NG Labs / Web application development

Web applications
for work people
return to.

Scope and develop a focused browser-based product around a real business task, its information and the people responsible for it.

Choose the task before the technology.

A web application gives people a place to do recurring work in a browser. That might be reviewing requests, updating a record, coordinating an approval or using information from more than one system. The scope starts with the action and the person responsible for it.

A marketing website and an operational product have different jobs. If the main need is to explain a service and receive enquiries, a website may be the appropriate shape. If someone returns to complete a task, review information or make a decision, an application may be worth exploring.

A useful first question: “Which task should a particular person be able to finish, and how will we know it is finished?”

A scope that names the boundaries.

Before agreeing an implementation, identify the current workflow and the information it uses. Map the primary journey, permissions and important exceptions. Keep assumptions visible so that unknown access, data quality or third-party behaviour does not become an invisible dependency.

  • Users and roles: who starts, reviews, changes or approves the work.
  • Records and states: the item being managed and the changes it may go through.
  • Existing systems: the real source of information, access owner and available interfaces.
  • First-release boundary: what is included, deferred or explicitly manual.
  • Acceptance: observable journeys, agreed devices and failure cases.

A prototype can test a difficult interaction or an uncertain process. Label its assumptions and what it does not connect to. A useful prototype helps make a decision; it does not establish production readiness by itself.

What belongs in the delivery plan.

The core experience

Agreed screens, actions and states for the primary user journey, including loading, empty, error and recovery behaviour. Keyboard operation and usable labels belong in those journeys.

Data and access

A record model, permissions and necessary data flows. Agree how accounts are owned, how access changes and what should happen when a request is not permitted.

Connections with a clear contract

For an agreed integration, identify the source record, destination, update direction, trigger and responsible owner. Define duplicate handling, failure visibility and a way to review the result.

Release and handover

Document the repository, hosting and service accounts, configuration responsibilities, test evidence and operating notes. Set the ongoing maintenance and support scope separately instead of leaving it implicit.

Review a working journey, including the awkward parts.

Acceptance should describe what a user can demonstrate. A successful path is only part of that review. An application also needs an agreed response when information is missing, an action is repeated, access is removed or another system cannot be reached.

A successful path

A permitted user opens the relevant record, completes the action and sees the resulting state on an agreed device.

A useful failure path

An unsuccessful update stays visibly unresolved, avoids an unintended duplicate and gives an agreed person enough context to follow up.

The exact tests depend on the product, data and operating environment. Security, accessibility and recovery requirements need a scoped review; a demonstration screen is not evidence of complete coverage.

Keep the product usable after handover.

The delivery conversation should cover who owns the domain, repository, hosting and external accounts, how the data can be exported, and who is responsible for routine maintenance and incidents. Identify access and documentation needed by the next person maintaining the product.

If customers need their own shared view, read about client portal development. If you are still defining roles and decisions, the client portal brief template is a practical starting point.

Questions before a project conversation.

Do we need a custom web application?

Start with the workflow. If an existing product handles it well, configuration may be enough. Custom development becomes a discussion when the roles, decisions or connections have a material requirement that an available tool cannot reasonably cover.

Can you connect our existing software?

We can scope an integration after reviewing the actual system, documentation, permitted access and test environment. The proposal should identify which records move, who owns them and how failures are handled. No specific third-party connection is included by assumption.

What information is useful for an initial enquiry?

Describe one recurring task, its users, existing tools and the result you need. Include known constraints and who approves the work. Leave passwords, credentials and private customer records out of the first message.

Updated 17 November 2026 / Buyer acceptance worksheet

Before handover, ask who can operate the next release.

A finished screen is only one part of accepting a web application. The company also needs a workable path to maintain it, recover agreed data and assign incidents. Agree these delivery items before the final review, rather than assuming that a repository link includes every operating responsibility.

Consider an illustrative supplier approval app. A new operator must release an approved change, identify an unresolved approval and export the relevant records. This is a planning example, not a claim about a delivered client system. The review should show what the receiving team can actually do, with gaps kept visible.

Five handover questions and the evidence to request
Delivery areaBuyer questionReviewable evidence
Company-controlled accessWho can grant and remove repository, hosting, domain and agreed service access?A company-named owner and a role/access record. Use the platform’s normal invitation flow; do not share a personal password.
Release instructionsCan the receiving team identify the current release and follow the agreed preview-to-release process?Versioned source, required tool/configuration names, preview steps, approval owner and the documented recovery path if release fails.
Data export and recoveryWhich records and files are covered, and what is excluded?A scoped export sample and dated restoration exercise in an isolated environment, with observations and unresolved limitations.
Failed work and incidentsWho notices and investigates an unresolved action?A safe example of the visible failure, the escalation owner and the evidence needed to reconcile it.
Maintenance and exitWho handles updates, costs and a future supplier transition?A separate scope for routine work, response arrangements, third-party dependencies and the handover/exit materials.

Rehearse one recovery, without risking live customer work

For the illustrative approval app, agree an isolated review environment and permitted sample data. Name one representative record and its related attachment. Ask the delivery team to explain which backup contains them, restore the agreed sample, and demonstrate what the receiving operator can inspect afterwards. Record the tested version, date, covered data and actual result.

A backup file existing somewhere is not restoration evidence. A successful sample restore is also not proof that every record, file, region or future incident is covered. Document missing dependencies, retention decisions and exclusions. Agree acceptable recovery objectives with the business and supplier; do not invent a universal deadline or treat this worksheet as a backup guarantee.

Keep credentials and responsibilities separate

The receiving company should know where configuration is managed and who can administer it. That does not require putting secret values in a document or public repository. Record configuration names, access owners and the approved method of transfer. Resolve access changes through the relevant platform, with the company’s authorised people involved.

For an external connection, distinguish the application owner from the other system’s owner. If a request times out, the operator should know who checks whether it completed before repeating it. The integration recovery checklist helps specify that decision; it does not establish that a particular integration is already implemented.

Copy one acceptance row

Delivery item → company owner → evidence and tested version → observed result → scope/exclusions → unresolved action and owner.

Download the blank web application handover review (.txt). Use it for the five delivery areas and the sample recovery exercise. Keep credentials and private customer records out of the worksheet.

This review is a procurement aid, not a security certification, legal opinion or replacement for your project agreement. A prototype may help review a journey, but it should be labelled with its missing production connections. If a handover item is not included in the quote, agree the extra work and owner before treating it as accepted.

hello@northgrowthlab.comРазработано компанией NG LABS