When theorists build theory, they design, construct and interpret imaginary experiments. In doing so, their activities resemble the three processes of evolution: variation, selection and retention.
Karl E. Weick
An effective strategist cultivates a capability to imagine a range of possible future states. Some of these states will require an extended act of imagination. A few of these more detailed imagined states will emerge as desirable target states.
We want to know more about the feasibility and viability of these few target states. To obtain reliable, reproducible or extensible results for this next stage of exploration, these desired imaginary states need to be subject to conceptual discipline.
It is through this discipline that the strategist converts their imagination into a theory of change. The specific form of the theory of change isn't a trivial matter: it will shape everything that follows. Any claim attached to the target future state will only be as good as the evidence generated through the discipline embedded for within theory of change.
The various lean start-up and agile methodologies provide much of this discipline. Some methodologies will be better than others, depending on the problem at hand. However, they are all sympathetic to working within the scientific method. This is because a theory of change produces testable (and falsifiable) propositions and testing is done via phases of experimentation.
The lean start-up and agile methodologies are ideal for proving/disproving the theory of change because their core mechanism is successive testing. This in turn enables modern data governance because the data output of sprint-based experimentation is controllable via the lean-start up and agile artifacts.
Embedding data governance by design within lean start up and agile methodologies produces data that is already partially de-risked. Innovation funding demands high returns and, because innovation is fundamentally uncertain, de-risked data reduces opportunity loss across the data product portfolio.
Controlling for instrument effects is one of the early benefits of locating data governed lean start-up and agile within a theory of change. We introduce instrument effects when we change the parameters of an experiment mid-phase. It is because changes are rarely introduced during the sprint time frame that makes lean start-up and agile effective controls.
Locating data governed lean start-up and agile within a theory of change also provides the basis for continuous improvement. A controlled experiment produces calibrated results, which reduces uncertainty and enables increasingly accurate estimates. These estimates can be quantised in dashboards to indicate the rate and direction of learning for decision making.
Better informing the next round of lean start-up experiments and agile sprints is another benefit from building in data governance. The entire existence of these methodologies is to move from conceptual model to live service delivery. The quality of their output is, as we well know, dependent on the quality of their input.
This becomes relevant when seeking to reduce the misfire and restart rate of lean start-up and agile projects. The data output of lean start-up experiments and agile provides us with the breadcrumb trail we need to improve the theory of change and get closer to establishing a meaningful, valued and sustainable business service.
A fifth reason for adopting a specific form for the theory of change is to ensure future usability of our output data in enterprise analytics projects. Because the value of data follows compounding behavior, future work will enjoy higher quality inputs. Better answers to previous questions gives us better questions for future answers.
Locating data governed lean start-up and agile within a theory of change gives us a continuing series of performance data of known quality that we can test for predictive fitness. With partially de-risked output data, we can complete the de-risking process at a later date and get ready for enrichment. This is how we get socially permissioned machine learning. It is also how we fulfill the enterprise analytics mission: produce facts and manage truth claims.
'Right and wrong are situational. In the appropriate situation, nothing is wrong. Without the appropriate situation, nothing is right'. Huainanzi