The 5 Gaps to being Data Driven
A tried and trusted approach
Business Problem
Let’s start with the business problem. You want to be data driven, you believe it will help deliver value but you don’t know where to start or want to review progress achieved to date. Once you have the vision to be data driven you need to assess how ready you are, and architect what you have to do to get there.
Gap 1: Measurement
Being data led means you measure your performance and understand how to improve your performance. Franklin Covey in his book “The Four Pillars of Execution” talks about two types of metrics:
1. Lag measures describe results — what you are trying to achieve (revenue, profit, market share)
2. Lead measures describe activities or sub-goals which can be acted upon. A good lead measure predicts success on the lag measure, and is something which you have control over influencing.
Eric Ries in the “Learn Startup” refers to these as vanity metrics and actionable metrics.
As an example, for an E-commerce business, conversion would be considered a vanity or lag metric. Knowing what caused conversion to go up or down would be metrics such as number of products on site out of stock, or number of products on site more expensive than competitors. Metrics which would clearly influence consumer purchase but also metrics which are within your team’s ability to take action to improve.
Michael Ross, founder of Dynamic Action takes this one step further: Data driven businesses are able to measure processes and understand why a process has failed, they measure that failure and assign ownership for ensuring the measure is improved.
Continuous thinking about how you measure business performance and creating new metrics to further understand business performance is a fundamental part of having an offensive data strategy. This strategy is more relevant for customer-focused business functions such as sales, marketing and operations.
Gap 2: Business Ownership
Whilst management might be keen to champion data, there is an inconvenient truth that they’ll be a potential blocker if they believe the above performance measurement approach might uncover poor performance. This is why it’s essential for your data vision is to have Board Level Sponsorship.
The Sponsor’s first goal is to ensure that the organisational mindset is open to transparent performance measurement, creating a culture in which this mindset is rewarded and actions to hinder this will not be tolerated.
The Sponsor’s second goal is to communicate the objectives and key results (OKRs) for a Business Owner.
Objective:
As a business team/department leader you adopt transparent, measurable & actionable performance measurement into your ways of working.
Key results:
1. You are able to articulate clear business performance problems which you want measured
2. You establish working groups to design & experiment with new metrics to measure performance
3. You assign owners to metrics
4. You take ownership of data quality ensuring business systems capture the data needed
5. You take ownership of a data & analytics project failures as well as its success.
The Sponsor’s third goal is to ensure the organisation is building its data literacy capability. The onus of data literacy shouldn’t be just on the analytics professionals an organisation hires, but also on the consumers of analytics across the organisational departments. Data literacy simply means being able to read, analyse, communicate and make decisions on data.
Things on the syllabus should be: different types of analytics, basic stats, how to interpret graphs and experimental design.
The Sponsor’s fourth goal is to provide focus. The chances are the list of use cases within the organisation to be more analytically-led is long. They all can’t be done at once — narrow the focus to a few initiatives that balance value versus feasibility. Build success and then ambition.
Gap 3: Data & Technology
Technology is a fantastic enabler, this article is deliberately not going to focus on this as much due to their there being lots of good content out there on data tech. Here are some key points when adopting an offensive data strategy:
1. Encourage data engineers and data architects to get closer to the analytical and business users. These individuals are brilliant and the more they know about what data gets captured, when, how and who does what with it the better they’ll be able to transform and model it.
2. A build on point 1, arrange your technical teams around delivering business outcomes / value streams. This is about building organisational maturity so that cross functional teams are working to deliver without dependencies on others.
3. Data is the new oil but only if it is well modelled & governed. Don’t under-value Data Architects. High performing Data Architects with clear requirements from step 1 are key to your data driven maturity.
4. Ensure governance around purchasing technology. Ensure business stakeholders aren’t being promised a silver bullet by Tech sales people. Ensure tech purchases are being driven by use case requirements — gap 1. Maintain data & tech spend appropriate to the 5th V of Big Data — value. You can scale your tech spend as you build success.
Gap 4: Analytics
Analytics Maturity Models are wrong. If you think descriptive and diagnostic analytics doesn’t have as much business value as advanced analytics then you are not doing it right, go back to Gap 1.
Although analytics terminology is becoming more mainstream, it’s still not consumer friendly. To sell analytics successfully to your business stakeholders you need to market like any other product, using simple language that resonates.
Think of analytics as a product. Who is the customer of that product, how do they want to consume the product and when. There are three types of analytical products:
Performance Intelligence. You understand how your business performing and why. You are monitoring the lead measures. This is known as descriptive & diagnostic analytics and it is business intelligence dashboard on desktop or mobile app. It provides the foundation for identifying and prioritising your analytics roadmap.
Action Intelligence. You are able to use data to take action in the present to improve performance. This is known as prescriptive analytics. This is typically consumed real-time such as an alert for someone to take action, triggered by a poor performing metric. Good design of this product will include a feedback loop to understand if alerts are correct or “false positives.”
Decision Intelligence. You are able to use data to automate decision making. This could be either human in the loop or human out of the loop. These are your ML & AI use cases, processing inputs into outputs. Authors note: this is a discipline best articulated by its founder Cassie Kozyrkov — Head of Decision Intelligence at Google.
Analytical maturity models are also wrong in presenting maturity in a linear flow. Different parts of an organisation can be engaged in the above three analytical products at any one time.
Gap 5: Execution
Up until this point, everything has been a cost. You deliver value from your data & analytics endeavours when it makes the business change processes and decisions.
For an analytical product to be a success you need:
1. An execution group the operational team who is responsible for driving analytical product into the business process. A classic change team whose activities are communicating the change, training users, and feeding back on product performance.
2. A steering group the governance group accountable for ensuring the analytical product is adopted by the business. This group meets regularly to monitor the adoption and impact of the analytical product through the use of a scoreboard, holding team members to account, clearing the path ahead.
3. The business change will involve trialling new processes, whether that’s new manual operational process, changes to a website/app via multivariate tests or, or automated decisions. You’ll need your analytics team on hand to measure these changes.
Finally you’ll also need the execution and steering groups to be brave & committed to trialling, failing and iterating.
Thanks for reading!
