Data and analytics: The fastest learning organisation wins | Deloitte UK has been saved
Limited functionality available
The pace of change will never be slower than it is today. It’s accelerating. For organisations to survive and thrive, there is a need to optimise for fast learning and to act on this learning. Continuously. Maintaining the status quo is standing still on the starting line, while more nimble competitors race ahead.
Data and analytics have a fundamental and increasingly important role to play in the new normal where speed to learning is key. It is the feedback loop. Like the dials on a car dashboard, data and analytics provide insight (e.g. speed, distance covered, fuel range) into the actions that people take every day, to achieve desired outcomes. This data and analytics feedback loop enables business agility. This feedback loop itself should be continuously delivered with agility. It enables an organisation to experiment, learn and pivot, to deliver better value, sooner, safer and happier.
The age of digital
We have passed the tipping point in the latest technology-led revolution. We reached the Age of Digital where work is increasingly unique and emergent. Unlike building an identical car every 60 seconds in an automotive factory, you don’t write the same software 100,000 times. You write it once, rewrite it a few times to improve it and a computer runs it 100,000 times. Also, an organisation is a Complex Adaptive System (CAS). A CAS is emergent and it is not possible to accurately predict cause and effect.
As the future is not predictable, as we don’t know what we don’t know, and as we can’t currently time travel, if we want the best possible outcomes, we need to optimise for learning and speed to learning. We want the cheapest cost of intelligent failure. Learning early and often enables work to be de-risked, it avoids ‘sunk cost fallacy’, it enables earlier value realisation and it enables business agility.
The importance of data and analytics in fast learning
Data and analytics are essential to be able to learn fast and act fast, and the bigger the gap from action to measurable reaction, the harder it is to determine causality. Data leads to Information (the what), it leads to Knowledge (judgement), it leads to hypothesised Understanding (the why), which enables a next action to be determined.
Therefore, the provision of data-led insights should itself also be optimised for speed to learning. Consider the irony of Data, Analytics or AI initiatives, which are enablers for business agility, being implemented in a ‘Think Big, Start Big, Learn Slow’ waterfall approach. In this traditional waterfall approach, there is big upfront planning and solution design at the point of knowing the least. The predictive plan has a focus on a predetermined solution and learning comes in at the last minute, when there is the least time to respond to the learning. The future, for unique emergent work, is treated as if it is predictable and deterministic. The focus is on the plan and the predetermined output, rather than on the desired outcome.
The reason that so many traditional waterfall projects are late or over budget is because it is a fundamental thinking error to approach emergent work in a deterministic manner. There will be unknown-unknowns. Things that people don’t know they don’t know. Rather than using agility as a competitive advantage, motivation and morale decrease as people work harder to deliver to the fixed plan and avoid a dreaded red RAG status, in the worst cases with a culture of fear.
For example, at one large organisation, the feedback was that every request from the business for data insights led to a lengthy project plan with a significant price tag. The data insights being requested were hypotheses themselves, with no way of knowing if the insights will prove to be useful or not until seen. With an inability to experiment, to test hypotheses, to optimise for the cheapest cost of learning, the business self-solved, building a shadow data function. Yes, they got quick access to data to drive timely decision making, but there was a lack of consistency across the organisation, a duplication of effort and complications around data governance and compliance to data and technical standards.
Clearly, there is a need to strike a balance, to aim for overall consistency with local optimisation. There is a need for both speed and control so that it is agile, not fragile. This enables quick decision making, the ability to pivot and ultimately, the fast flow of safe value.
If you have a data and analytics ‘transformation programme’ or large data initiatives, to optimise for speed to learning, approach them with agility.
Think big, Start small, Learn fast
Fall in love with the outcome hypothesis, not the solution. Getting to the outcome, if at all, will require experimentation.
First ‘Think Big’. Focus on the outcome hypothesis over the output. This is the North Star. For example, an outcome hypothesis might be that near real-time data analytics on product sales, pricing and customer satisfaction, with the ability to look at different segments, will enable faster and more insightful decision making, which will increase market share and profitability. The hypothesis is in business language and is measurable.
There will be several possible solutions, with no one-size-fits-all, depending on starting point and context. There is a short period of ‘just enough’ envisaging for high alignment and articulation of hypothesised value (which will be tested early and often and will continue to iterate). At this point, an outcome roadmap would be put together if it doesn’t already exist, especially if there any regulatory commitments.
Approaching work with agility does not mean that there is no planning. Quite the opposite, there is more planning, and it is continuous. And the roadmap is a roadmap of quarterly outcome hypotheses, expecting a wiggly path to optimise achieving them, based on learning. A traditional project plan plots a straight line from A to B, where B is a solution rather than an outcome, with late learning. When finally arriving at B it might turn out that B doesn’t optimally achieve the desired outcome.
For emergent work, it is worth maintaining optionality to the last responsible moment, to maximise value, time to value and for the cheapest cost of learning. Be wary of ‘Water-Scrum-Fall’. A deterministic project plan, with Big Up-Front Planning and Design, the word Sprint ten times in the middle and a Big Bang release, is not agile. It does not optimise for early and often learning and it does not realise value iteratively.
‘Start Small’, is focussed on the desired outcome hypothesis. Here you optimise for speed to learning, for the cheapest cost of intelligent-failure. Ask yourself the question, how many times has this been done before, in this exact context, with these people, at this time, with this data and these processes? Even Commercial off the shelf software is unlikely to have been installed a hundred times before in your exact context. Hence optimising for agility, for learning, for de-risking, is key, whether built or bought. Leverage others’ learning and lock in progress early and often, this will de-risk delivery and lead to an earlier value. It might turn out that a hypothesis was incorrect and the goal is to come to that conclusion as cheaply and as quickly as possible.
Have a multidisciplinary team, with people from a range of roles in ‘our business’ (it’s not ‘the business’ like it’s a separate entity). Have a team with people representing the customer, strategy, product, data science, engineering and so on. Each person is ‘T-shaped’, a specialist and a generalist, willing to step in to help others.
Run safe to learn experiments. Optimise for speed to learning. Data scientists may not know what algorithms are of most value and colleagues in customer-facing roles may not know the art of the possible, not knowing what to ask for. Don’t pass work between roles, with hand-offs. Instead, people go to the product, to the value, working on it together as a team, with all the skills needed. The primary personal identity is to the value being produced (e.g. ‘I work in the retail business’) rather than a job role.
The Learn Fast may also benefit from a federated approach, rather than fully centralised or fully decentralised. Instead, have a small central Data Analytics Centre of Excellence (CoE) and federated data analytics teams in the business lines. The CoE sets the guiding principles and the standards, ensuring that there are minimal viable guardrails. The CoE ensures that there is a wide road with high kerbs. This enables the federated teams to innovate and optimise for speed to learning, with right-sized governance, depending on risk profile and risk appetite. The federated teams can optimise to each business line’s value proposition and context. They enable the flow of value, as there is not one central team which can act as a bottleneck. It is a scalable pattern. Done well, there is global minimal viable consistency combined with local optimisation. The CoE also advances the craft, with oversight of innovative approaches and tools, as well as skill-based personal development of practitioners across the federated teams.
Repeat, repeat, repeat!
Repeat, taking an ‘S-curve’ approach. Amplify the experiments that work and dampen the ones that don’t. As progress is made towards the outcome hypotheses, don’t go from a pilot to a Big Bang (a deterministic mindset). Instead, gradually increase the rate of adoption or the scope, continuing to optimise for learning. Eventually, the 80/20 rule is reached, which is the view that the final 20% of value will take 80% of the effort. The ‘S-curve’ is starting to flatten out. At this point, make a judgement call. Perhaps it’s time to cease working on that desired outcome and consider pulling the next outcome on the backlog, which might generate more value, sooner.
To recap, data and analytics is the feedback loop which enables business agility. Therefore, to enable business agility, as the work is emergent and as organisations are emergent, approach the provision of the feedback loop with agility. Optimise for early and often learning and delivering slices of value, to try to prove an outcome hypothesis wrong as quickly and as cheaply as possible. Think Big, Start Small, Learn Fast.
Jon Smart leads Business Agility in the United Kingdom, helping clients deliver the outcomes “Better Value Sooner, Safer, Happier” through an emphasis on flow and the application of principles over practices and tools. He has more than 25 years of experience helping teams develop agility in contexts of change, work, and scaling up. Smart is the founder of the Enterprise Agility Leaders Network, a member of the Disciplined Agile Advisory Council, Business Agility Institute Advisory Council and Programming Committee for the DevOps Enterprise Summit, and a guest speaker at London Business School.