A customer asks for an update. Someone checks a spreadsheet, finds the latest file in a shared drive and replies by email. That is frustrating, but it does not yet tell you what software to buy.
Describe one recurring job before naming a product: “A customer submits a request, our coordinator checks it, a reviewer approves the current version, and the customer can see the next action.” Then ask whether the tools you already pay for can demonstrate that complete journey.
A configured service deserves a fair test. Custom development deserves a clear reason. The worksheet below helps collect evidence for either choice; it does not recommend a particular vendor or claim that an integration is available.
Compare the work, access and ownership
| Decision | Evidence to collect | What it changes |
|---|---|---|
| Recurring task | One real start-to-finish journey, including an exception | Establishes what must actually work, beyond a dashboard demonstration |
| Current tools | Demonstration using features and access already available | Shows whether configuration may be enough before buying new software |
| Customer access | Who may view, edit, download or approve each record | Exposes rules to check before comparing proposals |
| System boundary | Which system owns the record and which receives a copy | Prevents conflicting versions from becoming an unowned problem |
| Manual work | Frequency, handling time, corrections and follow-up | Gives a baseline for evaluating change, without assuming savings |
| Ownership | Who controls accounts, export, maintenance and support decisions | Makes the responsibilities after launch visible |
Mark each requirement as demonstrated, unsupported or still unknown. A feature name in a sales page is not the same as a demonstration of your task. Do not score an unknown as a failure or quietly treat it as included.
A worked example: artwork approval
Imagine a small supplier collecting artwork approvals. Customers need to see their own current artwork, comment and approve a version. This is an illustrative buying example, not an NG Labs client result.
If an existing service can demonstrate those exact journeys, including what happens when a copied file link reaches the wrong customer, a configured service may deserve a pilot before custom development.
Now change the requirement: an approval must identify a particular artwork version and order record, while two internal roles have different decision rights. Ask the existing supplier to demonstrate that behaviour. Document the unsupported steps and any proposed manual workaround. The extra complexity deserves a closer workflow review; it does not automatically make a new application the best answer.
Keep the allowed and refused access journeys alongside the workflow. A tidy customer screen does not establish that account boundaries work.
Measure today before estimating the benefit
For a clearly labelled planning example, 30 requests a week taking eight minutes each represent four hours of handling. That is a baseline input, not four hours of savings you can promise. Some work may remain, and setup, exceptions and maintenance take time too.
Record how the timing was collected and which tasks it includes. Separate necessary review work from avoidable copying or repeated questions. If the real issue is unclear responsibility, buying software may simply move that confusion into another interface.
Compare the whole agreed operating scope: configuration or development work, account charges where applicable, staff time, data migration, support, export and future change. Use actual supplier terms and your own observations; a generic article cannot establish those costs.
Run the smallest useful pilot
Bound the pilot to one type of request, one customer group, one review route and one awkward exception. Agree its acceptance evidence before the demonstration.
- The intended customer completes the task and can identify the next action.
- A different customer is refused access to the same record or file.
- The reviewer can identify which version was approved and by whom.
- An incomplete update remains visible with a named person responsible for checking it.
- The business can explain how its information would be exported or handed over.
Record what was observed, what remains unknown and who makes the buying decision. A limited pilot does not certify complete security, accessibility or recovery coverage. Identify any specialist review the actual data or workflow requires separately.
Choose from the evidence
Keep the existing approach when the problem is mainly inconsistent use of an adequate tool or unclear ownership. Consider custom development when an important, repeated workflow remains unsupported after a fair test, and the business is prepared to own the resulting application.
A deliberate manual step can be reasonable in an early release if its owner, visible status and review condition are explicit. Where systems exchange records, agree failure and recovery decisions before treating the connection as part of the solution.
Bring the completed worksheet to a client portal development conversation. A short evidence-led workflow brief is more useful than a feature wish list.
Download the pilot decision worksheet
Download the blank text worksheet. It contains no live form, account connection or data submission. Keep customer details and credentials out of any public brief.
TASK AND USEFUL COMPLETION: CURRENT TOOL / CONFIGURATION TO EVALUATE: CUSTOMER AND INTERNAL ROLES: APPROVED RECORD / FILE VERSION: ONE AWKWARD EXCEPTION: REQUIREMENT | OBSERVED EVIDENCE | DEMONSTRATED / UNSUPPORTED / UNKNOWN CURRENT HANDLING BASELINE AND HOW IT WAS MEASURED: MANUAL WORK THAT REMAINS AND ITS OWNER: PILOT BOUNDARY / EXCLUSIONS: ACCEPTANCE REVIEWER AND EVIDENCE: ACCOUNTS, EXPORT AND ONGOING SUPPORT OWNER: BUYING DECISION, REASON AND NEXT ACTION: