It is a common mistake, performed by many in the No Estimates community, to fail to understand that all risk comes from uncertainty. Uncertainty comes in only two forms:
- Epistemic uncertainty - is the lack of knowledge or information in any phase or activity of the modeling process or activity (Epistemology is the study of Knowledge). Epistemic uncertainty is characterized by alternative models. For discrete random variables representing a process or model, the epistemic uncertainty is represented by alternative probability distributions. To handle Epistemic uncertainty that creates reducible risk, we need to perform risk buydown activities to increase our knowledge of the uncertainty and risk. These can be:
- Build two in case one breaks,
- Build a prototype to confirm the production item has a higher probability of working,
- Overbuild the item.
- Aleatory uncertainty - is the inherent uncertainty due to the probabilistic variability or underlying statistical processes. This type of uncertainty is Irreducible, in that there will always be variability in the underlying variables. To handle Aleatory uncertainty that creates risk, we can Only provide margin:
- Schedule margin for the naturally occurring productivity of the work being performed by the team,
- Cost margin for the naturally occurring variances in the cost of the product or service,
- Performance margin for the produced item - more bandwidth, more storage, more CPU cycles to handle the naturally occurring demands on the system.
Epistemic uncertainty is due to limited data and knowledge of the process or project item. The epistemic uncertainty is characterized by alternative models. For discrete random variables, the epistemic uncertainty is modeled by alternative probability distributions. For continuous random variables, the epistemic uncertainty is modeled by alternative probability density functions. In addition, there is epistemic uncertainty in parameters that are not random by having only a single correct (but unknown) value.
Epistemic uncertainty comes from a lack of knowledge. This lack of knowledge comes from many sources. Inadequate understanding of the underlying processes, incomplete knowledge of the phenomena, or imprecise evaluation of the related characteristics are common sources of epistemic uncertainty. In other words, we don't know how this thing works so there is uncertainty about its operation.
Aleatory uncertainty is familiar to students of probability theory and includes the outcomes of tossing dice (Alea is a single Die used by the Greeks to gamble) and drawing cards from a shuffled pack. In statistics, aleatory uncertainty is present in almost all data that we obtain, produced by the random variability between the members of a sampled population or by random measurement errors.
Flipping a coin, rolling a die, drawing a numbered ball from the bingo machine is an aleatory uncertainty.
Getting to the Point
For all project work and product development - in the presence of uncertainty - reducible and irreducible - making decisions in the presence of this uncertainty, when using other people's money, with limited resources, any kind of deadline, and any mandatory capabilities from this work - requires - actually mandates - making estimates.
Let's look at some common fallacies of Agile Development and No Estimates with this knowledge:
- Slicing makes the amount of work smaller but does not reduce aleatory uncertainty. It can make the impact of Epistemic uncertainty have less impact. It's a good approach but is not a risk reduction strategy. It's a risk handling strategy. The aleatory risk is not addressed by slicing since the underlying processes are naturally occurring and cannot be reduced. All slicing does is make the visibility of the uncertainty finer-grained, but the cumulative impact of the un-handled uncertainties is still the same.
- Continuous delivery exposes risk faster but does not reduce Epistemic risk. Only direct, explicit risk reduction actions address epistemic uncertainty. Delivering frequently simply exposes the risk faster.
Any conjecture that credible decisions can be made with NO estimates - in the presence of risk created by Epistemic and / or Aleatory uncertainty is a fallacy
[1] “Epistemic versus Aleatory Judgment Under Uncertainty,” C. R. Fox and G. Ülkumen, Working Papers, Variants of Uncertainty and Judged Probability, UCLA Anderson School of Management, 2012.
Here are some resources for managing In the Presence of Uncertainty:
