Build log/Working with us

Discovery phase in software development: what happens and what you get

A discovery phase is short, fixed work before a build that turns an idea into a scope, a prototype, an estimate and a risk list, so you decide with evidence before committing the full budget.

The short answer

  • A discovery phase, sometimes called a discovery sprint, is a short piece of work before development that defines what to build, for whom, how, and roughly what it will cost.
  • At the end you should have a written scope, a clickable prototype of the main journeys, a technical approach, a ranged estimate and a list of risks with a plan for each.
  • For a new product, discovery usually takes a few weeks, and it saves money by finding expensive surprises while they are still cheap to change.

A discovery phase is a short, fixed piece of work at the start of a software project that turns an idea into a plan you can build from. It answers what to build, for whom, how, and roughly what it will cost, and it ends with a scope, a prototype, an estimate and a list of risks. It saves money because changing a decision on paper costs far less than changing working software.

What a discovery phase is

UK government services use discovery as the first of four phases, followed by alpha, beta and live, as described in the GOV.UK Service Manual. Studios and agencies use the same word for the work before a build. You may also hear it called a discovery sprint, an inception phase or a scoping phase.

Discovery is a few weeks of interviews, sketching, prototype testing and technical checks, done with the people who will use and pay for the product. Its outputs are short and practical, written so that a team can build from them.

What happens during discovery

  1. Kick-off. Agree the goal, the decision the discovery will support, the budget range and who decides.
  2. Interviews and observation. Talk to the people who will use the product, and watch the current process where you can. For construction software, that can mean standing on site while someone fills in a paper diary.
  3. Map the current process. Gather the spreadsheets, forms and emails people use today. They show the real data and the real exceptions.
  4. Define users and priorities. List the types of user and what each needs to do, then sort features into must have, should have, could have and won’t have for now, a method often called MoSCoW.
  5. Prototype and test. Build a clickable prototype of the main journeys and put it in front of real users.
  6. Technical investigation. Check integrations, data migration, hosting, security, and constraints such as working without signal (offline-first mobile) or rules for children’s data.
  7. Estimate and plan. Turn the agreed scope into an estimate, a release plan and a risk list.

What you should get at the end

A discovery phase should leave you with documents you own and could hand to any competent team.

OutputWhat it containsWhy it matters
ScopeUsers, features for the first release, what is out of scope, acceptance criteriaSets out what you are paying for
PrototypeClickable screens for the main journeys, tested with usersShows the product before it is built
Technical approachPlatforms, architecture, hosting, integrations and an outline data modelExposes hard problems early
EstimateA range of cost and time for the first release, with its assumptionsLets you set a budget and choose a contract model
Risk listEach risk, its likely impact and how to reduce itShows where the estimate could move
Release planWhat ships first and what followsGets something useful live sooner

The estimate should be a range with its assumptions written out. A single precise figure at this stage hides uncertainty that still exists.

A tested prototype is useful beyond the build, too. You can show it to investors, use it in early sales conversations, or put it in front of a few target customers to check they would pay.

How long discovery takes

For a new product, discovery usually takes a few weeks. A small product with one type of user and no integrations might need a week or two. A platform with several types of user, payments, integrations with existing systems or regulated data can take six weeks or more. The things that change the length most are these.

  • The number of user types and journeys.
  • Integrations, especially with older systems that have little documentation.
  • Regulation, such as UK GDPR, the ICO’s Children’s code for services children are likely to use, or CIS and VAT rules in construction.
  • How quickly decision makers and users are available.

A discovery that is short for the size of the product still produces a useful estimate, with wider ranges and more assumptions. That is fine, as long as everyone reads the estimate that way.

Why discovery saves money

The later a wrong assumption is found, the more it costs to fix. A changed decision during discovery means redrawing a screen. The same change after launch can mean migrating data, rewriting features and retraining users.

Discovery finds the expensive surprises while they are still cheap. Typical examples are an accounting system without the API the product depends on, users who work underground with no signal, a feature the founder assumed was essential that users don’t want, and a requirement that brings extra legal obligations. Estimark works offline on site, saving to the phone first and syncing later. A requirement like that shapes the whole architecture, so it has to be known before the build starts. CurioKids takes bookings in each family’s own time zone, which is straightforward to plan for from the start and fiddly to add to a booking system that assumed a single time zone.

Discovery also gives you a clean point to decide. You can go ahead, cut the scope to fit the budget, or stop, having spent a small part of what the build would cost. A defined scope also makes fixed-price quotes possible (fixed price or time and materials).

How to get the most from discovery

  • Make the person who will make decisions available every week.
  • Bring real material, such as spreadsheets, forms, quotes, invoices and emails.
  • Arrange access to the people who do the work day to day, as well as their managers.
  • Share your budget range, so the scope can be shaped to it.
  • Say what you already know is fixed, such as a launch date or a system you must keep.
  • Agree how outputs will be shared, so you can review the scope and prototype as they develop and comment early.

Who owns the discovery outputs

Make sure the contract gives you ownership of everything discovery produces, including the prototype, documents and designs. Then you can take them to any supplier, use them to raise money, or build the product in-house. Who owns the code covers how ownership of software and designs works in the UK.

Have an idea to test before you build? See how we scope an MVP.

Questions people ask

Is a discovery phase the same as a design sprint?

A design sprint is a specific five-day format for exploring and testing one idea with a prototype. A discovery phase is broader, covering users, scope, technical approach, estimate and risks. A discovery phase may include a design sprint as one part of it.

Can I skip discovery?

You can, if the product is small and well understood, or if you already have a tested prototype and a detailed specification. Skipping it on a larger or less certain project usually moves the same questions into the build, where answering them costs more.

Who should take part in a discovery phase?

On your side, the person who makes decisions about the product, someone who knows the current process in detail, and a few future users. On the supplier’s side, usually a product or delivery lead, a designer and a senior developer, so that design and technical questions are answered together.

Is a discovery phase usually fixed price?

Usually, yes. Discovery has a defined length and a defined set of outputs, which suits a fixed price. The build that follows may then be priced as fixed price or as time and materials, depending on how settled the scope is.

What happens if discovery shows the idea won’t work?

Then discovery has done its job at the lowest possible cost. The outputs often point to a smaller or different product that would work, or show that the idea needs more testing with customers before anything is built. Either way, you keep the research, prototype and documents.

Written by Flavoureak

We build and run Estimark, CurioKids and CalcStack, and we build web and mobile software for other businesses. Start a project.