When choosing business software developers in Devon for a customer portal, prioritise their approach to customer permissions, existing systems and ongoing support over the appearance of a demonstration. Ask each shortlisted team to explain one complete customer journey, including what happens when access is refused or an integration fails. Their answers will give you a stronger basis for choosing than a feature list or headline build price.

If workshops at your premises matter, make that a selection requirement. Otherwise, compare Devon-based and remote development partners against the same delivery criteria. Proximity can help with collaboration, but it does not establish whether a team can protect customer data or maintain the application.

Give developers a workflow they can challenge

Start with a task customers currently handle through emails, calls or spreadsheets. Describe who initiates it, what information they need, what your staff do next and where the outcome is recorded. A portal brief should explain the transaction, not simply request a dashboard, document area and messaging system.

For example, a hypothetical service business might want customers to approve quotations online. The useful scope includes which version they approve, who has authority to approve it, whether approval can be withdrawn and how staff receive the decision. Those details determine the build more clearly than saying customers need an approval button.

  • Name the first customer group and the one or two tasks the initial release must support.
  • List the records involved, their current location and the staff responsible for keeping them accurate.
  • Define success in observable terms, such as completed approvals or fewer requests for progress updates.

If you are still deciding which task deserves a portal, these customer portal examples can help frame the options. Once the purpose is settled, ask developers whether an existing platform could meet it before commissioning a bespoke application.

Test the permission model, not just the login

Authentication establishes who a user is. Authorisation determines what that user may see or change. A working login demonstrates only the first part. For a portal serving several customer organisations, the important question is how the application prevents one organisation from accessing another's records.

Ask for a permission matrix: a simple record of user roles and permitted actions. It should distinguish ordinary customer users, customer administrators and your internal staff. If customers have branches or external advisers, include those relationships rather than assuming everyone within an account should see everything.

Then ask how the rules will be tested. Hiding a link in the interface is not access control. The application must check permissions when it receives a request, including requests for documents, exports and changes to records. Ask for tests showing that a user cannot retrieve another customer's document by changing an identifier in a request.

Account recovery deserves equal attention. Establish who may invite users, change account administrators and remove access when someone leaves. Discuss multi-factor authentication, which adds another verification step, particularly for privileged accounts. The recovery process should not provide an easier route around those controls.

Ask what security testing is included, who performs it and how findings are resolved before release. For a portal handling sensitive information or consequential transactions, consider a separately scoped independent assessment. A scan or test report is useful evidence, not a guarantee that the application has no vulnerabilities.

Make integrations explainable before accepting a quote

A portal may display information from your customer relationship management system, accounts package or operational database. Establish which system is authoritative for each important record. Without that decision, two systems can hold conflicting versions of an address, order or account status.

Ask developers to map the direction and timing of each data flow. Does the portal read information only, or can customers change it? Must changes appear immediately, or is a scheduled update acceptable? What should customers see when the source system is unavailable or the displayed information is out of date?

Use a failure scenario in the supplier discussion. Suppose a customer submits a request, the receiving system accepts it, but the confirmation never reaches the portal. Retrying without safeguards can create a duplicate. Ask how the developer will identify the original request, confirm its status and retry safely.

Before accepting an integration estimate, ask what has actually been inspected: API documentation, available access permissions, sample records and a test environment. An API is an interface through which software exchanges data. Its existence alone does not establish that it supports every action your portal needs. Unchecked assumptions should be named in the proposal.

Ask for delivery evidence you can assess

Request examples involving comparable complexity, not just similar-looking screens. A relevant discussion should cover permission boundaries, connected systems, testing and post-launch changes. Where client confidentiality prevents a detailed demonstration, ask for an anonymised technical explanation or a walkthrough using demonstration data.

For your own project, agree milestones around working customer journeys. A milestone such as 'customer submits a request and staff process it in the existing system' is easier to assess than 'backend mostly complete'. Include error states and permission checks in the demonstration, not only the successful path.

Write acceptance criteria before development starts. For quotation approval, these might require the portal to show the correct document version, reject an unauthorised user and record who approved it and when. Agree who signs off each milestone and how defects differ from requests for additional functionality.

Usability also belongs in acceptance. Ask how the team will test with representative customers, including people using phones or assistive technology. Staff who understand the process may navigate a prototype successfully while customers struggle with its terminology, instructions or recovery messages.

Compare proposals on the same scope

Give shortlisted developers the same brief and ask them to separate discovery, design, development, integration, testing, launch and ongoing operation. Do not assume that a lower total includes the same work. Check exclusions, dependencies and the decisions your team must make before delivery can proceed.

Where a critical integration remains uncertain, commission a bounded discovery stage or technical investigation first. Agree its outputs, such as a tested connection, permission model and delivery scope. Ask whether you can use those outputs with another supplier if you decide not to continue.

FullyCoded builds customer-facing portals and API integrations. The studio is based in Cornwall and works with organisations across the UK. Its application development service is an option to assess if your shortlist extends beyond Devon, using the same requirements and evidence checks you apply to other suppliers.

Agree who runs the portal after launch

Ask who receives alerts for failed customer actions, integration errors and unusual access activity. A server availability check cannot establish that customers can complete a task. Agree which journeys need monitoring, who investigates failures and who communicates with customers while a problem is being resolved.

Separate support response targets from resolution commitments. Acknowledging an incident is different from restoring service. Establish coverage hours, escalation contacts and what happens outside the support agreement. Also agree responsibility for software updates, backups, restoration tests and changes to connected systems.

Include a handover and exit plan. Clarify ownership and access for source code, hosting accounts, domains, documentation and deployment instructions. Record third-party licences and how customer data can be exported. Ask what another competent team would need to maintain the application without relying on undocumented knowledge.

Use one scenario to make the final decision

Give each finalist the same customer journey and ask them to walk through permissions, data movement, failure handling and support. Note what they can demonstrate, what remains an assumption and what they propose to investigate. Name an internal owner who can resolve business questions throughout the project.

Before authorising the build, make sure that journey has agreed acceptance criteria, a priced scope and a named owner for failures after launch. That gives you something concrete to buy, test and operate, rather than simply a portal to unveil.