A warehouse-to-driver dispatch system for 500 daily UAE orders costs AED 192,000–312,000 in the worked software budget below. It includes one order import integration, parcel identification, loading scans, a driver app, dispatch visibility, proof of delivery and basic returns handling. An illustrative first-year software envelope comes to AED 313,000–433,000 after adding the monthly operations and launch allocations described here.
These figures use assumed work hours and rates. They are planning scenarios rather than surveyed supplier prices. Replace the inputs with estimates for your systems, devices and operational rules before approving a project.
What does 500 daily orders actually mean?
Imagine a fictional Dubai distributor operating one warehouse and 20 delivery vehicles. It receives 500 orders on a normal operating day. Most orders contain one parcel, but some contain two or more. Orders are packed before the dispatch cut-off and loaded by route.
For this scenario, an order averages 1.4 parcels. That means 700 parcels daily and 18,200 parcels over 26 operating days. Each parcel can produce several scan events. Pricing by orders alone would hide the extra handling and event volume.
The operation already has a warehouse system that records picking and packing. The new system receives packed orders, validates their parcel IDs and controls the handoff to drivers. It does not replace inventory valuation, purchasing or warehouse replenishment. If those capabilities are required, estimate them as additional product scope.
A customer mobile app is outside the first release. Customers receive a basic status view, while drivers use a mobile application and dispatchers use a web console. This gives the pilot a clear operational boundary.
Which records connect the warehouse to the doorstep?
| Record | Purpose | Stable identifier |
|---|---|---|
| Order | Customer request and promised delivery | Source order ID within the merchant account |
| Parcel | Physical handling unit | Parcel barcode or internal parcel ID |
| Shipment | Group of parcels moving together | Shipment ID |
| Route | Driver's assigned stop sequence | Route and version ID |
| Scan event | Observation or action on a parcel | Client event ID within the account |
| Delivery attempt | Result of visiting a stop | Attempt ID |
| Return | Reverse movement and receipt | Return ID linked to the original parcel |
Keep the relationships explicit. A three-parcel order can be partially delivered. A shipment can move to another route before departure. A return can be received days after the customer requested it.
Define whether the customer promise applies to the whole order or individual shipments. That affects the tracking page, exception messages and reports. The underlying data model should support the policy the operations team will actually follow.
Map the daily workflow before estimating screens
The day begins when the source system marks parcels ready. The integration imports them with their source references and records validation failures. A dispatcher groups ready shipments into a draft route. Loading staff scan parcels against that route, and the system flags missing, unexpected or already-loaded items.
The route becomes dispatchable only after the agreed checks pass. A driver acknowledges the handoff, sees the current stop list and submits attempt outcomes. Proof of delivery updates the parcel and order according to their relationships. Failed deliveries remain visible for review and a subsequent attempt or return.
A dispatcher can override a decision when authorized. The override should capture the reason and affected route version. A silent status edit makes later reconciliation difficult, particularly when the warehouse and driver disagree about custody.
Include the return flow from the start. At minimum, the team needs to know which parcel is coming back, why, who has it and whether the warehouse has received it. Refund approval and inventory restocking can remain separate decisions in the existing systems.
One-time development budget
Assume a blended rate of AED 200 per hour. The total includes one agreed warehouse interface and a defined pilot device set.
| Work package | Hours | AED |
|---|---|---|
| Discovery and workflow design | 100–160 | 20,000–32,000 |
| Warehouse/order integration | 140–240 | 28,000–48,000 |
| Parcel scans and loading controls | 120–200 | 24,000–40,000 |
| Driver application | 180–280 | 36,000–56,000 |
| Dispatch web console | 120–190 | 24,000–38,000 |
| Backend, access controls and audit history | 160–250 | 32,000–50,000 |
| QA, pilot and handover | 140–240 | 28,000–48,000 |
| Total | 960–1,560 | 192,000–312,000 |
The model includes Arabic and English driver workflows, a shared mobile codebase, one operating account and one warehouse. It assumes the source system exposes a usable interface or agreed export. The team owns its data and third-party accounts under the proposed handover arrangement.
A project with undocumented source fields, unreliable parcel labels or conflicting order states needs discovery work to resolve those issues. Ask for a revised estimate after the audit instead of treating the initial range as a fixed quote for unknown integrations.
The model excludes physical scanners, phones, label printers, fleet expenses, taxes and bespoke financial settlement. Obtain those separately if they are part of the rollout.
What pushes this system beyond the baseline?
| Change | Why implementation expands |
|---|---|
| Multiple warehouses | Transfers, source ownership and branch permissions |
| Multiple merchants | Account isolation, import mappings and reporting |
| Partial deliveries | Parcel-level outcomes and customer communication rules |
| Cash collection | Collection, deposit and discrepancy workflows |
| Rugged scanners | Hardware integration and real-device certification |
| Complex route optimization | Capacity, windows and replanning |
| Extended offline work | Local persistence, recovery and conflict handling |
| Carrier handoff | External tracking, labels and custody reconciliation |
Estimate each change with acceptance cases. “Offline support” can mean queuing a single completed stop or running an entire shift without connectivity. Those commitments have different implementation and testing costs.
For an additional warehouse interface, inspect the actual API or file format and failure behavior. A second integration that resembles the first may still use different identifiers, status meanings or update rules. Reuse the proven components after checking those differences.
A small JavaScript example for repeated scans
Mobile networks can lose responses after the server has saved a scan. The device then retries the same event. Use a stable event ID for the retry and enforce uniqueness in durable storage.
The example below prepares a canonical payload for one specific action, confirming that a parcel was loaded onto a route. It uses Node.js's built-in cryptographic module to detect reuse of an event ID with different content. It is a small request-preparation example. Database transaction handling follows afterward.
const { createHash } = require("node:crypto");
function prepareLoadingScan(input, trustedTenantId) {
if (
typeof trustedTenantId !== "string" ||
trustedTenantId.length === 0
) {
throw new Error("A trusted tenant is required");
}
const names = [
"eventId",
"parcelId",
"routeId",
"occurredAt"
];
for (const name of names) {
const value = input[name];
if (
typeof value !== "string" ||
value.length === 0 ||
value.length > 200
) {
throw new Error("Invalid field: " + name);
}
}
const timestamp = new Date(input.occurredAt);
if (
Number.isNaN(timestamp.getTime()) ||
timestamp.toISOString() !== input.occurredAt
) {
throw new Error("Use canonical UTC timestamp format");
}
const payload = {
tenantId: trustedTenantId,
eventId: input.eventId,
parcelId: input.parcelId,
routeId: input.routeId,
occurredAt: input.occurredAt,
action: "parcel_loaded"
};
const payloadHash = createHash("sha256")
.update(JSON.stringify(payload))
.digest("hex");
return { payload, payloadHash };
}
Derive the tenant and actor from verified server-side context. The event timestamp describes what the device reports. Store a separate server receipt time because a device clock can be wrong or a scan can arrive later.
The device creates the event ID once, saves the event locally and reuses both on retry. A second genuine physical scan can have a new event ID. Event deduplication alone therefore does not prevent two different events from attempting the same loading transition.
What must happen in the database?
Create a uniqueness boundary on the trusted account and event ID. In one transaction, attempt to insert the event, validate the parcel and route relationship, apply the loading transition, save the result and commit.
If the event already exists, compare its stored hash with the incoming hash. Matching content returns the saved outcome. Different content under the same key produces a conflict that the client must investigate. Use parameterized database queries.
For a new event, lock or conditionally update the relevant parcel assignment. A parcel already loaded onto another active route should produce a controlled exception. Save that outcome consistently with the event. A process crash before commit should roll back both the event and transition.
Enforce route versions or another concurrency rule when dispatch changes the manifest. If the worker scanned against an outdated draft, explain the conflict. Avoid accepting the scan and silently attaching the parcel to a route the worker never saw.
This workflow requires database integration testing. The included snippet only prepares and validates event content. Its checks demonstrate what is accepted at that boundary, without claiming to implement transaction-safe loading on its own.
Monthly costs at the stated volume
Assume 500 daily orders over 26 operating days, or 13,000 orders monthly. Assume four retained scan events per parcel and 1.4 parcels per order. That yields 72,800 scan events. Storage depends on payloads, evidence attachments, retention and access patterns.
| Monthly item | Assumed calculation | AED |
|---|---|---|
| Hosting, database and monitoring | Planning allowance | 2,000 |
| Maps and status-related requests | 39,000 requests × 0.02 | 780 |
| Paid customer messages | 13,000 orders × 1 × 0.10 | 1,300 |
| Evidence storage and retrieval | Allowance based on retention plan | 220 |
| Engineering support | 20 hours × 200 | 4,000 |
| Monthly software total | 8,300 |
Every rate in this table is an editable budgeting assumption. Actual providers may use different units, minimums and service tiers. A request can contain several billable elements, so confirm the billing model for the chosen architecture.
The listed software expense is about AED 0.64 per imported order. The per-successful-delivery figure depends on completed volume. At 12,000 completed orders, the same total is about AED 0.69. Label the denominator when comparing operations.
What is the first-year cost?
Twelve months at AED 8,300 produces AED 99,600. Add an illustrative launch allocation of AED 8,000 for label and mapping cleanup, AED 7,400 for shift training and AED 6,000 for restore and operational rehearsal. Those launch tasks are outside the development packages above.
The additional first-year allocation is therefore AED 121,000. Added to the build range, the total becomes AED 313,000–433,000 before the excluded physical and business costs. If support begins midway through the year, model the actual active months instead of charging twelve automatically.
Keep enhancement requests outside the support allowance unless the agreement explicitly includes them. A support hour used for incident recovery cannot simultaneously fund a new merchant integration.
Pilot a full operational day
Test a representative set of 100 orders before releasing the entire daily volume. Include multi-parcel orders, a missing parcel, a wrong-route scan, a failed delivery and a return received the next morning. Run the pilot on the intended device types.
Reconcile the warehouse-ready list with the loaded list and delivery outcomes. The route summary should explain every accepted parcel as awaiting loading, loaded, delivered, failed, returned or in a named exception state. Avoid an unexplained “other” bucket.
Have warehouse staff and drivers verify the screens in their preferred language. Test labels in realistic lighting, damaged barcodes and gloved handling where relevant. Measure the scanning time and exception resolution effort rather than relying only on a classroom demonstration.
Acceptance cases that belong in the contract
| Case | Expected result |
|---|---|
| Same event retried | Saved result returned |
| Same event ID, changed parcel | Conflict recorded |
| Two devices load one parcel onto different routes | One permitted assignment and one visible exception |
| Source imports an order again | Existing record reconciled under agreed update rules |
| Order contains three parcels and one fails | Partial state remains explainable |
| Device reconnects after dispatch changes | Version conflict is reviewed |
| Driver uploads proof twice | One effective delivery outcome |
| Return reaches the warehouse | Receipt linked to original parcel |
Validate access through normal application entry points. A driver should receive assigned route data, and warehouse permissions should match the site's responsibilities. Record who can override exceptions and verify that an override leaves an audit trail.
How to compare proposals fairly
Give vendors the same sample export, parcel-label examples, device list and state-transition policy. Ask for itemized integration assumptions and a list of fields the source system must supply.
Require an explanation of retries, out-of-order events, offline recovery and the process for finding a parcel's current custodian. Ask how the team will monitor pending imports and unresolved scans during a shift.
Compare the full first year, including account ownership, hosting, support hours, release responsibilities and data export. Agree how accepted work is demonstrated. A pilot involving the people who load and deliver parcels provides stronger scope evidence than a screen inventory alone.
Frequently asked questions
Can the first release use phone cameras?
Often yes, if real-device tests meet throughput and barcode-readability requirements. Price specialist scanner integration when the operation's devices or pace require it.
Can we keep the existing warehouse system?
Yes. This scenario preserves its picking and packing responsibilities and adds the warehouse-to-driver handoff. Specify which system owns each status and how corrections propagate.
Do 500 orders require microservices?
Order volume alone does not decide the architecture. Peak concurrency, reliability requirements and team operating capacity matter. A well-structured application with durable jobs can be a valid starting design.
What should we fund before the full build?
Fund a workflow and data audit, a loading-scan prototype and a representative import. Their results make the build estimate and pilot acceptance conditions much more concrete.
Top comments (0)