In Kanban we often talk about flowing value through the system and yet that’s somewhat misleading. The reality is that we can’t know whether a work item is valuable until we’ve actually finished the work and made it available to our customers.
The best we can do is optimize for learning whether what we’re working on is valuable more quickly and cheaply1. We have a hypothesis that the work item is valuable (an experiment) and so we’ve prioritized it and are working on it with the expectation that it will be valuable when we finish.
Even when we’ve finished, we don’t know if there’s value unless we take some measurements or observations around how it’s being used. We may have just built the most amazing feature but if nobody uses it then there was no value delivered.
So how can we improve our experiments to optimize for delivering more value? At a high level, there are two main approaches and we’re going to end up doing both to some degree.
- More planning is a strategy where we try to reduce the number of failed experiments by spending more time thinking up front.
- Shortening the time to value is a strategy that allows us to run more experiments in the same period of time. The faster we can run experiments the more quickly we can learn whether we are doing the right thing and the greater the chance that some value will get delivered.
When we focus on planning we are trying to reduce the number of failed experiments. If experiments are expensive or dangerous then more time reducing the risk of failure may well be warranted. When experiments are quick or inexpensive it usually isn’t.
We need to recognize that more up-front planning is a balancing act. More planning means lower risk for each individual experiment at the cost of running fewer experiments - any time spent on those planning activities is time not spent executing others. Less planning means that we’re spending more time building and delivering while increasing the risk that we may not be delivering the right thing. That the experiment may not deliver value at all. Finding that right balance between managing risk and delivering is an ongoing balancing act.
Time to value
The easiest way to shorten time to value is to make the individual work items smaller. How small can the items be and still have the expectation of delivering value? By making work items smaller we increase the number of experiments we can run and reduce the overall cost of each experiment.
In addition to making work items themselves smaller we can also look at improving our process to decrease the time it takes to run these experiments. Do we have many wait states in our process that slow the work down? Can we remove those? Once we’ve made the individual work items smaller then we need to look at how quickly we can get them through the system.
- We can’t know if a work item had value until it’s completed and we’ve observed or measured the benefit it delivers.
- It’s not possible for a work item to have value until it’s finished.
- Making work items smaller will reduce the time it will take us to prove our hypothesis.
- Improving the workflow by removing waste will also result in a shorter learning cycle.
- The earlier we can validate our experiments, the faster we can adapt to those results.
- Planning ahead is a balancing act between confidence in a single experiment and running lots of experiments. Too much planning increases risk in each experiment.
See also: Thinking in Bets by Annie Duke.