What to carry into the operating review
- Compare operating systems, not seat prices. Recruitment, local management, telecom, QA, compliance, integrations, reporting and continuity are part of the capability.
- An internal team maximises direct control but requires local management capacity. Outsourcing can compress setup work but never outsources the advertiser’s accountability for the offer and customer promise.
- A hybrid is often the most testable path: outsource an initial market or overflow lane, retain product and commercial ownership, then decide what should remain external.
- Demand evidence from either option. The question is not “Do we trust the team?” but “Can we observe, audit and improve the operation?”
An advertiser entering Africa rarely needs “calls” in isolation. It needs a repeatable capability: receive leads, reach customers in the right context, qualify and confirm orders, transmit accurate data, inspect quality, resolve exceptions and learn from delivery outcomes. Build-versus-outsource is a decision about how to own that capability.
External market research is useful because it reveals why a uniform playbook is risky. GSMA identifies trust as a challenge in African ecommerce, while the official Kenya and South Africa guides describe different payment, logistics, addressing and consumer environments. But local context alone does not make a provider effective. The operating model still needs product truth, clear ownership, connected data and management discipline.
1. Define what you are deciding to build or buy
Write the scope as outcomes and controls before asking for headcount or a quote.
- Market coverage: countries, service zones, languages, local hours and expected launch sequence.
- Workflow: lead response, qualification, order confirmation, upsell, reactivation, delivery follow-up and customer support.
- Decision rights: who may change a script, reject a lead, adjust an order, grant an exception or contact a customer again.
- Systems: tracker, storefront, CRM, dialler, order management, fulfilment, carrier and business intelligence.
- Controls: access, recordings where lawful, QA, coaching, audit trail, incident handling, continuity and data retention.
- Commercial output: statuses, reconciliation, invoices, source-level reporting and mature delivery cohorts.
If the internal proposal includes only agent salaries and the provider proposal includes a managed operation, the comparison is structurally unfair. Conversely, a provider should not label a basic staffing service “fully managed” when the advertiser will still design the workflow, supervise agents and repair every integration.
2. Score the decision against your actual constraints
| Decision factor | Internal build tends to fit when… | Outsource tends to fit when… |
|---|---|---|
| Launch sequence | The market is strategic, durable and the company can wait for entity, hiring and management setup. | The company needs a controlled pilot or staged entry before committing to permanent local infrastructure. |
| Volume shape | Demand is predictable enough to support a stable team and management layer. | Volumes are volatile, seasonal, source-dependent or likely to move between markets. |
| Existing capability | The company already has local HR, telecom, QA, IT, compliance and operational leadership. | Those functions would otherwise need to be established specifically for the launch. |
| Workflow sensitivity | The call contains highly proprietary judgement that cannot be sufficiently codified or audited externally. | The workflow can be documented, trained, sampled and improved under clear decision rights. |
| Multi-market need | The company is committed to a single large market and wants a long-term local organisation. | It needs to test or operate several markets without duplicating every support function. |
| Management attention | Leadership wants call operations as a core competency and can staff the management layer. | Leadership wants to retain offer and economics ownership while delegating day-to-day operations. |
Score each factor, attach evidence and name the assumption that could reverse the decision. Do not add the scores mechanically; some factors are gates. A legal, product-safety, data-access or continuity failure cannot be averaged away by lower cost.
3. Compare total cost over the same period and scope
Use a period long enough to include setup and stabilisation. Separate one-time, fixed, variable and failure costs.
Then connect the operating model to funnel economics. A lower seat or per-lead price is not cheaper if contact coverage, data accuracy, customer treatment or delivery contribution deteriorates. A higher managed fee is not justified unless the included management and controls are observable.
Normalise commercial units
- Define what counts as a processable lead and who pays for duplicates, missing fields or invalid markets.
- Distinguish staffed hour, productive hour, attempt, contacted customer, confirmed order and delivered order.
- State minimums, ramp-down notice, training charges, telecom, software, weekends, languages and management coverage.
- Model currency exposure, tax treatment and the cost of joining data across providers.
- Keep “success fee” definitions auditable; do not let the pricing model encourage pressured or low-quality confirmations.
Use scenario ranges, not one forecast. Test at least the launch case, expected case, traffic spike, volume decline and market pause. Inputs should be your assumptions—not vendor benchmarks presented without definitions.
4. Test a provider as an operating system
A capability review should examine artefacts and live demonstrations, subject to customer-data confidentiality.
- Market reality: Where are agents and supervisors located? Which languages are currently available? How is local context tested rather than asserted?
- Management: Who owns the account, floor, QA, training, workforce plan, incident response and technology?
- Workflow: Ask the provider to map one lead from intake through attempts, confirmation, fulfilment handoff, delivery feedback and correction.
- Quality: Review the scorecard, critical-failure rules, calibration, coaching evidence and dispute process.
- Data: Inspect role-based access, exportability, audit logs, retention, subprocessor handling, incident notification and end-of-contract deletion.
- Technology: Demonstrate authentication, retries, idempotency, monitoring, reconciliation and manual recovery—not just a dashboard.
- Continuity: Ask what happens during telecom failure, power loss, staff absence, volume spikes and loss of a site or vendor.
- Commercial alignment: Trace each billable unit back to a source record and test a sample invoice dispute.
Request evidence that can be inspected repeatedly. A polished reference call is useful; an auditable workflow is stronger.
5. Choose a staged model, not a permanent ideology
Internal operation
Best considered when the market is central to strategy, volume can support local leadership, the company wants customer operations as a core capability and it can build the surrounding controls—not only hire agents.
Managed outsourced operation
Best considered when an international company needs a faster controlled entry, specialised local coverage, variable capacity or several markets. The advertiser should retain product, claims, customer promise, acceptance policy, access governance and commercial accountability.
Hybrid operation
Useful patterns include an outsourced launch with an internal market owner; internal core capacity plus outsourced overflow; separate providers by market with one data model; or an outsourced front line with internal high-risk escalation. A hybrid only works when records and decision rights move cleanly across the boundary.
Build after learning
A pilot can reveal the languages, shift pattern, objection profile, management load, system requirements and viable service area before a permanent team is designed. If insourcing remains an option, contract for data portability, documentation, training artefacts and a transition path from day one.
6. Make reversibility part of the design
Whether moving to a provider or building internally, protect the customer and the evidence.
- Use advertiser-owned identifiers and, where practical, advertiser-controlled system accounts.
- Export lead, attempt, order, status, QA and change-history data in documented formats.
- Maintain current scripts, reason-code definitions, training records and process maps.
- Define access removal, data return/deletion, number portability where possible and customer-case handoff.
- Set an overlap period and reconciliation test before traffic is switched.
- Retain a tested fallback for lead intake and status delivery during transition.
The final decision should produce a named capability owner, a staged launch, measurable service boundaries and a credible exit. Outsourcing is not abdication; building is not automatically control. Control comes from clear decisions, competent management and evidence that follows the customer journey.