A delivery manager usually discovers a visibility problem through a phone call. The customer says the order has not arrived, the dispatch sheet says it left at 10am, and the rider is not answering. By the time three people have checked three systems, the customer has called again.
Visibility is not a map with a moving dot. It is the ability to answer four questions for any order at any moment: where is it, who has it, what happened last, and what happens next. If your team cannot answer all four without phoning someone, you have a data problem before you have a software problem.
Seven symptoms and what they usually mean
- Status only changes at the hub. Events are captured at dispatch and delivery, nothing in between. The system is a milestone log, not a tracking system.
- Customers know more than the office. Riders report by phone, so the customer hears first and the record is updated last.
- Failed attempts appear only at day end. There is no exception event, so a failure is recorded as a late delivery or not at all.
- Addresses are a phone call away. The order record lacks a landmark, a contact number that works, or a verified zone.
- Two systems disagree about the same order. No shared identifier, or a status vocabulary that differs between dispatch, rider app and accounts.
- Proof of delivery is disputed. Paper slips, photos on a personal phone, or a signature that cannot be linked to an order.
- Fleet utilisation is a guess. You know how many vehicles you own, not how many were productive in the last hour.
Cause and fix
No event model
Cause: the system stores a status field that gets overwritten. Fix: append-only events for pickup, arrival, attempt, exception, delivery and cash collection, each with a timestamp, actor and device. A status is a view over events, not the source of truth.
Coordination lives in WhatsApp
Cause: riders send location and exceptions to a group chat. Fix: keep the chat for escalation, but require the rider app to record the exception before the chat message is sent. If the app is slower than the chat, the app will lose.
GPS without an order link
Cause: the tracker reports a vehicle position, but the dispatch system cannot say which orders are on that vehicle. Fix: assign orders to trips, and trips to vehicles or riders. Without that link, GPS is a dashboard nobody can act on.
Address quality
Cause: free-text addresses with no zone, landmark or verified phone. Fix: structured address fields, a zone list that matches your routing, landmark and a phone check at order entry. Measure failed attempts by zone; the pattern is usually obvious within a month.
No exception vocabulary
Cause: every problem becomes "customer not available". Fix: a short reason list: wrong address, refused, no cash, shop closed, unreachable, damaged, rescheduled. Reason codes turn anecdotes into a fixable queue.
Paper proof of delivery
Cause: the rider collects a signature on a slip that reaches accounts three days later. Fix: capture proof in the app at the door: photo, signature or one-time code, with the order ID embedded. Sync when the network returns.
ETAs from straight-line distance
Cause: the estimate ignores traffic, stop density and service time. Fix: calculate from historical stop durations by zone and hour, and show a window rather than a single time. A wrong precise time damages trust more than an honest range.
Diagnostic sequence
- Pick twenty recent deliveries, including at least five failures.
- Reconstruct each timeline from every system: order, dispatch, rider chat, GPS, payment and accounts.
- Find the first missing event. That is usually where the process breaks, not where the complaint appeared.
- Compare the status vocabulary between systems and list every mismatch.
- Time-stamp the gap: how long between the real event and the recorded event?
- Check the device: low battery, poor connectivity and outdated app versions explain a surprising share of "missing" updates.
- Interview two riders and one dispatcher separately. Their workarounds are the real process.
What to instrument
- Order: created, confirmed, allocated, dispatched, delivered, returned.
- Trip: started, stop sequence, distance, fuel or charge where relevant.
- Stop: arrived, service start, service end, proof captured.
- Exception: reason code, free-text note, photo, next action.
- Cash: expected amount, collected amount, handover, variance.
Every event needs a device timestamp and a server timestamp. The difference between them exposes offline gaps that a single timestamp hides.
Bangladesh-specific constraints
Dhaka routing punishes straight-line estimates: a two-kilometre gap can take far longer than the map suggests at 5pm. Addresses often rely on landmarks and informal area names, so the zone list must match how riders actually navigate, not how a map provider labels the city. Cash on delivery adds a second reconciliation problem, because the delivery event and the cash handover are different events. Third-party rider fleets add a third: you do not control their app, so agree on an event feed and a dispute process in the contract.
Metrics
- First-attempt delivery rate, by zone.
- On-time delivery against the promised window.
- Status freshness: median minutes between a real event and the recorded event.
- Exception capture rate: failed attempts with a reason code.
- Proof-of-delivery capture rate at the door.
- Cash variance per 100 deliveries.
- Stops per hour and service time per stop.
Where this fits in our work
If you need the event model, rider app and dispatch view built or repaired, start with Operations & Industry Systems; the pricing page explains how field-system work is scoped. For the stock side of the same operation, read what inventory synchronisation costs.
Sources
- World Bank, Logistics Performance Index, accessed 17 November 2026.
- GS1, EPCIS standard for supply-chain event data, accessed 17 November 2026.
- Meta, WhatsApp Business Platform documentation, accessed 17 November 2026.
Responses (0)
No comments yet. Add the first one below.
Leave a response
Your email address will not be published. Required fields are marked *.