Affinity Estimating

In Affinity Estimation, all participants estimate different items at the same time and exceptions are discussed in more detail. It’s great for large backlogs or initial/distant estimates.

Pros/Cons

  • It is very fast, allowing to estimate large backlogs very quickly
  • It highlights assumptions and misinterpretations by exception
  • Everyone is involved in the estimate
  • It is very easy for experts to ‘lead’ the estimating

Tools

  • Set of estimation units in card form (Fibonacci, T-Shirt Sizes)
  • Additional recommended cards:
    • Infinite (Too big to estimate)
    • Question (Unable to estimate with existing information)
    • Red, Amber and Green cards for confidence estimating
  • A number of items to estimate, printed or written on pieces of paper/card

Some coaches may suggest doing this in complete silence, but in my personal experience I have found the conversation that results from this exercise helps create some shared understanding of the work which is worth the extra time it takes.

Flow

To set up, create the following - probably on a large flat surface:

If you have no reference point, start by finding a typical mid-complexity piece of work that you’re pretty confident with and place that in the middle of the wall/table, place that in the 5 column in the high confidence row.

Next all team members at once consider other work items and add this to the wall one by one, thinking about the relative complexity of the other items and how confident you are with your estimate.

Keep in mind that for the work item to be a complexity ‘5’ this should be about the same as a ‘3’ and a ‘2’ put together.

Any person in the estimating process can move any work item at any time, even if they disagree, however, if you find a work item moves back and forth, this should be discussed. This avoids unnecessary conversation about items that everyone agrees are the same size.

If someone feels they are unable to estimate, they can move the item to the question column. Then someone else can move it out of the question column providing they explain their estimate.

The above diagram shows an example outcome from an affinity estimation session. It’s quite common to add a numerical modifier to the levels of confidence when using the estimates to predict. For example:

  • Red = 2x Modifier
  • Amber = 1.5x Modifier
  • Green = 1x Modifier

We would then calculate optimistic-pessimistic estimates as follows:

As a result this total estimate for the work that we are able to estimate is 52-92points. With the knowledge that there are 2 items that need breaking down and adding to the estimate and 1 item that needs more information.

Do you think Affinity Estimation would work for your organisation? What problems could you forsee?