Northwind Grid Services - Offline Inspection App Pattern
Northwind Grid Services maintains 410 substations. Inspectors work up to 10 hours a day in areas with no cellular coverage. Each inspection captures roughly 60 fields and 4–6 photos, and the utility completes about 1,800 inspections per day. The business sponsor has asked for a model-driven app "because it ships faster," while operations insists the app must work with no network and must never lose an inspection. Inspectors sometimes swap tablets mid-shift, and two inspectors occasionally edit the same work order. Six years of history (about 2.9 million inspection rows) must be reportable by regional managers, and all data must remain in the UK region. The organization owns 500 Power Apps per-user licenses.
1. Recommend an app pattern and defend it against the sponsor's stated preference.
Read the worked answer
The deciding factors are offline correctness and the reporting shape, not build speed - so answer the sponsor's concern about speed but don't let it drive the choice.
Recommendation: a model-driven app with a mobile offline profile, scoped to a filtered subset - the inspector's own work orders for a window such as −7 days to +14 days. Do not sync history; 2.9M rows is a reporting problem, not an app problem.
Reasoning:
- Model-driven apps give you security roles, views, forms and audit for free, and offline sync is a supported, configurable profile rather than custom code.
- Canvas is justified only by a bespoke UX requirement. If chosen, you must prove in a POC that photo capture and attachment upload survive offline use, and that all gallery queries delegate - otherwise the app silently truncates data.
- Data residency is satisfied by the environment's region; confirm the tenant's default region before the environment is created, because it cannot be moved later.
- Device swapping forces a fresh sign-in and re-sync; make that a tested scenario rather than an assumption.
Escalate the requirement "never lose an inspection" to a measurable SLA (e.g., a submission is durable once the user sees a queued state) rather than a promise.
2. Define what you would validate before committing, and the condition that would reverse your recommendation.
Read the worked answer
Run a 3-week POC with 20 inspectors and explicit exit criteria, because the gating risks are behavioural and volume-related, not feature-related.
Test script and measures:
- One 10-hour shift, 40 inspections, 200 photos, airplane mode throughout. Measure sync duration after reconnect; target under 3 minutes on the corporate tablet.
- Two devices editing the same work order offline; confirm you can detect and resolve the conflict rather than silently overwriting.
- Kill the app mid-sync; verify no duplicate inspection and no data loss.
- Device swap mid-shift; verify recovery path and re-sync.
- Manager reporting on 6 years of data - confirm it runs outside the app.
Reverse the recommendation if: photos or attachments cannot round-trip reliably under offline conditions, or sync exceeds the agreed SLA. In that case the fallback is a thinner model-driven offline app plus a separate attestation step, or a licensed field-service mobile app if the organization already owns that capacity.
Also confirm licensing during envisioning: 500 per-user licenses cover inspectors and managers, but not warehouse/support staff, and not external or shared-device scenarios.