Keep the operational records separate
A network alarm is evidence from infrastructure. A service problem describes affected service behavior. An incident coordinates restoration. A trouble ticket communicates a reported issue across customer or partner boundaries. A work order authorizes field activity. An appointment reserves a time with a customer or site. Linking these records preserves their different owners, states, clocks, and audit obligations.
Service and customer impactProduct, circuit, SLA, customer sites, affected users, criticality, redundancy, and communication commitments.
Logical networkService path, VLAN, IP, optical channel, topology, dependencies, alarms, tests, and performance history.
Physical resourcesCabinet, rack, chassis, card, port, splitter, fiber segment, tower, antenna, power, battery, serial, and spare.
Field contextSite, coordinates, access hours, keys, hazards, permits, landlord, contact, photos, route, and offline map.
Build the service-to-resource graph first
Dispatch quality depends on knowing which physical resources support the affected service. TMF639 provides a standardized resource-inventory interface and TMF674 models sites. Inventory should contain stable IDs, resource specifications, lifecycle state, parent/child and connectivity relationships, installed location, serials, ownership, warranty, maintenance, and provenance. Discovery data is not automatically inventory truth; reconcile and approve changes.
Fault-to-work-order flow01 DetectCollect fault evidenceAlarm, performance anomaly, customer report, partner notice, power event, test, and weather context.
02 CorrelateFind common causeTopology, time, shared resource, site, dependency, duplicate reports, maintenance, and suppression.
03 AssessCalculate service impactCustomers, circuits, redundancy, critical sites, SLA, safety, revenue, and estimated scope.
04 QualifyDecide whether field work is neededRemote test, reset policy, diagnosis confidence, asset, location, access, skills, parts, and permit.
05 DispatchAssign a feasible crewPriority, skill, certification, shift, route, time window, vehicle, tool, stock, and workload.
06 ExecuteGuide offline field workSafety check, asset scan, test steps, measurements, photos, parts, notes, signatures, and timestamps.
07 VerifyProve restorationNetwork test, telemetry recovery, service KPI, customer confirmation, soak period, and residual risk.
08 LearnClose every systemInventory delta, cause, resolution code, parts, labor, SLA outcome, repeat risk, and knowledge update.
Correlate before multiplying tickets
A fiber cut can produce thousands of alarms and customer contacts. Group by shared resource, topology, timing, geography, and maintenance state. Keep each customer/partner ticket linked to one service problem or incident while preserving individual communication and SLA. TMF656 covers service problems, TMF724 incident lifecycle, TMF621 trouble tickets, and MEF 113/124 support inter-provider trouble-ticket flows.
Qualify the work before dispatch
A work order should carry diagnosis hypothesis, affected asset and site, required skill/certification, safety class, access constraints, customer appointment, test procedure, estimated duration, parts/tools, rollback, acceptance criteria, and evidence checklist. If asset, location, access, or task is unknown, route to a qualification queue rather than sending a technician with a vague description.
Routing is constrained scheduling
Distance is only one variable. Model technician shift, skills, certifications, territory, union or contractor constraints, vehicle, stock, tool calibration, site access, customer time window, task precedence, expected duration, traffic, weather, break rules, overtime, SLA risk, and emergency preemption. OR-Tools supports vehicle routing with time windows, capacity, pickups/deliveries, and optional visits, but the objective function must reflect operational priorities.
| Dispatch signal | Required data | Optimization effect | Fallback | Unsafe shortcut |
|---|
| SLA at risk | Contract clock, severity, pause rules, target and projected restore. | Raise priority cost as breach approaches. | Duty manager override with reason. | Use ticket age as the only clock. |
| Specialized optical fault | OTDR skill, fiber certification, test equipment and splice kit. | Restrict feasible crews and reserve tools. | Pair local technician with remote expert. | Assign nearest generalist. |
| Customer appointment | Confirmed window, contact, access, duration and travel. | Hard or penalized time-window constraint. | Offer new slots through appointment workflow. | Promise a slot before checking capacity. |
| Remote tower site | Route, weather, daylight, access vehicle, permit, safety and battery. | Bundle compatible work and add risk buffer. | Delay safely or mobilize specialized crew. | Optimize only road distance. |
| Required replacement part | Part compatibility, van stock, warehouse, reservation and return core. | Add pickup, capacity and precedence constraints. | Transfer stock or split diagnosis/repair visits. | Assume inventory records are exact. |
| Major outage | Common cause, affected services, crew status and restoration plan. | Freeze low-priority work and coordinate teams. | Incident command with manual dispatch. | Let each ticket dispatch independently. |
SLA is a clock model, not a color
Store acknowledgement, response, arrival, restore, resolution, and communication targets separately. Define start, pause, resume, stop, service hours, exclusions, severity changes, customer delay, force majeure, and partner dependency. Maintain both contractual and operational clocks. Forecast breach probability from diagnosis, travel, access, parts, and work duration instead of waiting for a dashboard to turn red.
Detect
Ack
Triage
Assign
Travel
Access
Repair
Verify
Close
The mobile app is offline-first by contract
Download a bounded job packWork order, asset/site snapshot, map, contacts, hazards, procedure, forms, permitted history, parts and version tokens.
Record an append-only field journalState transitions, scans, measurements, GPS, photos, parts, notes, signatures and local timestamps get stable client IDs.
Synchronize idempotentlyUpload media in resumable chunks, send events with idempotency keys, resolve conflicts by field ownership, and retain acknowledgements.
ODK Collect demonstrates mature offline form concepts: locally available forms, finalized submissions, offline maps, media, datasets, and later synchronization. A telecom app needs the same reliability plus work-state control, encrypted local storage, remote wipe, device attestation, schema versions, and explicit conflict handling.
Do not use last-write-wins for work
The server owns assignment, SLA and cancellation; the technician owns captured measurements and media; inventory changes require validation; customer contacts may update appointment details. Synchronize commands and observations as events. If a job is cancelled while offline, preserve completed evidence and route it to review rather than deleting it or reopening the job silently.
Evidence should prove the task
Require evidence according to work specification: asset barcode/NFC, before/after photo, calibrated measurement, port or fiber test, serial installed/removed, GPS within allowed precision, safety checklist, customer signature where lawful, and exception reason. Hash media, preserve capture time and device, and prohibit gallery imports for evidence types that require live capture.
Parts inventory is a reservation system
Track warehouse, van, technician custody, reserved, installed, removed, defective, returned, and consumed states. Scan each transfer and preserve compatible resource specification. A work order reserves expected stock but permits controlled substitution. Closure should update installed-resource inventory only after evidence and validation, not merely decrement a spreadsheet.
Verification must use service telemetry
A technician marking “fixed” is not proof that the circuit recovered. Run approved tests, check alarms and performance, validate optical/power/radio levels, confirm service path, monitor a soak window, and obtain customer confirmation when appropriate. Auto-close only when field evidence and service health agree. Otherwise create follow-up work or return the incident to diagnosis.
Appointments and communication share one timeline
TMF646 models appointment slots and characteristics. Keep proposed, offered, accepted, travelling, arrived, access failed, completed, cancelled, and rescheduled states. Notify customers from authoritative events, share realistic arrival windows, minimize personal data, and record delivery. A technician's location should be exposed only when policy, safety, and purpose allow it.
Inter-provider faults need contract boundaries
Enterprise connectivity often crosses access carriers, wholesale providers, tower companies, power utilities, and contractors. MEF LSO Cantata/Sonata and trouble-ticket standards provide a basis for partner automation. Exchange product/circuit references, impact, severity, test evidence, milestones, commitments, and closure while keeping internal topology, workforce details, and sensitive customer data appropriately scoped.
Measure restoration quality
Track detection-to-ticket, triage, dispatch, travel, access delay, first-time fix, repeat fault, no-fault-found, appointment success, SLA attainment, parts accuracy, inventory correction, offline sync delay, evidence rejection, safety exceptions, customer updates, mean restore time, and service health after closure. Segment by asset type, region, contractor, cause, shift, and network technology.
What I would build
An event intake and correlation layer; service/resource/site graph; incident and trouble-ticket adapters; qualification rules; work-order service; appointment API; skill/shift/stock registry; routing solver; offline encrypted mobile app; media/evidence store; parts ledger; network test integration; SLA clock engine; customer/partner notification hub; and verification workflow. Every state change would carry actor, event time, source, reason, version, and correlation IDs.
The principle
Field-service automation succeeds when the right qualified team reaches the right asset with the right context and parts, can work without connectivity, and returns evidence that both infrastructure and customer service recovered. Optimizing dispatch alone only makes bad work arrive faster.