In this article

A team can have a convincing case for acquisition work, retention work, and a better paid offer, and still have capacity for only one substantial project. Saying all three are priorities leaves the real decision to whoever books the engineer first.

Prioritization is the allocation decision: which work gets scarce capacity, what waits, and what evidence would justify changing that commitment. A growth model can show which relationships matter. It cannot tell you how much effort a particular intervention deserves, whether a dependency will block delivery, or what the business can afford to postpone.

Begin with the capacity and outcome you actually have. “Six weeks” means little until you know whose time is available and what else must fit into it.

Name the constraint precisely

If engineering capacity is scarce, the team’s best next move might be a change that mostly needs commercial or operational work. If support capacity is already exhausted, an acquisition campaign that creates more assisted setups can worsen the bottleneck. If cash is limited, the timing of costs and receipts can matter more than a larger benefit a year from now.

State the outcome and horizon alongside the constraint. More retained paid accounts at quarter end is a different objective from near-term cash or evidence that a new segment can succeed. Be honest when the company needs more than one view. A longer-term product investment may be worthwhile while still being unaffordable this month.

Reserve capacity for obligations before comparing discretionary bets. A broken billing flow or a serious reliability issue has a different status from an optional growth experiment. Put it in the plan with its reason and cost. Otherwise the growth backlog will appear feasible only because essential work has been omitted.

Compare contributions to the same outcome

Visits, activation percentage, and upgrades are intermediate measures. An acquisition idea expressed in visits cannot be compared directly with a retention idea expressed in saved accounts.

Translate each proposal into a plausible contribution to the chosen outcome, keeping the calculation visible. For a retained-paid-account goal, an acquisition proposal depends on how many additional entrants will become paid and remain so. A retention proposal depends on the population at risk and the share the intervention might help. An upgrade proposal needs a view of later cancellation as well as initial purchase.

Use ranges where the effect is uncertain. Label them as estimates, not confidence intervals, unless they come from a statistical procedure that supports that interpretation. Keep the evidence for the problem separate from the evidence for the proposed solution. Knowing that customers struggle does not tell you how much your design will help.

Intercom’s problem-definition principle emphasizes understanding the customer problem and revising that understanding as the team learns. At prioritization time, this becomes a practical distinction between funding a delivery commitment and funding the work needed to choose one.

A score can organize the discussion

Intercom’s original RICE framework combines reach, impact, confidence, and effort. Reach is tied to a period, and effort includes contributions from across the team. Those definitions prevent some common mismatches, such as comparing a month’s audience for one idea with a year’s audience for another.

Use the score to expose disagreement. If a project wins because someone assigned it very high impact, examine that estimate. If another loses because its effort is high, ask whether there is a smaller version that answers the main question.

Do not mistake the resulting order for a schedule. Two projects may require the same specialist at the same time. A lower-scoring dependency may unlock several others. A project with a low total effort estimate may still require a scarce person’s uninterrupted attention. Capture those constraints outside the single score rather than hiding them inside an arbitrary confidence adjustment.

Spend the scarce capacity deliberately

Consider an illustrative subscription product choosing work for the next six calendar weeks. After maintenance and existing commitments, it has six engineering person-weeks available, plus limited design and analysis support. The objective is additional retained paid accounts at quarter end.

CandidateIllustrative contribution estimateEngineering needOther cost or uncertainty
New acquisition channel20–50 retained paid accounts2 person-weeksAdvertising spend; untested audience intent
Reusable setup configuration25–45 retained paid accounts4 person-weeksWhether repeated setup causes the observed drop-off
Clearer upgrade comparison10–35 retained paid accounts2 person-weeksWhether successful free users want the paid offer

All numbers are teaching estimates. They describe hypotheses, not expected results from real projects.

The table makes several plans possible. It does not establish that combining acquisition with reusable setup yields 45–95 additional accounts. Their effects may overlap, and the new channel’s customers may behave differently. Adding project estimates without considering those interactions overstates the plan’s certainty.

Suppose direct observation and task histories support repeated setup as a substantial obstacle, while the acquisition channel is untested. The team could commit four engineering person-weeks to a scoped reusable-setup change. It could spend limited marketing capacity on a small audience test, with an explicit spend cap, before allocating engineering to the full channel. The remaining engineering capacity can cover measurement and rollout risk rather than being filled automatically.

This is a defensible allocation if the evidence and staffing assumptions hold. It is not a rule that retention always wins. If recurring use is already strong and qualified demand is scarce, the channel could deserve the main commitment instead.

Buy information when it can change the decision

A learning task earns capacity when its result could change what you fund. Before scheduling it, write down the decision it informs.

For the acquisition channel, a small test might reveal that the intended audience is too expensive to reach or has the wrong need. That could avoid a larger integration effort. For the upgrade comparison, conversations and a limited offer test might reveal that the proposed paid distinction is unimportant. More polished comparison copy would then be the wrong project.

Put a boundary on the investigation. State what evidence is sufficient to proceed, stop, or narrow the work, and when the team will review it. “Learn more about pricing” has no natural finish. “Determine whether this segment needs the proposed team capability before building its purchase flow” is a usable commitment.

Allow for different feedback speeds. Acquisition can produce early response data while retention requires mature cohorts. A review after two weeks should not automatically favor the project with the fastest visible metric. Decide in advance which early signals permit a safety stop and which outcomes need more time.

Make the next review a decision, too

Record the chosen work, the capacity assigned, and the assumptions carrying the decision. Include what is deliberately waiting. That makes the opportunity cost visible when a new request arrives.

At review, compare evidence with those assumptions. If customers use reusable setup but still fail at the next task, the original intervention may have addressed a real inconvenience without resolving the growth constraint. If delivery is blocked by a dependency, the remaining capacity may need to move. If nothing material has changed, finish the committed work instead of rescoring the whole backlog.

The plan for the next six weeks should make it possible to decline a good idea for a specific reason: the relevant capacity is committed to a better-supported opportunity, and there is a defined point at which new evidence can change that choice.

Sources

Something we should clarify?

Suggest a correction