**How large portions of the work should be estimated**

In software development, it is difficult to understand the complexity of a project. It is therefore a good idea to split up the task in smaller parts where the largest part will take less than 40 hours. Large and complex chunks of work need to be split in smaller parts. My observation in time estimation is that complex tasks seem to be 90% done and stay that way much longer than tasks that are less complex. This is the consequence of undertaking an endeavor in which you have no experience

I will present two methods that I use, both separately and together. We have used planning poker to estimate the most likely value and Lichtenberg to handle the variations.

### Planning Poker

Planning poker is a consensus-based estimation technique. The method was first described by James Grenning but it gained popularity when Mike Cohn wrote the book *Agile estimating and planning.*

When an estimation session is done, one person describes the requirements for the group. The group discusses and specifies how to fill the requirements. Each participant makes an individual estimate but does not say it. They use playing cards with pre-set numbers. The denominators are 0, ½, 1, 2, 3, 5, 8, 13, 20, 40, 100 and “Don’t know”. The values can be in either hours, days, ideal days or story points.

When everyone has played their card, the card is shown and the group discusses the estimates with those who provided estimates that were radically different from the rest. After discussion notes are written (which might be valuable for future development) a new estimation is done. This time the estimates are more similar. If this doesn’t happen, the process is repeated. It is not necessary for the whole group to have the same number because the objective is a qualifying assessment rather than a precise time period.

### Lichtenberg’s Method

Professor Steen Lichtenberg developed a mathematical method to estimate time which also provides a measure of uncertainty in addition to an assessment of the likely value. Minimum and maximum values specify the limits of uncertainty. These dimensions should be such that there is more than a 1% probability that the actual result will fall outside the min-max range. The estimates are based on certain conditions, stated or unstated, such as who should do the job. The uncertainty is processed only on those parts which are most important for the project’s total uncertainty. Because there are so many project unknowns, the min-max range will be large. The project’s end date and total cost are not stated, only a probability assessment is made.

When doing an estimate, specific activities are discussed with a view defining what is included and how it was and will be done. The group tries to estimate how long it will take one of the members to achieve it. This is done by the person or persons who will be doing the activity later.

When the most likely value has been agreed upon, the group estimates how long it will take to do this if everything works out great. Then the group discusses the worst case and estimates the difficulty level of the activity. When estimates of the “most likely” value, “best case” value and “worst case” value are obtained, the following formula is used to calculate the expected value.

This calculates the probability that the team will be within a 95% interval with the following formula.

So let’s look at an example.