DEV Community

Qtim
Qtim

Posted on

Before You Estimate Software: Define the Product Boundary and Acceptance Criteria

“We need an account area” can mean a login and three screens. It can also mean roles, password recovery, documents, payments, notifications, audit trails, and a CRM integration. Two teams can hear the same request and price two very different products.

We align on the decisions that change scope and the way the result will be checked before estimating a project. The goal is a shared boundary for the first delivery. Later features can remain open while the team estimates the work in front of it.

Before an estimate, we discuss five things: a user scenario, the first delivery, exclusions, dependencies, and acceptance criteria.

Start with a scenario, not a feature label

Feature names are convenient shortcuts. They rarely tell a delivery team who acts, what they need to do, or what outcome matters.

For example, a document-handling flow can be described through an actor, an action, and a business outcome. This gives the team a starting point for questions without pretending that a generic illustration is a client case.

Who can upload the file? Which document types matter? Is a returned document visible to the supplier? Where does the history live? Some answers become features. Others become constraints or assumptions in the estimate.

A 2025 study in Information and Software Technology of 116 completed software projects found insufficient analysis and planning negatively associated with outcomes in modernization work and new service or product projects. Its sample has clear limits, so it cannot predict the result of one particular project. It does make an important operational point: early analysis belongs to delivery management, not paperwork.

Give the first delivery a visible edge

A first release needs both inclusions and exclusions. Without both, a statement such as “the account area is in scope” leaves space for each participant to fill in the blanks.

For a document-handling flow, the first delivery might include corporate sign-in, a list with statuses, PDF upload, and an in-product notification. It might exclude self-service registration, multi-step approval routes, OCR, email or push notifications, and changes to the CRM itself.

The list does not freeze the product. It turns a new request into a conscious decision. A team and a customer can compare the request with the agreed boundary, then include it now, postpone it, or update the estimate.

Our public services page describes analytics and discovery as an early stage of product work. A useful input for that stage is a user path, first-release functions, exclusions, and external connections. Explore Qtim’s product development services.

List dependencies as assumptions

Dependencies often change an estimate more than one extra interface screen. They include access roles, data sources, API availability, legacy-data migration, document retention, security requirements, and test environments.

Put them in their own section. Mark what is confirmed and what needs a decision. For example: “The CRM provides document status and supplier ID through its current API” is confirmed. “The security team approves the role model” still needs an answer before work begins.

An unknown does not make estimation impossible. It should become an explicit assumption. This makes the estimate easier to revisit when a condition changes.

A 2025 empirical study published by IEEE identifies client engagement, prioritization, and strategic planning among factors that support requirements elicitation. It does not prescribe one process for every team. It supports making open questions visible before work begins.

Separate facts from assumptions

An estimate often contains both. A team may know that an existing identity provider is available. It may not know whether a new user type can be connected to it, whether a legal review is required, or whether the current API exposes the data required by the new flow.

These states deserve different labels. A confirmed fact can be used as an input to an estimate. An assumption should name the condition and what will happen if it changes: “This estimate uses the existing identity provider; a separate access model requires a scope review.” The wording is useful because it identifies the point at which a discussion becomes a change decision.

This also gives a customer a fairer way to compare proposals. The total number is one part of a proposal. The first-delivery boundary, assumptions, exclusions, and change process explain what that number actually covers. Two estimates can look similar while describing different deliveries.

Acceptance criteria are not a design prescription

Acceptance criteria should state the outcome that needs to be observed. They do not need to decide the internal architecture, database schema, or every UI detail before the delivery team has investigated them.

For example, a criterion can require that an authorized user completes an action, receives a visible result, and that the result is available to the person responsible for the next step. The team can still choose the implementation that meets performance, security, and maintainability constraints.

This distinction protects both sides. The customer keeps control over the result that matters. The delivery team can make technical choices with the context required to make them responsibly. When a technical choice changes the visible result or the boundary, it returns to the product discussion instead of quietly changing the scope.

Use five questions in the estimation conversation

Start with the scenario, then move to the first delivery, exclusions, dependencies, and acceptance criteria. If a question has no answer yet, agree how to find it and when to return to it. Then ask the team which assumptions materially affect the estimate.

At the end of the conversation, the output is not certainty about every future release. It is a traceable basis for the next decision: estimate the first delivery, run a short discovery step for the risky part, or postpone a feature whose value is still unclear.

That is why the conversation should return to these questions when a meaningful condition changes. A new integration, a new role, or a new compliance constraint may be a small note in a meeting and a large change in delivery work. Keeping the boundary visible lets the customer and the team discuss that change in the same language.

Write acceptance criteria as observable outcomes

“Make document upload convenient” is difficult to test. Different people will picture different results.

For a document-handling flow, acceptance criteria can state who has permission to act, which inputs are allowed, which result appears afterward, who can see or change the status, and which event is recorded. Another product may require offline work, a qualified signature, several languages, or response-time limits. The useful property is testability: each condition can be demonstrated or tested.

It also helps to keep two lists. One covers the behavior of a particular scenario. The other covers delivery-wide conditions such as compatibility, security, documentation, migration, monitoring, and release readiness.

A 2026 systematic review in Information and Software Technology examines clarity, ambiguity, and testability as quality aspects. The product boundary remains a decision for the people responsible for the delivery.

Five questions before an estimate

Discuss these five questions in an estimation conversation:

  1. The user scenario and intended outcome.

  2. Features and platforms included in the first delivery.

  3. Explicit exclusions.

  4. Dependencies and assumptions.

  5. Observable acceptance criteria.

If one of the questions has no answer, agree how to get it and when to return to it. “Who approves access roles?” then stops being an unspoken line in the estimate.

The product will change anyway. These agreements make change visible. They separate a new decision from the normal work required to deliver the decision already made.

A 2025 comparative study in the Egyptian Informatics Journal identifies ambiguous requirements as the leading cause of prolonged analysis and project failure in its comparison. The finding does not assign a universal cost to ambiguity. It is a reason to surface ambiguity before work begins.

Return to the account area. The team still does not have a finished product after this conversation. It has a sound object of estimation: a supplier scenario, a first-delivery boundary, named dependencies, and a result that can be checked.

An estimate has no single object until a feature has a scenario, a boundary, dependencies, and a testable outcome.

Qtim can help map the first delivery before you request a quote. Discuss your product scope with our team.

Top comments (0)