What to carry into the operating review
- Approval is an intermediate state, not the commercial outcome. A confirmed order that never ships or delivers does not create the same value as a completed order.
- Every rate needs an explicit numerator, denominator, cohort date and maturity window. “Delivery rate” without those definitions is not comparable.
- Report both stage-to-stage conversion and accepted-lead-to-delivery. The first locates the constraint; the second protects the business view.
- Keep call outcomes, fulfilment statuses and courier outcomes separate. That makes ownership clearer without pretending one team controls the entire journey.
In a lead-to-delivery operation, “approval rate” can become the loudest number in the room. It arrives quickly, is easy to compare by source and appears to describe call-center performance. But approval is only one transition in a longer commercial system. A stronger review asks two questions at the same time: where does the funnel lose orders, and what happens to contribution after every loss and cost?
This distinction matters in cross-border African operations because the caller, advertiser, fulfilment partner and last-mile carrier may be separate companies. Market context also changes the customer journey. GSMA’s work on African ecommerce identifies trust as a material adoption challenge, while the International Trade Administration’s Ghana ecommerce guide highlights last-mile delivery and digital-identity verification among practical challenges. A call can reduce uncertainty, but it cannot erase stock, routing, address or courier problems downstream.
1. Define the funnel before comparing performance
A status should describe an observable event. Start with a shared dictionary that all parties can map to, even when their internal systems use different labels.
| Stage | Minimum definition | System of record |
|---|---|---|
| Submitted lead | A form or order event reached the advertiser or tracker. | Tracker or storefront |
| Accepted lead | The record passed agreed technical checks and entered the calling queue. | Call-center CRM |
| Attempted | At least one documented outbound attempt was made under the agreed schedule. | Dialler or CRM |
| Contacted | The intended customer—or an agreed authorised respondent—was reached. | Call outcome |
| Qualified | The customer and order meet the advertiser’s written acceptance rules. | Call outcome |
| Confirmed | The customer explicitly agrees to the final product, quantity, price, payment and delivery details. | Call outcome plus audit trail |
| Fulfilment accepted | Stock and operational rules allow the order to enter fulfilment. | OMS or warehouse |
| Dispatched | The carrier has accepted the parcel and a shipment identifier exists. | Fulfilment or carrier |
| Delivered | The carrier records completed delivery under the agreed proof standard. | Carrier or fulfilment |
| Settled | Collected funds have reconciled, net of agreed adjustments. | Finance or fulfilment |
Do not let “approved” absorb several of these states. If one market counts a verbal “yes” as approval while another requires a validated address and final price, the rates describe different events. Version the definition whenever the acceptance rule changes.
A useful status is reproducible: a second reviewer looking at the same evidence should reach the same outcome.
2. Use stable denominators and mature cohorts
Publish the formula next to the number. Two approval views are useful, but they answer different questions:
Contact-to-confirm rate = confirmed orders ÷ contacted leads
Accepted-lead approval rate = confirmed orders ÷ accepted leads
Dispatch rate = dispatched orders ÷ fulfilment-accepted orders
Delivery rate after dispatch = delivered orders ÷ dispatched orders
Lead-to-delivery rate = delivered orders ÷ accepted leads
Stage-to-stage rates locate a constraint. The end-to-end rate prevents local optimisation: a team cannot declare victory because confirmation improved if lead-to-delivery or contribution deteriorated.
Cohort by when the lead entered the system
A daily operational dashboard may show events that happened today. A performance review should normally group records by lead creation or queue-acceptance date and follow that same cohort through the journey. Otherwise, today’s deliveries are compared with today’s newly approved orders even though they are not the same customers.
- Use one timezone per market and store the original timestamp plus UTC.
- Define when a cohort is mature enough for delivery reporting; keep immature cohorts visibly provisional.
- Preserve late carrier updates instead of rewriting the historical export without an audit note.
- Deduplicate with an agreed rule, but retain the original records and the reason for exclusion.
- Segment by market, offer, product variation, source, affiliate/subID and fulfilment lane before drawing a conclusion.
Important: A percentage without the underlying count can be misleading. Every view should expose both the numerator and denominator, especially for small segments.
3. Read rate movements as hypotheses, not verdicts
A rate movement points to a place to investigate. It rarely proves a single cause on its own.
| Observed pattern | Questions to test next |
|---|---|
| Contact falls; confirmation among contacts is stable | Did lead age increase? Are phone formats valid? Did source mix, dial schedule, retry policy or telecom reachability change? |
| Contact is stable; confirmation falls | Did the offer, price, landing-page promise, script, language, stock variant or traffic audience change? |
| Confirmation rises; delivery falls | Are acceptance rules too loose? Are customers hearing the final price and timing? Did address quality, dispatch delay, courier lane or attempted-delivery behaviour change? |
| Confirmed orders fail before dispatch | Is stock available? Did order sync fail? Are service zones, duplicate rules or fulfilment cut-offs rejecting orders? |
| Delivery is stable; contribution falls | Did acquisition cost, average order value, discounting, cost of calling, fulfilment cost, returns or collection cost change? |
The correct next action is usually a joined record review: the tracker event, call timeline, final order, fulfilment status and carrier outcome for the same order ID. Aggregate dashboards show where to look; sampled records explain what happened.
4. Assign responsibility without breaking the partnership
Affiliate marketing separates roles: publishers promote advertiser products or services and earn commission under the programme, while a network may track and administer the relationship. The operating chain can add a call center, fulfilment partner and carrier. Reporting should preserve that separation.
- Advertiser: product truth, price, stock, acceptance policy, approved claims and customer promise.
- Affiliate or media source: audience, creative, landing-page representation and valid tracking identifiers.
- Call center: lead handling, attempts, script adherence, customer verification, evidence and status transmission.
- Fulfilment: stock acceptance, picking, packaging, dispatch timing and handoff.
- Carrier: delivery attempts, proof, reason codes, cash handling where applicable and final status.
Use those responsibilities to route investigations, not to auto-assign blame. For example, “customer refused” can originate in a misleading ad, a pressured confirmation, a long dispatch delay, a changed delivery fee or a poor delivery attempt. The final courier code is an observation, not a root-cause analysis.
5. Run a review that ends in a controlled action
Daily: protect service
Watch queue age, attempts, contactability, system errors, unprocessed records, stock exceptions and stale courier statuses. The goal is to prevent avoidable loss, not to explain the entire market every morning.
Weekly: diagnose cohorts
- Freeze the cohort cut and maturity rule.
- Review counts and conversions from accepted lead through delivery.
- Segment only where the sample is large enough to inspect responsibly; always show the count.
- Listen to or audit a defined sample across positive and negative outcomes.
- Join selected records to fulfilment and carrier events.
- Choose one operational change, name its owner and specify the guardrail metrics.
Monthly: review unit economics
Add acquisition, calling, fulfilment, delivery, collection, return and product-cost assumptions appropriate to the commercial agreement. A simple contribution view is more useful than optimising approval alone:
The exact accounting boundary is a commercial decision. Make it explicit, use the same boundary when comparing cohorts and keep taxes, refunds, chargebacks and fixed overhead visible if they are excluded. Our unit-economics calculator provides a transparent starting model rather than a performance promise.