Customer portal development: what to include

Choose the task customers actually repeat

Review recurring enquiries: requests for a status update, a copy of a document or a change to an order. Identify which can be answered reliably without staff interpretation. A portal does not help if customers still need to call because its information is incomplete.

Interview a small sample of customers rather than assuming they want another account. For infrequent, low-risk interactions, a simpler secure handoff may be more appropriate than a full portal.

Design the first visit

Explain what the person can do, show the current status and make the next action obvious. Use language customers already recognise from your service. Internal stage names are rarely meaningful outside the business.

Account invitation, password recovery and support must work as carefully as the main screen. A locked-out customer has not received self-service.

Decide permissions before layouts

Test information boundaries

Changing a link or guessing a document address must not reveal another customer's information. Access checks need to apply to downloads and background requests, not only the visible navigation. Ask the developer to demonstrate these negative tests.

Use realistic file sizes and names during testing. Decide which file types are accepted, how uploads are handled and how old documents are replaced or retained.

Keep notifications useful

An email should tell the customer that something needs attention and link them to the right place. Avoid putting sensitive details into a notification unnecessarily. Make sure failed notifications are visible to staff and that a resend does not create duplicate actions.

Run a customer pilot

Watch customers complete the target task without coaching. Record where they hesitate and what they still ask staff. Compare the specific type of enquiry before and after the pilot; do not promise a percentage reduction without measured evidence.

A useful first portal is often narrow. Add more features only when they solve observed customer needs and the underlying operational data is reliable.