Skip to main content
Resources

Bespoke software buying & planning

How to choose a software development company in the UK

Choose a software development partner by comparing business understanding, scope, relevant work, ownership and support. Includes a practical supplier example.

Veda Software desk and laptop cover for How to choose a software development company in the UK

10 minute read · Veda Software · · Updated

Contents

A good software development company should understand the problem you are trying to solve, explain what you are buying and show relevant work you can assess. Whether you are improving an established business or commissioning your first product, compare suppliers against the same brief before comparing their prices. This guide covers the questions to ask about scope, delivery, ownership and support, then uses a fictional buying example to show how to judge the answers.

What to look for in a software development partner

Start with fit for your project. Ask each company to explain the business problem in its own words, identify the people who will work with you and show relevant work it is authorised to share. A polished portfolio is a starting point; discuss what the team actually delivered and how that experience relates to your requirement.

Use five questions for your shortlist: what will the first release achieve; what is included in the price; who makes decisions and keeps you informed; what will you own at handover; and who supports the software after launch? Ask for clear exclusions and dependencies as well as promised features. This gives both an established business and a founder a common basis for comparing offers.

For a UK engagement, confirm the contracting organisation, working arrangements, availability for decisions and the people responsible for delivery. A location or company label alone does not establish project fit. Check that the proposed way of working suits your team, rather than assuming every supplier will provide the same level of product guidance or ongoing support.

Start with the responsibility you need to buy

Decide whether you need someone to define a product, deliver a known scope, add capacity to your team or take over an existing system. Name the decisions your organisation will retain and those you expect the supplier to own. A team offering implementation capacity is not automatically taking responsibility for discovering the right workflow, coordinating suppliers or operating the application after launch.

Write a short common brief before comparing proposals: the users, the outcome, existing systems, access constraints and the next decision you need to make. Include the inconvenient facts, such as a supplier that cannot yet offer a test environment. These are useful comparison inputs. A long feature list with those uncertainties omitted can make estimates appear comparable when they rest on different assumptions.

Give both suppliers the same small workflow

Our fictional buyer wants a service-request portal. An employee submits a request, their manager approves it and an external system receives the approved instruction. Employees must not approve their own requests or see another team's restricted records. If the external system is unavailable, staff need to understand what has happened and how the request will be recovered.

The buyer asks two fictional suppliers, A and B, to explain that same journey, including its failure cases. This is a discussion brief, not a demand for an unpaid prototype or access to another customer's production system. The buyer can inspect an appropriately shareable example, discuss a proposed design or commission a bounded assessment where the uncertainty warrants it.

Choose a workflow that exposes your project's important dependency. For a rescue project that may be recovering a failed release; for a migration it may be reconciling one difficult record. A visually impressive but unrelated demonstration gives less evidence about those decisions than a modest example that makes the actual constraints visible.

Comparison 1: what happens to a normal request?

Supplier A's fictional proposal says the portal includes approvals and an integration. Supplier B describes the request states, identifies where the manager's authority is checked and separates approval from confirmation that the external system received the instruction. B can show a demonstration using a simulated external system, but has not tested the buyer's actual connection.

Record A's answer as a claim awaiting explanation. Record B's answer as a proposed approach with an inspected example, not a verified integration. B has made the boundary easier to discuss; that does not establish compatibility with the real system. Ask both suppliers what access they need, which external behaviour they assume and what result would count as acceptance.

Comparison 2: what happens when a request gets stuck?

Suppose the external system receives an approved request, but the portal loses the connection before receiving confirmation. Retrying could create a duplicate; doing nothing could leave the request unresolved. Supplier A says failed requests will retry automatically. Supplier B proposes checking the external reference before retrying, while explicitly noting that the external system's lookup and duplicate-handling behaviour is unknown.

Neither answer is enough to approve this dependency for a full build. A has not explained the ambiguous outcome; B has identified information it still needs. The buyer records a follow-up: obtain the external documentation and test access, identify who owns that investigation and agree how an unresolved request will reach an operator. A specific unknown is more useful than a confident promise that hides it.

Comparison 3: who can approve or see sensitive records?

Supplier A shows an employee screen without an Approve button. Supplier B explains how the application would reject an approval attempt from that employee even if the request were sent without using the screen. B offers an authorised, sanitised example of a permission test from a different system; the buyer has not inspected a test for its own organisation's role model.

A hidden button demonstrates an interface choice, not the complete permission boundary. B's example explains a testing approach, not proof that the buyer's rules are implemented. Write the acceptance question plainly: can an employee approve their own request or read another team's record? Identify who will confirm the roles and who will verify the implemented behaviour before release.

NCSC supplier-assurance guidance supports tailoring evidence requests to the risk and reviewing assurance as circumstances change. It does not make a generic questionnaire or a certificate proof that this workflow is safe. Keep the request proportionate and respect confidentiality; another client's credentials, records and private test results are not appropriate buying evidence.

NCSC: Supplier assurance — having confidence in your suppliers

Comparison 4: a deployment breaks the request journey

Supplier A promises ongoing support. Supplier B names who would monitor the journey, who can restore a previous release and how unresolved requests would be checked after recovery. B's proposal assumes the buyer provides an out-of-hours contact; A has not described a support window. These differences are responsibilities to resolve, not a basis for inventing a reliability score.

Ask both suppliers to describe one failed-release scenario and the recovery evidence they would provide. Clarify the support period, escalation route, account access and any responsibility retained by your team. Compare the included operating work alongside development and launch. A lower proposal may leave your organisation with work it is able to do; the decision should make that transfer explicit.

Comparison 5: another team needs to take over

Supplier A offers a copy of the source code at completion. Supplier B proposes a buyer-controlled repository, documented build and release steps, an account-access inventory and a handover exercise. B excludes transfer of the external supplier's subscription because the buyer must arrange it. That exclusion belongs in the comparison before a contract is agreed.

Source access alone does not show that a successor can operate the application. Ask what a new maintainer would need to build, test, deploy and diagnose a request, and when those materials will be kept current. Clarify control and permitted use of code, accounts and third-party components with the people responsible for the agreement. Do not infer those arrangements from a sales presentation.

GOV.UK's technology guidance encourages adaptable choices and considering existing systems, ownership costs and lock-in. It also encourages testing assumptions about technologies, security and interfaces with prototypes. These are useful questions for this comparison; the government guidance is not a mandatory procurement standard for a private UK business.

GOV.UK Service Manual: Choosing technology — an introduction

Use case studies to ask better follow-up questions

A relevant case study helps you identify what to investigate, but its published scope is not proof of every proposed capability. Veda Software's Powerleague print-portal case study describes venue accounts, central visibility and a supplier ordering flow, alongside workflow mapping, design, build and rollout. A buyer evaluating a portal could ask which responsibilities are comparable and how the proposed approval or integration differs.

The published page does not establish the permission tests, recovery arrangements or contract terms for our fictional service-request portal. Keep those questions open. For any supplier's work, distinguish its actual contribution from a familiar logo or an adjacent project. Where references can be shared with permission, ask about the specific responsibilities and working decisions you need to understand.

Veda Software: Powerleague print portal

Choose the next commitment the evidence can support

In this fictional comparison, B has explained more boundaries, but the real integration remains unresolved for both suppliers. The buyer should not average attractive interface work against that critical unknown. It can ask A to clarify its approach, obtain the missing external information and then decide whether a bounded integration assessment is justified. This example does not select a winner on the buyer's behalf.

For each important question, record the answer, the evidence you inspected, its limits, the unresolved point, an owner and a next decision. Keep claim only, example inspected, proposed approach and verified for this project distinct. Unknown means information is missing; failed means an agreed check was attempted and did not meet its requirement. Neither should silently become a pass.

A discovery phase can be useful when it resolves a consequential uncertainty with defined outputs and an exit decision. It need not be a default extra phase where the scope and dependencies are already understood. Match the commitment to what remains uncertain, compare assumptions and exclusions on the same basis, and carry the agreed responsibilities into delivery reviews and acceptance. Revisit the selection assumptions when the workflow or supplier arrangements change.

A shortlist record that leads to a decision

For the fictional service-request comparison above, create one row per essential question: independent approval, external receipt confirmation, recovery ownership and successor handover. In each row record supplier A’s answer, supplier B’s answer, evidence inspected, unresolved dependency, responsible person and the next decision. This is an evidence record, not a points contest.

An inspected demonstration may justify a focused assessment but cannot settle an untested connection. For example, assign the client’s operations owner to obtain API test access, and ask each shortlisted supplier to describe what a bounded test would produce. Agree in advance that duplicate creation or an unrecoverable timeout prevents committing to that integration design. If access remains unavailable, compare a manual first-release handoff with the cost of waiting.

Separate selection from later governance. Before appointment, check whether the supplier can explain and evidence the proposed responsibility. After appointment, the governance guide helps set decision rights, release evidence and escalation. Keeping those two decisions separate makes this shortlist useful even when you already have an internal technical team.

NCSC: Supplier assurance — having confidence in your suppliers

How this relates to Veda Software’s work

Use the Powerleague ordering narrative already discussed as a prompt for relevant questions, rather than a substitute for evidence from your own proposed workflow.

Veda Software: Powerleague print portal

Key takeaways

  • Compare suppliers against the same representative workflow and failure cases.
  • Separate a claim, a demonstration and evidence verified for your own project.
  • Give critical unknowns an owner and a decision before committing around them.
  • Compare operating and handover responsibilities alongside the build scope.

Business Improvement Partnership

Make the next improvement part of how your business works.

Work with us on the systems, software and automation that can save your team time. Start with your priorities and agree a practical plan for ongoing improvement.

Veda Software · Self-serve ordering portal build

One portal, every venue: decentralising print ordering across a national estate

Print ordering across a national estate ran through an awkward external experience and one central team. Veda Software designed and built a self-serve ordering portal: every venue orders directly, and the central team no longer has to push each order through by hand.