StackLift
All playbooks

Playbook

Software Estimation Playbook

A good software estimate connects business intent to scope, architecture, risks, delivery phases, and change rules. It should be a delivery baseline, not a loose guess.

Best reader

Teams comparing software build options and budget ranges

Outcome

A structured estimate that is easier to review, approve, and execute.

Use this sequence

1

Capture the project intent and target user groups.

2

Group features into must-have, should-have, and later phases.

3

Map integrations, data, roles, and non-functional requirements.

4

Score delivery risk before committing to a number.

5

Convert the approved scope into phases and acceptance criteria.

Estimate the workflow, not a feature list

Feature lists hide important cost drivers. Workflows reveal roles, edge cases, data movement, and business rules.

Primary user journey

Admin and support flows

Exception paths

Reporting and audit needs

Separate range from commitment

Early estimates should be ranges. Commitment comes after architecture review and assumption validation.

Low and high range

Assumptions

Exclusions

Decision points

Make changes explicit

A scoped estimate should explain how future changes are handled so budget does not drift silently.

Baseline scope

Change request process

Approval owner

Impact on timeline

Start in 60 seconds

Turn this checklist into a scoped plan.

Answer a few questions and get an AI-assisted, architect-reviewed scope, cost range, and timeline for your software project.

Frequently asked questions

Why do software estimates vary so much?

Because scope, integrations, roles, data, security, and quality expectations vary. A structured estimate makes those drivers visible.

Can AI create a reliable estimate?

AI can create a strong first pass, but the estimate becomes reliable after senior architect review.

The complete guide

Software Estimation Playbook: the complete guide

Updated November 18, 2026 · Published by , Founder, StackLift AI. Written with AI assistance; how we make content.

Short answer

To estimate a software project, write down the intent and users, group features into must-have, should-have and later, map integrations, data, roles and quality needs, score delivery risk, and then turn the approved scope into phases with acceptance criteria. Give a range early, commit only after architecture review, and record how changes will be handled.

What is a software estimate, and what should it contain?

A software estimate is a reasoned prediction of what it takes to build something: the scope, the effort, the cost, and the time. It is not a promise. A useful one also says what it assumes, what it leaves out, and what could move the number.

The software estimate glossary entry gives the short definition. In practice, a defensible estimate has six parts, listed below. If a quote you receive is missing most of them, you are looking at a guess.

  • Scope: the features and workflows included, in plain language.
  • Assumptions: what you and the team believe to be true, such as access to an API or a decision-maker's availability.
  • Exclusions: what is deliberately not included.
  • Range: a low and a high figure for cost and time, not a single number.
  • Risks: the factors most likely to push toward the high end.
  • Change rules: how new requests are priced and approved.

How do you estimate a software project step by step?

The five steps below turn an idea, a requirements document, a backlog or a reference product into an estimate you can review. Each step has an output, so you can see whether it was really done.

  1. Capture the project intent and target user groups. Output: a one-page summary of the goal, the users, and what success means.
  2. Group features into must-have, should-have and later. Output: a prioritized scope list with a first release defined.
  3. Map integrations, data, roles and non-functional requirements. Output: a systems list, a data and permissions outline, and quality targets.
  4. Score delivery risk before committing to a number. Output: a risk register with the top items and how each could change cost or time.
  5. Convert the approved scope into phases and acceptance criteria. Output: phases, deliverables and the tests that define done.

If you want a fast first pass, the free estimate walks you through the same inputs feature by feature and returns cost, timeline and team in about five minutes, with no signup.

Why estimate the workflow instead of a feature list?

A feature list says "user accounts, dashboard, reports." It hides the cost. A workflow says who does what, in what order, with which data, and what happens when something goes wrong. That is where roles, exceptions, integrations and business rules show up.

What a workflow view reveals that a feature list hides
Look atWhat it revealsCost effect
Primary user journeyScreens, validation and data per stepCore build effort
Admin and support flowsBack-office tools nobody listedOften a large share of the work
Exception pathsRefunds, failures, edits, approvalsRules and testing effort
Reporting and audit needsData models and history trackingDatabase and export work
Roles and permissionsWho can see or change whatSecurity design and testing

The related approach in the production readiness checklist is worth reading next. It lists quality needs that estimates forget, such as monitoring and rollback.

Which estimation methods exist, and when does each fit?

Teams use several techniques, and the better ones combine two and compare the results. People who will do the work should be part of the estimate, since estimates prepared by others tend to be less accurate. Historical data from similar projects helps most.

Common estimation methods compared
MethodHow it worksBest forWatch out for
AnalogyCompare with a similar finished projectEarly ranges when you have historyDifferences that look small but are not
Bottom-upBreak work into tasks and sum themWell-defined scopeMissed work between tasks
Top-downStart from a total and divide itBudget-first planningHides complexity
Three-pointBest, likely and worst case per itemItems with real uncertaintyOptimism in the best case
Story pointsRelative complexity, calibrated by the teamOngoing delivery planningTreating points as hours
Expert judgmentSenior people size the workUnusual or new problemsIndividual bias; use more than one voice

StackLift prices by deliverable complexity, measured in story points, not by hours. Scope, budget and delivery path are agreed before engineering begins.

What is the difference between a range and a commitment?

Early in a project, almost nothing is known, so any honest number is a wide range. As requirements firm up and risks are tested, the range narrows. This idea is often called the cone of uncertainty. Asking for a tight number on day one does not make the project more predictable. It only hides the width.

A commitment is a promise to deliver a defined scope by a date for an agreed price. It belongs after architecture review and after the biggest assumptions are validated. An estimate becomes a commitment when someone senior has examined the risks and agreed to stand behind it.

  • Range: low and high figures, used for budgeting and comparing options.
  • Assumptions: the conditions the range depends on, with an owner for each.
  • Decision points: moments when you confirm scope and the range narrows.
  • Commitment: a baseline scope, price and timeline, approved in writing.

Every StackLift estimate is reviewed by a senior architect before it becomes a commitment. The estimate, proposal, project and client portal share one baseline, so later change is explicit.

How do you score delivery risk before committing to a number?

Risk is why two estimates for the same idea differ. Score each factor as low, medium or high, and say how it could change the range. Do this before you price, not after the first surprise.

Risk factors to score
FactorLow risk looks likeHigh risk looks like
RequirementsWritten, stable, agreed by one ownerVerbal, shifting, several owners disagree
IntegrationsDocumented API with sandboxUndocumented or legacy system, no test access
DataClean, owned, accessibleScattered, dirty, permissions unclear
TechnologyWell-known stack the team has usedNew platform or unproven AI behavior
ComplianceRules identified and plannedUnknown rules, late review
Decision speedNamed approver, quick answersCommittee sign-off, slow feedback

High scores do not mean you should not build. They mean you should plan a spike, a prototype, or a Scope Sprint to reduce the uncertainty first.

How do you turn the estimate into phases and acceptance criteria?

A phase is a chunk of delivery that ends with something usable and a decision. Acceptance criteria are the checks that decide whether a deliverable is done. Without them, "done" becomes an argument.

  • Name each phase by outcome, such as "customers can sign up and pay," not by activity.
  • List deliverables for the phase and the tests that prove each one.
  • Include non-feature work in the phase: CI/CD, monitoring, security review and rollback.
  • Mark the decision at the end: continue, adjust scope, or stop.
  • Keep the first phase small enough to deliver in weeks, not quarters.

StackLift offers delivery plans in the estimate: MVP-first, Standard, Fast Track and Priority. Most first versions take 6–16 weeks depending on scope. To see how a first release is shaped, read MVP vs. full build.

How should changes be handled so the budget does not drift?

Scope will change. The question is whether it changes in the open. A good estimate names the baseline scope, who may request a change, who approves it, and how a change affects time and price. See the change request glossary entry.

A simple change-control template
StepWhoOutput
RequestAnyone on the projectA short description of the change and why
AssessArchitect or leadEffect on scope, cost, time and risk
DecideNamed approverApprove, defer or decline, recorded in writing
Update baselineDelivery leadA new shared baseline everyone can see

Compare billing models before you sign. Fixed scope vs. hourly billing explains how each one handles change.

What mistakes make estimates wrong, and when should you get help?

  • Estimating only the happy path and ignoring exceptions.
  • Forgetting non-feature work: testing, deployment, monitoring, documentation, handover.
  • Treating a range as a promise, or demanding one number too early.
  • Letting the person who will not do the work set the estimate alone.
  • Skipping a risk review on integrations and data.
  • Not recording assumptions, so no one notices when they stop being true.
  • Leaving out running costs such as hosting, support and AI usage; check model prices.

Bring in help when the scope is large, when several systems must connect, when compliance applies, or when you have received quotes that differ widely and cannot tell why. Our cost guides show the drivers by project type, and you can contact us to talk through a scope.

Questions people ask

Why do software estimates vary so much between vendors?

Because scope, integrations, roles, data, security and quality expectations vary, and vendors make different assumptions about each. A structured estimate makes those drivers visible, so you can compare like with like instead of comparing totals.

Can AI create a reliable software estimate?

AI can produce a strong first pass quickly by breaking a project into features and drivers. It becomes reliable after a senior architect reviews the assumptions and risks. StackLift's estimate works that way: AI drafts, an architect reviews.

How accurate can an early estimate be?

Early estimates should be ranges because little is known yet. Accuracy improves as requirements, integrations and risks are confirmed. If someone gives you one precise figure from a short call, ask what it assumes.

What is the difference between story points and hours?

Story points measure relative complexity of a deliverable, while hours measure time spent. Pricing by complexity keeps the focus on what you receive rather than how long the work takes. The two should not be converted one to one.

What should I give a vendor so I get a good estimate?

Share the goal, users, main workflows, systems to connect, data sources, compliance needs and any deadline. For AI projects, use the AI project discovery checklist first. Gaps are fine if you label them.

Is a fixed price always safer than time and materials?

Not always. Fixed scope works when the scope is clear and changes go through a defined process. When the problem is still being discovered, a phased approach with a clear first deliverable can be safer than pricing everything up front.

When does an estimate become a commitment?

After the scope is baselined, the major risks have been reviewed by a senior architect, and the price, timeline and acceptance criteria are agreed in writing. Before that point it is a range for planning.