The short answer
A discovery phase is a short, fixed piece of work at the start of a software project that turns a business problem into a defined scope, a technical approach, a set of risks and a reliable estimate, before anyone commits to the full build. For a commercial project it typically runs 2 to 6 weeks and, in the Australian market, commonly costs around $10k to $25k (AUD, ex GST), more for large enterprise systems. Its main job is to replace a guess with a price you can hold a supplier to.
Key takeaways
- Discovery answers what to build, for whom, how, and what it will cost, before the expensive part starts.
- Outputs are concrete: a prioritised scope, user flows or prototypes, an architecture outline, an integration and data map, a risk register, and an estimate or fixed-price proposal.
- Government guidance puts discovery at 4 to 8 weeks (UK GDS) or 6 to 8 weeks (Australia's DTA); a defined commercial product usually needs 2 to 6 weeks.
- Typical Australian market range is about $10k to $25k ex GST, driven by team size and duration; this is a range, not a quote.
- A good discovery can end with a recommendation not to build, which is a cheap outcome compared with a failed project.
What is a discovery phase?
A discovery phase is a short, time-boxed piece of work that defines what a software project should build, how, and at what cost, before the build is committed. Discovery in software development is sometimes called scoping or inception. It sits between “we have an idea or a problem” and “we’ve signed a contract to build it”.
The UK Government Digital Service describes discovery as understanding the problem before committing resources to a solution, and is firm that you “should not start building your service” during it. Australia’s Digital Transformation Agency takes the same line in its service design and delivery process: prototyping and testing belong to the next stage, not discovery.
Commercial software projects borrow the idea but usually aim it at a narrower target: a scope and price precise enough to sign.
What happens during the discovery process?
The discovery process moves from understanding the problem, to examining the real systems and data, to designing and pricing a solution. A typical sequence for a commercial project:
- Kick-off. Agree the goals, the decision maker, who needs to be interviewed and which systems and data are in scope.
- Interviews and observation. Talk to the people who do the work today, not only managers, and watch the current process, including the spreadsheets and workarounds.
- Systems and data review. Look at the actual APIs, databases, documents and access arrangements that the new software will depend on.
- Scope and journeys. Write up the user groups, key journeys and a prioritised feature list with acceptance criteria.
- Prototype. Sketch or click through the main screens with users and adjust.
- Architecture and risks. Outline hosting, components, security and data residency, and list the known unknowns. For AI work, run a small feasibility test on real data.
- Estimate and playback. Present the scope, plan and price, and agree what happens next.
What does a discovery phase produce?
A set of documents and designs that someone other than the author could use to build the system and price it. If you can’t hand the outputs to a second supplier for a quote, the discovery wasn’t finished.
| Output | What it contains | Why it matters |
|---|---|---|
| Problem and outcomes statement | The business problem, who has it, and measurable success criteria | Stops the project drifting into building features nobody needs |
| User groups and key journeys | Who uses the system and the main tasks each performs, often as flow diagrams | Reveals the real workflow, including the workarounds |
| Prioritised scope | Features ranked must, should, could, won’t, with acceptance criteria | Becomes the contract scope for a fixed price |
| Clickable prototype or wireframes | Screens for the main journeys | Cheap to change now, expensive later |
| Integration and data map | Every system to connect, how, who owns it, and data quality notes | Integrations are the most common source of overruns |
| Architecture outline | Hosting, main components, security approach, Australian region choices | Surfaces cost, compliance and residency decisions early |
| Risk register | Known unknowns with a plan for each | Tells you where the estimate could move |
| Estimate and delivery plan | Phases, milestones, team, timeline and a price or range | The decision you actually came for |
For AI projects, add an evaluation plan and often a small technical spike to test whether the model can do the task on your real data. Our guide to evaluating an LLM application explains why that test set should exist before development starts.
How long does discovery take?
For a defined commercial product, usually 2 to 6 weeks; government service discovery runs longer because it involves more user research. The UK Service Manual says “around 4 to 8 weeks is typical”, and the DTA says discovery “usually takes between 6 and 8 weeks, depending on the size and complexity.”
| Project type | Typical discovery length | Main driver of time |
|---|---|---|
| Internal tool replacing a spreadsheet process | 1 to 2 weeks | Mapping the current process |
| Customer-facing app or portal with 1 to 2 integrations | 2 to 4 weeks | User journeys and integration detail |
| AI assistant or RAG knowledge base on company documents | 2 to 4 weeks | Data audit, access rules and a feasibility spike |
| Legacy system replacement | 4 to 8 weeks | Understanding what the old system actually does |
| Government digital service | 6 to 8 weeks (per DTA guidance) | User research across all user groups |
These are typical ranges based on the published guidance and common market practice, not fixed rules. Discovery overruns most often because of slow access to people, systems or sample data, not because of the analysis itself.
What does a discovery phase cost?
In the Australian market, commonly around $10k to $25k (AUD, ex GST) for a commercial project, and more for large enterprise or legacy work. Treat this as a typical range, not a quote. VT Digital’s 2026 enterprise cost guide cites $15,000 to $25,000 for a discovery that produces “proper specs and accurate numbers”.
The arithmetic is simple: discovery is mostly senior time. Re:Sourced’s 2026 guide puts senior software engineer contract rates in Sydney at about $800 to $1,100 a day, ex GST, and Robert Walters’ 2026 salary guide lists NSW senior full stack contractors at $800 to $1,000 a day. Agencies charge more than contractor rates to cover overheads.
| Discovery shape | Effort | Illustrative cost (AUD, ex GST) |
|---|---|---|
| Small: 2 weeks, one senior engineer plus some design time | about 12 to 14 days | $11k to $18k |
| Medium: 3 to 4 weeks, senior engineer and designer or analyst | about 20 to 25 days | $18k to $32k |
| Large: 6 to 8 weeks, small team including an architect | about 45 to 60 days | $40k to $80k |
Day rates assumed: roughly $900 to $1,300 blended, a mix of contractor and agency rates, multiplied by the days shown. Your supplier’s rates and team shape will differ; ask them to show the same breakdown.
As a share of the total, discovery is usually a small fraction of the build it defines. On a $150k project, a $15k discovery is 10%. The comparison that matters is with the contingency a supplier would otherwise add to a vague brief, and the cost of building the wrong thing. For wider budgeting, see how much custom software costs in Australia.
Why does discovery make a fixed price possible?
Because a fixed price is only safe for both sides when the scope is defined, and discovery is what defines it. Without it, a supplier quoting a fixed price has two options: add a large contingency to cover what they don’t know, or quote low and recover the difference through change requests.
Discovery changes the maths in three ways:
- Scope becomes testable. Features have acceptance criteria, so both sides can tell when something is done.
- Unknowns become named risks. Each can be priced, tested with a spike, or explicitly excluded.
- Integrations are understood. The supplier has seen the APIs, data and access arrangements rather than assuming them.
This is the logic behind “paid discovery, then fixed price”. It doesn’t suit every project: where requirements will keep changing as you learn, time and materials may be the better model. Our comparison of fixed price vs time and materials covers when each fits.
What makes a discovery good or bad?
A quick checklist for judging a proposal or a finished discovery:
- Fixed duration and fixed fee, stated before it starts
- Named senior people doing the work, not only a salesperson
- Your staff who do the work today are interviewed, not just managers
- Real data, documents and systems are examined, not described
- Outputs are yours to keep and specific enough for another supplier to quote from
- Risks and exclusions are written down
- Includes an honest option to stop or change direction
- Ends with a clear price or range and a phased plan
Warning signs: a discovery that produces only a slide deck, one where every recommendation needs the same supplier’s proprietary platform, or one that can’t say what would make the project a bad idea.
What comes after the discovery phase?
A decision: proceed to build, change direction, or stop. If you proceed, the discovery outputs become the basis of the build contract, usually a fixed-price proposal for the defined scope, followed by detailed design and development in stages.
In government, the next stage has a name. The DTA’s service design and delivery process moves from discovery to alpha, where teams prototype and test the ideas formed in discovery, before beta and live. In commercial projects the steps after discovery usually look like this:
| After discovery | What happens |
|---|---|
| Decision | The business reviews the scope, price and risks and decides whether to go ahead |
| Contract | A build contract or schedule of work is signed against the discovery scope |
| Design | Detailed UI design and technical design for the first stage |
| Build in stages | Development in milestones, each with a demo and acceptance against the agreed criteria |
| Launch and support | Release, handover, then maintenance and improvement |
Should discovery ever end with “don’t build”?
Yes, and it’s one of the most valuable outcomes. The UK Service Manual lists stopping as a legitimate end to discovery and notes that it saves resources. Sometimes the research shows an off-the-shelf product fits, a process change solves the problem, or the benefit doesn’t justify the cost. Learning that for $15k is far better than learning it after spending $300k. Our build vs buy guide helps frame that decision.
How All Webbed Labs approaches this
Every project we take on starts with a paid discovery at a fixed fee, run by the senior engineers who would do the build, under NDA. It ends with a written scope, designs, an architecture outline in Australian cloud regions by default, a risk register and a fixed-price proposal. You own the outputs whether or not you proceed with us. For AI projects, our AI readiness assessment is a productised version of the same process. See how the rest of delivery works on our methodology page, or start with custom software development.
Frequently asked questions
Do we own the discovery outputs?
You should. Ask for the scope document, designs, architecture notes and estimates to be yours to keep, so you can take them to another supplier if you choose. A discovery that only works with one vendor is a sales exercise.
Can discovery be skipped for a small project?
For a genuinely small, well-understood job, a few workshops and a written scope may be enough. The risk rises with integrations, regulated data, AI components and multiple user groups, which is where a proper discovery pays for itself.
Is the discovery fee credited against the build?
Some suppliers credit it, others don't. Either model is fine as long as it's stated up front. What matters more is that the outputs are useful to you whether or not you proceed with the same supplier.
What do you need from us during discovery?
Access to the people who do the work today, a decision maker who can settle scope questions quickly, samples of real data and documents, and access to any systems that need to be integrated. Slow access is the most common reason discovery overruns.
Is a discovery phase the same as a scoping phase or product discovery?
Largely, yes. Suppliers call it discovery, scoping, inception or a discovery workshop series. "Product discovery" is sometimes used more narrowly for ongoing research into what users need in a product team. Whatever the name, check what it produces and whether you can take the outputs to another supplier.
How is discovery different from a proof of concept?
Discovery works out what should be built and what it will take. A proof of concept tests whether one specific technical idea works at all. For AI projects, a small proof of concept is often part of discovery when the key risk is whether the model can do the task well enough.
Prices on this page are typical Australian market ranges for planning purposes, not quotes. Every project is priced after scoping.
Sources
- How the discovery phase works , UK Government Digital Service, Service Manual
- Discovery stage: exploring the problem , Digital Transformation Agency
- Enterprise Software Development Cost in Australia: 2026 Guide , VT Digital
- Contractor day rates in Australia: 2026 guide , Re:Sourced
- Salary Guide 2026 (Australia, mid-year edition) , Robert Walters