Key takeaways
  • A healthy backlog has three layers: Discovery, Refined and Ready.
  • Five anti-patterns turn it into a graveyard.
  • Groom on a fixed cadence, with clear Definitions of Ready and Done.

Most backlogs are graveyards. Ideas that seemed urgent in January, still open in May. Stories nobody can remember writing. Bugs that have been "P3 — will fix next sprint" for two years. The team has stopped looking at it; the PO is afraid to declare bankruptcy; engineering treats whatever's in flight as the real backlog and ignores the rest. The artefact is dead. But it's still claiming a slot in the tool.

A healthy backlog is the exact opposite. It's a working queue: small, alive, three-layered. Items move forward (or out) every week. Refinement takes fifteen minutes, not an hour. Engineers never have to ask "which of these should I pick up next?" because the answer is on the top.

The three layers.

Diagram · The three layers of a healthy backlog
The three layers of a healthy backlog — Discovery, Refined, Ready A horizontal funnel showing the three layers of a healthy backlog plus the In-flight and Done states. Layer 1 — Discovery: ideas and problems, no stories yet, lives at outcome level, the messy end of the queue. Layer 2 — Refined: shaped initiatives broken into draft stories, has a clear outcome, lacks final acceptance criteria. Layer 3 — Ready: fully-formed stories that meet Definition of Ready, picked up by engineering within one or two sprints. Then In-flight (this sprint) and Done (shipped). Items flow left to right and items also exit upward via Defer. The PO Piece owns the gates between Discovery, Refined and Ready. The PM Piece owns what enters Discovery. The team owns what enters In-flight. LAYER 01 Discovery Ideas · problems No stories yet LAYER 02 Refined Stories drafted Shape known LAYER 03 Ready DoR met Pick up < 2 sprints IN FLIGHT This sprint Team owns DONE Shipped DoD met flow forward · weekly · stale items either move or exit PM PIECE curates entry PO PIECE owns the gates TEAM owns delivery Items also exit upward via Defer — clean exits matter as much as clean entries.
A healthy backlog flows weekly. Discovery curates problems. Refined turns problems into draft stories. Ready is the engineer-pickup zone — every item meets Definition of Ready. The PO Piece owns the gates between each layer.

Layer 01 · Discovery. The messy end of the queue. Ideas, problems, customer signals, hypotheses. No stories yet. Items live at outcome level ("reduce checkout abandonment", "support mobile sign-in for the trial cohort"). The PM Piece curates what enters here; the PO Piece curates what leaves.

Layer 02 · Refined. Items that have been shaped into draft stories. The team knows what shape the work is, roughly what it'll cost, what the risks are. But acceptance criteria are still loose, and dependencies haven't all been mapped. Items live here for a sprint or two while they ripen.

Layer 03 · Ready. Stories that meet the Definition of Ready. Acceptance criteria are sharp, dependencies are unblocked, design is in place where needed. The team can pick them up within one or two sprints without going back to product for clarification. The PO Piece keeps this layer at roughly two-to-three sprints' worth of work — full enough that engineering never starves, light enough that nothing stale piles up.

Discovery curates problems. Refined turns problems into draft stories. Ready is the engineer-pickup zone. The PO owns the gates.

The five anti-patterns.

  • The mausoleum. Every idea ever raised, still open. Half of them are obsolete; nobody dares close them. Closes blocked by hope.
  • The fire hose. Items shoved straight into Ready without refinement. Engineers picking up under-specified work, asking three questions, blocking on answers.
  • The wish list. Stakeholders push items in; nothing exits. Backlog grows monotonically until it collapses under its own weight.
  • The immutable plan. Backlog set in stone six months ago. Refresh blocked by political cost of "changing direction".
  • The shadow backlog. The real work lives in a side spreadsheet, a Notion doc, or the lead engineer's head. The tool's backlog is theatre.

If two or more of these are present, the function isn't short of a backlog tool. It's short of a PO Piece with the seniority to close items, push back on stakeholders, and keep the queue moving. The artefact follows the seat. Buy the seat first.

Definition of Ready, Definition of Done.

The gates between layers aren't vibes. They're defined. Definition of Ready is the checklist an item passes to enter the Ready layer. Definition of Done is the checklist a story passes to exit In-flight as actually-shipped.

A typical DoR: persona named, outcome stated, requirements scoped at 5–10 items, AC objectively checkable, dependencies identified, design ready where needed, T-shirt sized. A typical DoD: AC verified, automated tests green, deployed to production behind a flag if not GA, observability in place, documentation updated, no open critical bugs in the linked issue list, accepted by the PO.

If your team doesn't have these written down, half the friction in refinement and acceptance comes from people disagreeing about thresholds nobody set. Write them on a single page; pin them in the tool. Then enforce them. A story that doesn't meet DoR doesn't enter Ready, even if the stakeholder asking is the loudest one. A story that doesn't meet DoD isn't done, even if the engineer is keen to move on.

How often to actually groom.

Forty-five minutes a week is the floor for most teams. The point isn't to refine every item — it's to keep the gates moving. The PO walks the layers, asks the questions that unblock the gates, and closes the items that have aged out. Once a quarter, a longer session: declare bankruptcy on the older Discovery items, re-cast the priorities against the latest roadmap, and reset what "two sprints' worth" means.

If your backlog grooming takes longer than that every week, you're refining items that shouldn't be in the queue yet. If it takes less, your PO probably isn't really doing the job. They're just sorting.

Where this lands.

A healthy backlog is the visible output of a strong PO Piece. If the backlog is leaking (wish list, mausoleum, shadow backlog), the fix isn't another grooming session. It's the seat above the grooming session. The Backlog Set Piece (1–2 weeks, fixed scope) reshapes the queue, writes the DoR/DoD, and proposes a grooming cadence the team will actually keep.

That closes the Product Artefacts series for now. Future issues will go deeper on specific artefacts the team builds around the backlog: release plans, RACI charts, the metrics tree. Then back to opinion and diagnostic content. All issues →