A backlog feels responsible because nothing has been forgotten. Every idea, request, fix, experiment, and improvement has a place. The list grows, the team stays busy, and the business appears to have more opportunities than it can handle.
But a backlog only answers one question: what could we do? It does not tell you what deserves attention now, what must wait, or what should never be done. When every request remains visible and every item is described as important, the list stops helping the business choose. It becomes storage for unresolved decisions.
I want a priority system to create a clear order under real constraints. It should connect work to an outcome, limit how much can be active, expose the cost of starting something new, and make it safe to remove work that no longer matters. The goal is not to organize a longer list. The goal is to make fewer, better commitments.
Start with the constraint the list is trying to ignore
Most businesses have more possible work than available time, money, or attention. That is normal. The problem begins when the planning process pretends the constraint does not exist. Ten projects receive the same priority, several people begin each one, and progress slows everywhere because the operation has accepted more active work than it can support.
Before ranking tasks, define the limiting resource. It may be the owner’s decision time, a developer’s capacity, the number of campaigns the team can measure properly, or the amount of operating cash available for experiments. The exact constraint can change, but it needs to be named. Otherwise, planning becomes a negotiation over preferences rather than a decision about capacity.
A useful plan is allowed to disappoint good ideas. If the team can complete two meaningful improvements this month, the plan should contain two meaningful improvements, not eight projects labeled high priority. A priority becomes real when something else is required to wait.
- Name the resource that limits how much work can move at once.
- Set an active-work limit that reflects ordinary capacity, not a perfect week.
- Reserve room for recurring operations and predictable exceptions.
- Treat a new urgent request as a decision about what will stop or move.
Connect every candidate to an operating outcome
Tasks sound important when they are described by activity. Redesign the page. Add the supplier. Build the report. Automate the workflow. Each one may be useful, but the verb does not explain why the business should spend capacity on it now.
Rewrite the candidate as an outcome. The page change should reduce a specific source of customer confusion. The supplier launch should fill a documented assortment gap. The report should make a recurring decision faster or more accurate. The automation should remove stable, repetitive work without hiding an exception that still needs judgment.
This does not require a complicated forecast. It requires enough clarity to compare unlike requests. When two projects compete for the same capacity, I want to know which business result each one is meant to improve, what evidence supports the need, and how we will recognize progress. An item with no clear outcome is not ready to compete for priority.
Separate obligations, improvements, and experiments
A single undifferentiated backlog creates false comparisons. Fixing inaccurate product information is not the same kind of work as testing a new acquisition idea. A customer promise with a deadline is not competing on equal terms with a future design improvement. When everything lives in one ranked list, whoever argues most convincingly can make optional work appear mandatory.
I separate the work into three broad groups. Obligations protect commitments the business has already made: customer communication, supplier requirements, accurate information, and work needed to keep the operation reliable. Improvements strengthen a known part of the business. Experiments test an uncertain opportunity with a defined limit.
The groups still share capacity, but they deserve different questions. An obligation is evaluated by consequence and deadline. An improvement is evaluated by the size and certainty of the operating benefit. An experiment is evaluated by the quality of the evidence it can produce relative to its cost. This keeps a speculative idea from borrowing urgency from work that protects the current business.
Rank with evidence, not volume
A request does not become more valuable because it has been repeated in several meetings. Repetition may reveal a real pattern, but it may also mean the same opinion has traveled through multiple channels. I want the decision to return to evidence.
For operating problems, look for frequency, consequence, and reach. How often does the issue occur? What does it cost in time, margin, customer trust, or supplier confidence? How many orders, products, or people does it affect? For growth work, add confidence and reversibility. What makes us believe the opportunity exists? How cheaply can we learn whether we are right? Can we stop without creating a large ongoing obligation?
You do not need to turn every decision into a scoring formula. A score can create false precision when the inputs are guesses. A short comparison is often enough. The purpose is to expose the assumptions so a familiar, exciting, or politically popular request does not win by default.
- What observable problem or opportunity does this address?
- How large is the consequence if we do nothing this cycle?
- What evidence supports the proposed response?
- How much capacity will the work consume, including follow-up and maintenance?
- What smaller action could produce the next useful piece of information?
Make the cost of starting visible
Starting feels productive. A new project creates meetings, documents, assignments, and early movement. Finishing is harder because the last part often contains testing, cleanup, documentation, customer communication, and the decisions that turn activity into a usable result.
Every active project also creates carrying cost. Someone has to remember the context, answer questions, review changes, and decide what happens next. When work pauses, that context decays. Restarting requires the team to rebuild an understanding it already paid to create. A business can spend a great deal of attention maintaining unfinished work without noticing the expense.
That is why I want a visible rule for starting. Before a new item becomes active, name the owner, define the result, identify the first complete milestone, and show where the capacity will come from. If the active limit is already full, finish, pause, or cancel something deliberately. Do not quietly add another commitment and hope the constraint disappears.
Use a now, next, and later view
A ranked list of one hundred items suggests a level of certainty the business does not have. The order near the bottom will change long before the team reaches it. Maintaining that detailed ranking consumes time while giving old assumptions the appearance of a plan.
I prefer three horizons. Now contains the small number of commitments currently receiving capacity. Next contains the few items most likely to earn that capacity when something finishes, along with any preparation they need. Later contains possibilities worth remembering but not yet worth planning in detail.
The value is not the labels. The value is the honesty. Now is a commitment. Next is a current expectation, not a promise. Later is storage, not a queue. An item in later can move forward when new evidence appears, remain where it is, or disappear without the business breaking an agreement.
Keep the now view small enough that everyone can explain it. If the active work requires a complicated dashboard merely to remember what has started, the business probably has too much in motion.
Review aging work and remove it without guilt
Old backlog items create a strange kind of pressure. They have survived several reviews, so they begin to feel like commitments even when nobody has chosen them. The team may keep moving them forward simply because deleting them would force a decision.
Age is information. If an item has remained untouched while the business continued operating, either the consequence was smaller than expected, the outcome was not important enough, or the proposed solution never became convincing. None of those conclusions require shame. They require an update.
Review older items on a schedule. Ask whether the underlying problem still exists, whether the evidence has changed, and whether the requested solution is still the smallest sensible response. Close duplicates. Combine requests that point to one root problem. Archive ideas that have no current owner or evidence. If a real obligation has aged, move it into an explicit plan rather than allowing it to hide in the backlog.
Deletion is part of prioritization. A system that can only add work is not preserving opportunity. It is accumulating indecision.
Close the loop when priorities change
Priorities should change when the facts change. A supplier issue may affect a customer promise. A campaign may produce evidence that deserves a faster response. An operating failure may expose a larger risk than the current project. The problem is not changing the plan. The problem is changing it silently.
When new work moves ahead, record the reason and the tradeoff. Which fact changed? What moved out of now? What happens to the partial work? Who needs to know? When will the displaced item be reconsidered? This short explanation protects trust and prevents the team from treating every interruption as a permanent emergency.
It also improves future planning. If priorities change every week for the same reason, the business may be underestimating recurring work, leaving no capacity for exceptions, or using urgency to avoid making a stable choice. The pattern is more useful than the explanation for any single change.
Turn the backlog into a decision tool
Start with the active work. Count the real commitments, including the projects that have gone quiet but still require a decision. Compare that number with the capacity available during an ordinary week. Finish, pause, or cancel until the operation can support what remains.
Then clean the candidates. Give each one an outcome, evidence, owner, and rough cost of completion. Separate obligations from improvements and experiments. Move only the few likely next commitments into a near-term view. Let the rest remain possibilities without pretending they have already earned a place on the calendar.
Finally, establish a review rhythm. Look at active work often enough to remove blockers and finish. Review next when capacity is about to open. Review later occasionally to find new evidence, merge patterns, and delete ideas whose moment has passed. The system should spend most of its energy moving important work to completion, not polishing the inventory of work that may never begin.
A business will always have more useful things it could do. That is not a planning failure. The discipline is choosing what matters now, protecting enough capacity to finish it, and allowing the unchosen work to remain unchosen.