How to Run a Backlog Grooming Session

Open your experiment backlog right now. How many ideas are in it? When was the last time someone reviewed the full list? How many of those ideas still align to a current business objective? If the backlog has more than thirty items and nobody has reviewed it this quarter, it has almost certainly stopped being a planning tool and become a dumping ground. Groom quarterly, ideally in the first two weeks after new business objectives have been set.

Before the Session

Review the backlog yourself first. Flag obvious issues.

Checkpoint: As you scan, count the duplicates. If you find more than two, ask yourself why they were not caught when they were submitted.

Update context. Pull in new evidence. Check active and completed experiments.

Checkpoint: Can you quickly produce a list of experiments completed since the last grooming session and their outcomes?

Running the Session

Step 1: Objective alignment (15 minutes). Which ideas align to current objectives? Are there objectives with no ideas?

Checkpoint: Can you filter your backlog by business objective right now? If the answer is 'no, we would have to read through and mentally sort them', then every prioritisation conversation starts with guesswork.

Step 2: Remove the dead weight (20 minutes). Archive ideas older than two quarters with no activity, duplicates, ideas overtaken by results, ideas that no longer align.

Checkpoint: When you archive an idea, where does it go? Can someone find it six months from now if it becomes relevant again?

Step 3: Reassess priorities (30 minutes). Reprioritise what remains against current context.

Step 4: Identify gaps (15 minutes). Objectives with weak or empty pipelines? Underrepresented areas?

Checkpoint: At the end of the last grooming session, were any gaps identified? If so, were they addressed before this session?

Step 5: Confirm the top of the pipeline (10 minutes). Top five to ten ideas with owners and target dates.

What Good Looks Like

After a good grooming session, the backlog is lean, current, and aligned to strategy. If the checkpoints surfaced problems, they probably fell into one of two categories: the backlog is not connected to anything else in the programme, or it has no structural enforcement.