Select Page

Jumpstart your prioritisation sessions: classify WSJF fields using Affinity Grouping, and align to numbers afterward. It’s a big step up on HiPPO, and offers enough fidelity without getting lost in false precision of analysis.

When you start with WSJF, you often find that you have a large number of items to order. As reducing this list to a manageable few is the output of this process, there’s no easy way to short cut it. So you need a way to get an ordered list with a group of experts from both customer and deliver domains in a relatively short space of time.

Remember all you’re doing is working out what you will progress now, next, later and never with increased objectivity, forcing a conversation about Why? that goes beyond Because I said so. High precision is not necessary, as we’re usually talking about high uncertainty contexts in both value and sizing, and it’s only the final order that matters, not the actual number.

### Sidebar Anecdote

I worked in an environment which had a horribly complicated formula for WSJF, made up of almost 20 factors. Disentangling this to understand what was going on, after I’d removed factors that cancelled each other out, I realised that there was a big fudge factor applied to the whole sum. This was indicative of Someone Very Important who had objected that their number just wasn’t big enough. So the team… made it bigger.

This is the kind of problem that my old UX skills recognised. One that’s regularly addressed with Card Sorting and Affinity Grouping.

## Running a Prioritisation Session

The first time I run this with a group, I will usually facilitate, to force the Why? conversations to happen, considering 1 candidate at a time. More mature teams or low stakes environments are less risky, so can be left to collaboratively work on the entire set at once, using cards on a table.

I set up a wall with 7 rows, labelled with T-Shirt sizing, from XXS at the bottom to XXL. Or I might seed the group’s imagination with categories such as:

1. Shoes
2. Bicycle
3. Moped
4. Compact Car (eg a Mini)
5. SUV
6. Minibus
7. Articulated Lorry

Then we will go through the entire set of candidates, one WSJF factor at a time.

Within each factor, I start by asking the domain experts to pick an example that’s about in the middle. And I put that in the Medium/Compact Car row. Now we take all the other candidates in turn and start probing for where it falls:

Is it larger or smaller than that first one?
Smaller? OK. A little bit smaller?
A definite step smaller. One or two steps?
Can you imagine something that’s smaller still? Yes?
OK, let’s leave a gap below and call that XS

With each new candidate, we test against what’s come before

Is it about the same size as that one?
Is it between these two?

and as needed, we might move some of the previous candidates as we learn more about the whole set’s distribution.

Card Sorting in Action

You might recognise this as something akin to reference class estimating. When you have a previous session to recall, you have previous data as references, even (and more effectively) if you have inspected and adapted your understanding since that session.

Once you have the whole set distributed for the first WSJF factor, take a note of each candidate’s value, clear the wall and move onto the factor, repeating until all 4 factors are complete.

## Aligning to Numbers

To produce a sortable stack rank of options, SAFe uses the formula

$\frac{UserBusinessValue+TimeCriticality+RiskReduction|OpportunityEnablement}{Job Size}$

(I have a lot of time for Joshua Arnold’s improvement on this)

and recommends using the modified Fibonacci series as decreasingly precise buckets in each term.

Going from T-shirt sizes to mFib buckets is easy, using the following handy reference table:

XXS XS S M L XL XXL
1 2 3 5 8 13 20
and you simply do the maths on the backend – building a little Excel formula should be trivial for this.

### Weighting Factors

In your context, it might be reasonable to weight one or more of the factors. One customer in a highly disruptive media market strongly weighted for the RR/OE factor, to reflect that they valued learning about customers over immediate revenue.

Using Card Sorting and Affinity Grouping, it’s perfectly possible to WSJF score 20 well-understood candidate items in a short afternoon.

## Bootnote

SAFe City: Sketchnote by Stuart Young

I started using this when running the SAFe City Simulation up against a tough timebox, as a countermeasure to teams getting stuck in analysis.

This simulation workshop has participants practice WSJF prioritisation and roadmapping in a safe setting.

Recommend this post