Backlog

backlog — Everything known, in one order

The product backlog: the whole of the known work in a single ordered list, with the reason for the order written down. It is groomed occasionally, not every iteration.

It is ordered, not prioritized — and the distinction is the whole point. Scrum changed that word deliberately: ordering by priority is only one technique among several, and rarely the best one. A position reflects value, risk, cost, and dependency together, weighed against each other, which is precisely what a single priority label cannot express.

So there is one list, every item has exactly one position, and that position is the only statement of sequence the document makes. No priority labels, no tiers, no parallel "critical" list — each of those is a second opinion that can contradict the first, and the moment two disagree the list stops being trusted and people fall back to asking someone.

A total order is uncomfortable on purpose. It forces the comparison between two things you would rather call "both high", and that comparison is the entire value of the exercise. The Why here column is where it gets resolved: not "P1", but why above the item below it.

It deliberately creates no issues. That is sprint's job, and the separation matters: work becomes an issue at the moment someone commits to doing it, and everything before that moment is a plan that can still change cheaply.

Ordering is the entire value. A list of everything that must be built is not a backlog; it is an inventory.

Where it goes

<docs>/BACKLOG.md — the product backlog, groomed in place.

<docs> is the - Docs: line in .steering/tech.md, which defaults to docs/. In a multi-repo product it points at the shared documentation repository instead, so product-level truth has one home rather than one per repo.

Steps

1. Read the epics. Start from the walking skeleton — it is always first, because it is what discovers integration problems while they are cheap.
2. Weigh value, risk, cost, and dependency together — not size alone, and not enthusiasm. The item that would most change the plan if it went badly goes as early as its dependencies allow.
3. Force a total order. No ties. When two items feel equal, that is the comparison worth making, not the one to avoid — ask which you would drop if only one could ship, and order by the answer.
4. Flag the ones that block others. A slip on a blocking item moves everything after it, so mark it on its row. It is an attribute of the item, not a second ranking.
5. Note roughly how many issues each item is worth. One backlog item usually becomes several issues — a change, the tests that were missing around it, the migration it forces. An item nobody can size at all is a signal: it needs a spike, or it needs to go back to epics.
6. Record the order and the reasoning, so the next grooming starts from the argument rather than from scratch. Then hand to sprint, which decomposes the top items into issues.

Template

# Product backlog

- Last groomed: <YYYY-MM-DD>
- Ordered, not prioritized. Position reflects value, risk, cost, and dependency together. There is no separate priority field, and adding one would contradict this.

| # | Item | Epic | Blocks | Rough size | Why here |
| :-- | :-- | :-- | :-- | :-- | :-- |
| 1 | <item> | EPIC-n | #3, #7 | <~n issues> | <the risk it retires, or what it unblocks> |
| 2 | <item> | EPIC-n | — | <~n issues> | <why above #3 specifically> |

## Unshaped
Items that cannot be ordered yet because nobody knows what they are. This is a queue to empty, not a tier — anything sitting here is undecided work, and it does not get built while it stays here.

- <item> — <what has to be answered before it can take a position>

Rules

• The walking skeleton goes first. Always. Reordering it later is how integration risk gets discovered at the end.
• One list, one position per item. No priority labels, no tiers, no separate critical list. A second statement of sequence can contradict the first, and then neither is believed.
• "Ordered" rather than "prioritized" is not pedantry. Priority reads as one dimension; a position has to absorb value, risk, cost, and dependency at once. The Why here column is where that reasoning lives, and a backlog without it is a ranking nobody can argue with — which means nobody can correct it either.
• No ties. "These two are both high" is the comparison being dodged.
• This skill creates no issues. Ordering is cheap and reversible; an issue is a commitment. Keeping them separate is what lets the plan change without leaving debris in the tracker.
• A backlog item is not an issue. One item commonly becomes several — the change, the test coverage it turns out to need, the migration it forces. Sizing them 1:1 at this stage is how a plan quietly becomes wrong, because the decomposition has not happened yet and cannot be guessed accurately.
• Items are coarse on purpose. Fine-grained ordering of work nobody has decomposed is precision without accuracy.
• Every "Next up" row states why now. Without it the order is preference, and preference gets re-argued.
• Re-sequence when reality diverges. A backlog that has not moved in a month is either a finished plan or an ignored one.

Red flags

• A priority column has appeared beside the position column. One of them is now lying, and people will learn which by trial.
• Half the list is flagged as blocking. Then blocking means nothing, and the flag has become a way of saying "important" without paying for a position.
• The order matches the epic numbers. Ordering by the sequence things were thought of is not ordering.
• "Unshaped" never empties. Those items are being deferred without anyone deciding to defer them.
• Items read like issue titles. Then they have been decomposed too early, and the decomposition is now frozen into an ordering decision that has to be redone.
• The backlog has been created as issues. Then the tracker is a plan rather than a commitment, and every future search filters through work nobody chose.
• Nothing is in "Not ready". Either the plan is unusually complete, or items with unresolved dependencies are being ordered as though they were actionable.