Choosing a development partner is not about finding the most persuasive presentation. The decision affects the product, data, internal capability, and the cost of changing direction. Compare evidence against a specific problem, not generic promises or a list of technologies.
A sound selection process makes the desired outcome, measurement, decision ownership, and eventual handover explicit. This guide offers a practical scorecard; it does not replace legal, financial, data protection, or security review for the specific engagement.
1. Define what you are buying before comparing firms
Adding capacity, contracting for an outcome, and exploring uncertainty are different purchases. If the brief mixes them, proposals will not be comparable.
- Capacity: you need specific roles working alongside your team under shared direction.
- Outcome: the supplier takes responsibility for delivering agreed scope and acceptance criteria.
- Discovery: you need to understand users, process, data, risks, and options before committing to a full build.
The UK government's buyer guide recommends describing the situation or problem, technical requirements, user needs, budget, and selection criteria. It also separates essential from desirable requirements and calls for the same evaluation method and weighting to be applied to every candidate.
Source: GOV.UK — Digital Outcomes and Specialists buyers' guide
2. Build the scorecard before proposals arrive
Choose criteria and weights before seeing offers. This reduces the temptation to reshape the evaluation around a preferred supplier. There is no universal weighting: it should reflect impact, uncertainty, and the capabilities you already have.
- Problem understanding: reframes the objective, separates assumptions from facts, and identifies the first questions to answer.
- Actual team: introduces the people who would participate, their availability, responsibilities, substitutions, and relevant experience—not only sample profiles.
- Delivery approach: proposes verifiable stages, frequent demonstrations, a definition of done, risk management, and a route to stop or change direction.
- Technical quality: explains architecture, testing, code review, observability, accessibility, data, deployment, maintenance, and reversible decisions in proportion to the project.
- Security: integrates requirements and secure practices into the development lifecycle, manages dependencies and vulnerabilities, and can provide evidence appropriate to risk.
- Collaboration and transparency: makes progress visible, handles disagreement with evidence, records decisions, and shares knowledge with your team.
- Continuity and exit: clarifies repositories, documentation, environments, data, licences, intellectual property, support, handover, and service recovery. Specific terms require professional review.
- Whole-life cost and value: exposes assumptions, included work, dependencies, recurring costs, and conditions that would change the estimate; initial price is not the only cost.
3. Ask for evidence resembling your engagement
A case study is useful only when it reveals the context, intervention, constraints, and what the proposed team actually did. Ask for a reference, work history, or comparable case, and validate what can be checked without requesting another client's confidential information.
The GOV.UK Service Manual recommends evidence of using agile methods to design, build, and deliver, plus a commitment to pass expertise to permanent staff. Make transfer observable through joint work, operational documentation, decision reviews, and the ability to deploy and maintain without hidden dependency.
Source: GOV.UK Service Manual — Working with contractors or third parties
4. Evaluate how the team works, not only what it says
A conversation with the people who would deliver the work reveals more than a polished proposal. Give every candidate the same scenario: a critical dependency changes, a user hypothesis fails, or a data constraint appears. Observe their questions, what they make visible, and how they balance scope, quality, and risk.
- What would you investigate in the first two weeks, and which decision would each result support?
- How will you demonstrate working software, and how will we know a stage is complete?
- Which risks could stop the project, and when would you test them?
- Who reviews code, architecture, accessibility, privacy, and security?
- How do you handle incidents, vulnerabilities, and dependencies after launch?
- What must the internal team learn to operate and evolve the solution?
Where uncertainty is high, a paid, bounded discovery or technical exercise may produce better evidence than unpaid speculative work. Define deliverables, ownership, confidentiality, and exit criteria in advance.
5. Treat security as a demonstrable practice
NIST presents the SSDF as a common vocabulary that purchasers and suppliers can use to discuss secure practices throughout the development lifecycle. It does not turn certification into a guarantee or require every practice to be applied identically; it helps request processes and evidence proportionate to risk.
Source: NIST — Secure Software Development Framework 1.1
CISA's Secure by Demand guide advises assessing product security as well as the supplier's enterprise security, with questions before procurement, requirements in the agreement, and ongoing assessment. Topics include authentication, logging, updates, vulnerability handling, and transparency.
Source: CISA — Secure by Demand Guide
6. Watch for warning signs
- A fixed solution appears before the team understands users, process, data, and constraints.
- The sale depends on one person, but the delivery team does not participate in the evaluation.
- A final date and price are promised despite fundamental uncertainties and no plan to reduce them.
- Activity is presented as progress: hours, tickets, or sprints without verifiable outcomes.
- Testing, operations, security, debt, incidents, or knowledge transfer are avoided.
- Access to code, data, environments, or documentation depends exclusively on the supplier.
- The proposal competes on rate alone and does not explain assumptions, scope, recurring costs, or trade-offs.
7. Make a decision you can explain
Score every criterion on the same scale, record the evidence, and separate facts, inferences, and open questions. Document why weights were chosen, which risks are accepted, and which conditions must be resolved before work starts. If no candidate meets the essential criteria, revisiting the brief is a valid outcome.
Selection ends with a collaboration hypothesis, not certainty. The agreement, governance, and first stage should test early whether the team understands the problem, delivers quality, communicates transparently, and increases—rather than reduces—your organisation's capability.
Custom software vs SaaS: how to decide
Codiloop can help you scope the challenge, define a verifiable first stage, and transfer knowledge from the start.

