has the answer, and it has taken time for our model to develop through a consultative approach with key members of the team and important stakeholders in the business.
That said, the model around which an analytics team should operate within a business or organization has now formed more solidly in my mind. It has six steps which form a ‘lifecycle’ around business decision-making. These six steps inform how the team should operate, what skills are required, how the team should be structured and what types of profiles and skills should be present on the team. I realize that many of those who read this article will not necessarily have access to the resources required to fully comply with this model, often because the organization they serve is too small to justify it, but nevertheless I believe the overall picture is still of interest.
Step 1: Understand the decision-making needs of the business
Why? One of the common reasons why analytics teams do not function optimally is a lack of understanding of the decision-making that occurs in the business they serve. If an analytics team is overly staffed with technically or academically minded individuals, this ‘conceptual gap’ is highly likely to affect the potential for impact of the team, because analytics will often be driven by the personal interests of those on the team and not against the requirements of the key internal clients.
How? At least one of the senior members of the analytics team should be closely connected to the decision-making organs of the business, be it the CFO, CHRO or the other committees or groups which are tasked with making the critical business decisions. Regular forums should be set up where the key priorities of the business are communicated as well as feedback on where analytic intelligence is lacking. This will allow the formation of an analytics agenda, and the building of data structures and tools which will best serve the future needs of key internal clients. The individual responsible for this should ideally be sourced from the business itself, be analytically and strategically minded (though not necessarily technically skilled), and have a passion and drive to develop analytics and analytic capabilities within the business.
Step 2: Construct measures to support business decision-making
Why? My regular readers will likely be tired of hearing this from me, but analytics success depends on effective measurement. Many analytics teams struggle today because they are not accessing accurate measurement data. Upon understanding the decision-making needs of the business, the first question to be asked should be ‘What would we need to measure to understand this?’. Measurement can be a fiendishly difficult problem. Determining the appropriate measure is a delicate balance of mathematical and statistical principles set alongside considerations of data systems and data capture processes as well as human behaviors. It’s often a compromise, but it needs to have strong expertise and judgment to get it right.
How? Determining measurement approaches to address a particular business problem requires participation from many skill sets on the analytics team. A measurement expert is an absolute necessity. For example, this could be a psychometrician if it is a people and skills problem, or a marketing metrics expert if it is a sales or marketing problem. Alongside the measurement expert, input is needed from a data expert, usually an engineer who understands the transactional systems and the flow of data, an analytics professional such as a data scientist who understands how to work with data in conducting analytics, and finally the individual from Step 1 who will be able to properly explain and translate the decision making needs of the business to the technical experts.
Step 3: Capture data in transactional systems to allow tracking of measures
Why? If the measures that arise from step two are new, then more often than not it will require the entry of a new type of data at an atomic level within the organization. This will require implementation as a systems level, as well as an understanding of the rules and logic required and the adjustments needed to human processes to ensure the data is captured accurately.
How? The data engineer is a foundational role in any analytics team. They will act as the key liaison with the system administrators and experts, and they need to have an important say on how the fields and entry rules are constructed within transactional systems to allow for accurate and reliable data flows. This is not their only critical role (see next step).
Step 4: Engineer data for regular reporting and analytics of measures
Why? The process needed to translate transactional business data into measures that are useful for analysis and decision making can be laborious. Some careful thought around this can make a huge difference to the efficiency and reliability of an analytics operation. Can the transactional data be pre-processed on a regular basis (hourly, daily, weekly, monthly) to create tables or views that are aggregated at a level allow for more rapid analysis and insight? How should these views be designed? What demographics or cuts should be available?
How? A capable data engineer can work wonders in both understanding the needs for pre-processing of data and in actually creating and designing the required aggregated data sources.
Step 5: Conduct analysis to address business questions
Why? This part is obvious.
How? This deserves a fuller treatment which I will focus on in a future article. Clear lines needs to be drawn between
- Regular standardized reporting, which requires data science, data engineering, visualization, UX and software development skills in data automation and provision. Ideally, almost no team time should be spent on delivering regular reporting manually. External vendor products may help plug gaps here, but my experience is that no vendor tool satisfies the entire plethora of needs here. The best teams will design this for themselves.
- Ad hoc analytics, which require business intelligence professionals who understand the data systems of the organization and can use them to satisfy the specific, non-standard questions which come in from business customers.
- Advanced analytics, which requires deep knowledge of advanced statistical methods and fluency with advanced data science tools to allow data to be analyzed and processed using these advanced methods,
Step 6: Translate analysis for the business consumer
Why? One of the main reasons why analytics does not have impact in business and organization is that the results are not well understood, and often conclusions are drawn which are clearly wrong due to the lack of effective communication of results. To have impact in an organization, an analytics team needs to be able to draw on skills that facilitate clear translation of the results in a way that can be understood by business leaders, many of whom may not have the knowledge or skills to translate for themselves. This is particularly important when advanced methods of analysis are being employed.
How? The translator is a critical role in an effective analytics team. Translators have an understanding of organizational strategy and decision-making, strong general problem solving skills, a passion for and an interest in analytics and a client-service mindset. Analytics translators are currently the most difficult profiles to find because the role is so new and not well understood. In my experience, the most effective and able translators are sourced from the business itself. Translators provide overall leadership and direction and work with the technical skillsets (data science, measurement, engineering, visualization and design) to find the best possible solution to the client’s needs. Translators maintain long term client relationships within the business to allow readjustment of the analytics approach as needs change.
This is my first articulation of the analytics value lifecycle, and no doubt it will develop further, so please consider it a work in progress. Over the coming weeks and months, I also intend to spend a bit more time fleshing out the specifics of some parts of the lifecycle and the roles and profiles of individuals needed to effectively staff an operation like this.