1. LandOffer
  2. Blog
  3. All articles
  4. Explain an Estimate as a Range: Communicate Assumptions and What Would Change It

Workplace Communication

Explain an Estimate as a Range: Communicate Assumptions and What Would Change It

Explain conditional software estimates with stable scope, focused effort, confirmed capacity and separate waiting time. Revise the same scenario after evidence without inventing statistical confidence.

9 min readLandOffer team

Explain an estimate as a range by naming the deliverable, the conditions behind each endpoint, and the evidence that would make you revise it. Separate hands-on effort from elapsed working days and calendar dates. Say whether capacity and dependencies are confirmed, and identify when you will revisit the estimate. A range can support a planning decision without becoming a delivery commitment.

This LandOffer workplace guide uses one fictional software feature to compare three conditional estimates and then revise the applicable scenario after investigation. All work, capacity confirmations, observations, and conversations are invented examples. The arithmetic is explicit; the ranges are judgment-based planning bounds, not statistically calibrated confidence intervals or actual project results.

Workshop planners comparing three wooden pathways toward the same model pavilion and examining an uncertain join

Define what the range is estimating

“Two to three weeks” leaves important questions unanswered. Does that mean one person's full-time effort, elapsed time with interruptions, a review-ready change, or a production release? State the unit and completion condition so implementation effort cannot silently become a finish-date promise.

Start by defining what must be finished for the estimate to count as complete. In the fictional example, the feature is a repository documentation checker for relative Markdown links and heading fragments. Completion means the agreed checks, automated tests, usage instructions, and a reviewed change ready to merge. It excludes external website availability checks, automatic link repair, production deployment, and cleanup of existing documentation errors.

All three scenarios below retain that same scope and completion condition. They differ in what the team discovers about reusing the repository's existing parser. A smaller feature is a legitimate option, but it must be labeled as a scope alternative, not presented as the low end of an estimate for the original feature.

NASA's Software Engineering Handbook version D guidance on estimate conditions identifies assumptions, functionality, complexity, reuse, technology maturity, and uncertainty as relevant to software cost estimates. Those requirements belong to NASA's context. Here they serve as prompts for explaining an estimate's basis, not as a mandatory standard for every software team.

Keep effort, capacity, waiting, and dates separate

Effort is the work required, expressed here in focused engineer-hours. Duration depends on when that effort can occur, how work is ordered, and what must wait for someone else. Four hours of review activity and four days waiting for an available reviewer are different inputs.

For this simplified example, engineer Hana has four focused hours per working day allocated to this feature. Her manager, Emil, has confirmed that allocation by moving an unrelated maintenance task to the next cycle. Ordinary meetings and other duties occupy the rest of her day. The allocation covers implementation, testing, documentation, and responding to review; it does not assume evenings or overtime.

Reviewer Ivo has tentatively indicated a two-to-three-working-day queue after the implementation package is ready. That waiting period is an assumption in the initial estimate, not yet a reserved slot. Hana's review-response effort is included in the hour range, and the queue is serial waiting time with no overlapping feature work credited against it.

For calculation only, the team assumes Monday–Friday working days, no holidays or leave, and four usable feature hours every working day. The model rounds focused work up to whole working days, then adds the separate review queue. Real schedules may need a more detailed dependency model. Do not divide a team-sized hour total by a headcount and assume all tasks can run in parallel.

Three estimates for the same feature

The following filled estimate is fictional revision 1. Each range starts when implementation can begin after a separately time-boxed investigation. The investigation itself is not hidden inside the implementation range: it is a four-hour preparation allocation that has not yet been spent at revision 1.

Scenario and condition Remaining feature effort Effort at four hours per working day Separate review wait Modeled elapsed working days
A: Existing parser exposes the required link and fragment information directly 24–32 hours 6–8 days 2–3 days, assumed 8–11 days
B: Existing parser is usable through a compatibility wrapper 36–48 hours 9–12 days 2–3 days, assumed 11–15 days
C: Existing parser requires a replacement adapter for the same agreed checks 52–68 hours 13–17 days 2–3 days, assumed 15–20 days

The team has not assigned probabilities to A, B, and C. These are conditional scenarios, not the minimum, average, and maximum of a validated distribution. No probability has been calibrated for these intervals or for their combined outer endpoints.

Hana's basis is a task breakdown compared with two earlier parser changes in this fictional repository. Those comparisons inform professional judgment; they are too limited to establish a calibrated forecast. The low endpoints assume the expected integration path works with limited correction. The high endpoints allow more implementation and test adjustment within each stated parser condition. None is a worst-case guarantee against every possible discovery.

The initial stakeholder message reads: “For the same link-and-fragment checker, implementation plus review is currently modeled at 8–11, 11–15, or 15–20 working days, depending on parser reuse. Hana's four-hour daily allocation is confirmed; the two-to-three-day review queue is still assumed. We will spend four preparation hours checking the parser interface, then select and revise the applicable scenario. No completion date is committed.”

The three rows tell the stakeholder which parser condition drives the estimate and what the investigation will establish next.

Choose an investigation that can change the estimate

A useful investigation has a question, an owner, a time box, and a recorded result. “Do more research” does not explain which uncertainty will shrink or what happens when the time box ends.

In the fictional case, Hana uses the four preparation hours to examine whether the existing parser returns the link target and heading-fragment information required by the agreed checks. She records that the parser can be reused, but its output requires a compatibility wrapper. She also records a remaining uncertainty about how much adjustment the wrapper's test cases will need.

That result selects scenario B. It does not change the feature's accepted inputs or remove the fragment checks. If the investigation had failed to establish a usable integration path, Hana would retain the unresolved scenarios or propose a further bounded investigation rather than narrowing the range for presentation purposes.

The UK Infrastructure and Projects Authority's cost estimating guidance treats estimates as evolving with scope and evidence and discusses ranges reflecting uncertainty. Its context is major infrastructure cost estimation, not a formula for software duration. The transferable discipline here is to document why the estimate changed, rather than borrowing infrastructure confidence bands or assuming every investigation must produce a narrower answer.

A planner revising a measured wooden pathway after examining the join, with the earlier alternatives retained beside the shared model

A completed revision after new evidence

Fictional revision 2 preserves the same feature and completion condition. The four-hour investigation is now completed work. The remaining implementation estimate for scenario B becomes 40–44 focused hours, with this explicit breakdown:

Remaining activity Focused engineer-hours Basis in the fictional revision
Compatibility wrapper 8 Required interface translation identified during investigation
Link and fragment checks with automated tests 16–20 Known parser output; remaining variation is test adjustment
Integration and usage instructions 12 Existing command path and agreed usage scope retained
Review responses and final corrections 4 Included hands-on response allowance; separate from waiting
Total remaining effort 40–44 8 + 16–20 + 12 + 4

The new lower endpoint rises from 36 to 40 hours because the wrapper work is now concrete. The upper endpoint falls from 48 to 44 because the integration path is understood better. This is a narrower range for the same scenario and scope, rather than a new optimistic scenario replacing the old one.

Ivo also confirms a two-working-day review queue for this planning window, replacing the earlier two-to-three-day assumption. At four hours per working day, 40–44 hours requires 10–11 working days of focused work. Adding the serial two-day queue gives 12–13 elapsed working days from implementation start. The four completed investigation hours bring projected total hands-on effort to 44–48 hours; they are not remaining work and are not added to the future duration again.

The team's fictional stakeholder explanation is complete:

“Revision 2 replaces scenario B's 36–48 remaining hours and 11–15 working days with 40–44 remaining hours and 12–13 working days. The scope is unchanged. We confirmed the need for an eight-hour compatibility wrapper and narrowed test-adjustment uncertainty. Hana's four-hour daily feature allocation remains confirmed, and Ivo has now confirmed a two-day review queue. Four investigation hours are already spent; total projected hands-on effort including that work is 44–48 hours. This remains an estimate, not an approved delivery promise.”

The fictional revision record retains revision 1, the four-hour investigation result, the task breakdown, Emil's capacity confirmation, and Ivo's review-window confirmation. The 36-hour lower bound becomes 40 hours because the wrapper is confirmed; the upper bound falls from 48 to 44 hours as integration uncertainty narrows.

Translate working days into a calendar carefully

Only map a range to dates after checking the actual start, work calendar, leave, and dependency windows. In the simplified fictional model, implementation starts on a Monday labeled Working Day 1. Twelve to thirteen working days places completion on Tuesday or Wednesday of the third working week, counting the start day. Weekend days contribute no feature capacity.

That mapping is illustrative, not a dated commitment. A holiday in the modeled period or a reviewer unavailable in the required window can shift the calendar result. If capacity is confirmed for only one week, do not silently extrapolate it across the entire estimate.

The GAO Schedule Assessment Guide's overview describes a schedule as a model of when activities and major events are expected, and emphasizes analyzing the effects of change. Its government-program context is broader than this small example. The relevant lesson is to expose the time model and dependencies instead of treating a work estimate as a calendar date by itself.

Explain what would change the range or the commitment

Name revision triggers while the estimate is still useful. For this case, the triggers are a different parser interface than the investigation established, additional accepted link formats, more test adjustment than the stated allowance, reduced focused capacity, or a review queue longer than the confirmed window. A new feature request changes scope; it should receive a separate estimate or an explicit rebaseline.

Suppose product owner Noor wants the checker ready by Working Day 10 for an internal demonstration. That is a target. Revision 2's 12–13-day model does not support committing the full reviewed feature by Day 10. Hana explains the mismatch and asks Noor whether to move the demonstration or separately define a smaller demonstration scope. Neither option is treated as approved merely because it appears in the discussion.

A formal commitment requires the appropriate decision owner to accept scope, capacity, dependencies, and the remaining delivery risk through the team's actual process. Calling the lower endpoint a commitment, or the upper endpoint a guaranteed deadline, would erase the uncertainty the range was meant to communicate.

Before sharing your next estimate, state its completion condition, unit, capacity basis, dependencies, and revision trigger. Save the task arithmetic, capacity confirmations, dependency window, and prior revision alongside the updated estimate. Then ask for the planning decision the estimate supports, keeping targets and commitments explicitly separate. More workplace guides are available through LandOffer.

Sources and Further Reading