Searching for the Estimation Grail: Relative Estimation

Relative Estimation – Description

Relative estimation is also known as Silent Grouping, Affinity Estimating and Team Estimation Game. It is also well described in this handbook by Steve Bockman.

This method is often used by experienced teams that are not satisfied with the classical planning poker approach. Depending on the actual implementation of this method, a team can reduce amount of noise and empty discussions, estimate more items per session and achieve more accuracy.

Typical Process

A typical game session consists of two rounds. Since the process is turn-based, all participants have to know their position (people usually gather in circle).

Round 1

This round has nothing to do with actual estimating. It has everything to do with positioning items compared to each other.

A session starts with 1 card laying on the table – a reference card – and the deck of items to estimate.

First participant picks a card from the deck and moves it either on the left or on the right to the reference card. Left – easier, right – more complex. Then turn goes to the next participant.

Next participant has to check the table – if she agrees with the sequence, then she picks next card from the deck and puts it to the desired position (left, right, between, etc). If she disagreed with the cards positioning on the table – she would have to pick a card from the table instead of the deck (discussion happens at this moment). Turn goes to the next participant.

This repeats until all cards are on the table (the deck is empty) and until all players agree on the sequence of the cards (no more shuffles, everyone is skipping his turn).

Round 2

We have all items laid on the table in the sequence of their complexity. It’s time to assign some estimates now.

The first participant now attempts to assign the actual estimates – either by putting cards on top of them, or by dragging piles of cards into predefined table columns, or by whatever is described in the particular method. When she finishes – turn goes to another participant, who then either skips if she agrees or adjusts the estimates if she disagrees (another moment of discussion).

Applying Relative Estimation for Distributed Teams

So good so far.

But how do you actually implement this method for location distributed teams?

There are few options available. I will briefly describe them

Google Spreadsheets

Yes, that’s right. Odd but true. You can use Google Spreadsheets for relative estimation. Although it just requires some preparation. You can draw some squares which would represent the items to estimate, then “pile” them in the corner of the screen, put reference block to the middle of the screen and invite everyone to the document. This solution is neither elegant nor fancy, but it works. And you can also have some fun by drawing some figures of pacman.

The huge drawback is the difficulty of session setup – you would need either to create a template or do a lot of copy-pasting, adjusting titles to fit the box and to make them readable, etc.

Trello

Some people prefer using Trello for such kind of need. Preparation is simpler thanks to the user-friendly interface. You just create items same way you do in your issue tracker (you’re lucky if you’re using the same software), then you create a board with size columns (Fibonacci or T-Shirts), and one extra column for keeping the not-yet-estimated items.

Special Apps

Some time ago together with a friend we created and open-sourced the tool (Agile Estimation) for relative estimation for distributed teams. It works best of all described methods since its only purpose is to implement relative estimation method for distributed teams. You can either use a deployed version, or install it in-house.

Here is the demo:

Conclusion

This method is a good fit for distributed teams with a decent maturity. It makes it possible to estimate nearly 30-40 items per hour. This is also my personal preference.