Back to all posts
Process7 min read·October 25, 2026

How to Estimate a Software Project Without Guessing

TB
ThynkBlox Team
Engineering

To estimate a software project, define what must work, break it into testable pieces and estimate each piece as a range. Include design, testing, release work and specific allowances for unknowns, then map that effort onto the people and dependencies involved. Treat the result as a forecast with stated assumptions, and update it as those assumptions are tested.

Why do software estimates fail?

An estimate becomes misleading when its scope or assumptions stay hidden. “Build customer accounts” might mean a sign-in screen to one person and invitations, permissions, account recovery and account deletion to another.

A second problem is confusing effort with elapsed time. Work that takes several days of focused effort can span longer when it depends on feedback, access or another supplier.

A third is estimating only the visible feature. Ask whether the proposal includes error handling, testing, accessibility, deployment and corrections after acceptance testing.

Start with this question: What exactly does the estimate promise to include? A precise total cannot compensate for an unclear answer.

What should you define before asking for an estimate?

Write a short scope brief that explains the outcome, the users and the boundaries. Describe what someone must be able to do, rather than listing screens alone.

For example, “a customer can reset a forgotten password and regain access” is more testable than “password reset page.”

Your brief should cover:

  • Outcome: What problem should the release solve?
  • Users: Who uses it, and what may each person do?
  • Included journeys: Which actions must work from start to finish?
  • Acceptance conditions: What observable results count as complete?
  • Constraints: Which devices, integrations and existing systems must be supported?
  • Exclusions: What will this release deliberately leave out?
  • Dependencies: What access, content or decisions must someone else provide?

If important interactions remain unclear, include design work before committing to the full build. Why startups need UX before code explores that decision.

How do you break the project into work you can estimate?

Split the project into pieces with a clear result and a clear way to check it. Keep breaking down an item when it contains several unrelated decisions or nobody can explain what finishing it involves.

For a booking product, “booking management” is too broad for a useful discussion. Separate viewing available slots, reserving a slot, cancelling a booking and managing availability.

Then include the supporting work: permissions, notifications, failure handling and testing. Give shared work its own place so it is neither omitted nor counted under every feature.

Ask the people expected to do the work to explain each range. If they cannot identify what makes an item straightforward or difficult, it needs more definition or investigation.

Keep the first release focused on a complete user journey. How long it takes to build an MVP provides related guidance on release scope.

Why should you use ranges instead of a single number?

A range communicates that the outcome depends on conditions. For each item, write a lower and upper estimate alongside the assumptions that separate them.

Define what the bounds mean. For example, the lower end might assume the documented integration behaves as expected; the upper end might include adapting to missing fields and inconsistent responses.

Do not describe the upper bound as protection against every possible event. A different integration, a new user role or a redesigned journey changes the scope and requires another estimate.

Use these questions to explain each range:

  • What must be true for the lower end to hold?
  • What work accounts for the upper end?
  • Which events fall outside both bounds?
  • What could we check now to narrow the range?

Avoid presenting a midpoint as a promise. Keep the range and its conditions visible wherever the estimate is shared.

What does a worked estimate look like?

Illustrative example only: suppose a small booking release has an agreed design, one user role and no payment collection. The figures below are invented to demonstrate the method, not benchmarks.

An effort day means one person's focused working day. It does not mean a calendar day.

Work itemIllustrative effort rangeMain assumption
Confirm journeys and acceptance conditions2–3 daysDecision-maker available
Build availability and booking flows5–8 daysBooking rules agreed
Add cancellation and notifications3–5 daysNotification service accessible
Test journeys and correct defects3–5 daysTest accounts and environments ready
Prepare release and handover1–2 daysRelease access available
Named integration contingency1–3 daysExisting service needs limited adaptation
Total effort15–26 daysListed scope and assumptions hold

The total is a sum of effort ranges, not a delivery schedule or a statistically calculated confidence interval.

To propose a delivery window, arrange the work in dependency order and check actual availability. Testing may overlap with development, while final release checks depend on completed journeys. Ask the partner to show that sequence.

How should you allow for unknowns?

Separate work you know you must do from uncertainty about how much work it will take.

Testing is planned work. An undocumented integration response is an uncertainty. Do not put both into an unexplained “buffer.”

Keep a short risk list with an owner, a possible effect and a next action. Where an unknown could change the approach entirely, estimate a limited investigation first. Its output should be evidence: a tested connection, a sample import or a confirmed constraint.

For remaining uncertainty, attach contingency to a named risk. Explain what it covers and what would trigger a revised estimate.

Avoid counting the same risk twice through both a wide task range and an additional allowance. Also keep waiting time visible in the schedule: more development effort does not resolve a delayed approval.

Should you choose fixed price or time and materials?

Choose by considering how much scope can be agreed and how much flexibility you need. Check the written terms; the contract label alone does not define the arrangement.

QuestionFixed priceTime and materials
What is agreed?A price for defined deliverables, subject to the termsPayment for agreed time and resources used
What needs careful definition?Scope, acceptance, exclusions and change handlingRates, reporting, spending limits and stop conditions
How should changes work?Assess their effect before authorising themReprioritise work and review the remaining forecast
What should you monitor?Accepted deliverables and unresolved scope questionsCompleted outcomes, time used and remaining work

Consider fixed price when deliverables and acceptance conditions are sufficiently clear to agree in writing.

Consider time and materials when learning and reprioritisation are expected, with regular demonstrations and a spending limit. You can also commission a defined investigation before choosing the build arrangement.

What common mistakes should you avoid?

Dividing effort by headcount. Check which tasks can actually run together before shortening the schedule.

Hiding essential work to meet a deadline. Reduce scope explicitly. Keep testing, release preparation and other agreed completion conditions visible.

Comparing proposal totals without comparing scope. Check whether each includes the same journeys, responsibilities and support.

Treating future maintenance as included by default. Agree what happens after launch and estimate ongoing work separately.

Keeping an estimate after its assumptions change. Record what changed and forecast the remaining work again. Do not silently stretch the original commitment.

Before accepting a proposal, confirm that you can explain its scope, exclusions, ranges, dependencies, contingency, acceptance conditions and change process.

What questions do founders ask about software estimates?

Can you estimate a project before every detail is known?

Yes, as an early planning range. List unresolved decisions and identify which could materially change the scope. Narrow the estimate after those questions are answered.

What if the deadline cannot move?

Work backwards from the deadline. Identify the smallest complete outcome, reserve time for verification and release, and agree what can be removed. Ask whether the resulting scope fits the available capacity.

How often should an estimate change?

Review it when scope changes, an assumption fails or completed work provides useful evidence. Show work accepted, work remaining and the reason for any movement in the forecast.

What should you ask a development partner to explain?

Ask them to walk through one uncertain item: its boundaries, dependencies, range and next check. Then ask how they will report progress and seek approval for changes. Questions to ask a software development agency can help structure that conversation.

To decide how much reserve to hold for the uncertain items, see how much contingency a software budget should include.

Ready to build?

Let's turn these ideas into your next product.

Start your project