Busting Agile Prediction Myths
When first hearing about the processes that agile teams (particularly Scrum) follow, listeners might reach the incorrect conclusion that teams using an agile process cannot provide estimates, predictions, or commitments about what they will deliver.
This can be particularly dangerous when the team members using the agile process don’t understand that they can—and should—provide estimates, predictions, and commitments to their stakeholders.
In their recent eBook, Top 30 Agile Myths Busted, the TeamPulse team at Telerik call out and describe two myths about agile prediction:
- Myth #5: Agile projects don’t provide budget estimates.
- Myth #7: Agile means no commitment.
In their essay on budget estimates, they explain that agile teams use time-boxing (fixed time and resources) within each sprint and not what they call scope-boxing (fixed intended scope given time and resource constraints). They correctly identify that trying to fix all three is always inaccurate, and their point is that each particular sprint has a fixed budget. In effect, the expenditure of resources over time is fixed.
What they don’t cover in this essay is that while budget per unit time is completely predictable with this model, you are faced either with uncertainty about when the project ends or uncertainty about what is ultimately delivered.
If you define a budget estimate as “What will we spend?” then, yes, their assertion is true. If you define a budget estimate as “What will we spend in order to achieve X?” then this assertion is not true, because you don’t know how long you will be spending (at a predictable rate) to achieve X.
In their essay on agile commitments, the authors seem to take the position that “commitments are only as good as the paper they are written on.” Instead of showing how commitments can be made, they point out that commitments using waterfall processes are rarely met (true, both in my experience and as reported by the Standish Group in their annual Chaos reports), so agile is no different.
I agree with their statements, but they missed an opportunity to acknowledge that commitments can be made even when following an agile process.
Commitments are made by making predictions; assessing the risk of inaccuracy in those predictions; and based on those predictions, making a promise that incorporates the identified risks. This is true for both waterfall and agile processes. Predictions are based on estimates—the error in prediction is the sum of the errors in the underlying estimates—and the central limit theorem demonstrates that the overestimated and underestimated tasks will tend to cancel each other out. The larger the project, the more powerful this effect becomes.
You can manage risk and commit to a subset of what you predict. The iterative delivery and conspicuous feedback loops it provides allow you to refine—consistently upwards—what you are able to commit to as the project progresses.
This creates a virtuous cycle, increasing management’s faith in your commitments (and therefore increasing management’s ability to plan effectively), and increasing your team members’ well-being as they continue to deliver and overdeliver.
Meanwhile, your stakeholders are getting the commitments they require.
