Kaspi.kz: Building an Ecosystem Around Everyday Customer Needs

Kaspi.kz illustrates connected value between consumers and merchants. The lesson for a smaller business is not to build a super app. It is to identify adjacent tasks that become easier when information and service continuity are connected. Distinguish the company's reported scale from our interpretation of the mechanisms behind it.

The consumer app and merchant-facing Kaspi Pay connect payments, commerce and financial services. A consumer needs a convenient way to choose, pay and receive a result. A merchant needs demand, payment acceptance and operating support. Solving related tasks can make an ecosystem useful, but adding unrelated services can create complexity rather than value.

The strategic choice is to solve connected customer jobs in a coordinated system. Repeated useful interactions may support cross-adoption, while merchant participation can improve consumer choice. This is a platform-economics interpretation. It is not proof that any two services create a network effect or that bundling alone causes profitable growth.

The relevant evidence snapshot is year-end 2025. During 2025, the acquisition of control in Hepsiburada changed the scope of consolidated reporting. The evidence cards below explicitly refer to Kazakhstan; Turkish operations and group consolidated figures must not be silently mixed with them [1].

The 2025 Form 20-F reports approximately 15.7 million Average MAU and 764,000 Active Merchants for the Kazakhstan apps, and 77 Monthly Transactions per Active Consumer at year end [1]. Average MAU is a defined activity measure, not all registered customers. Active Merchants are not simply marketplace sellers. These measures have different denominators and cannot be added together.

Use platform economics to inspect both sides of the interaction and jobs-to-be-done to identify the linked task. Ask who benefits from another participant joining, what coordination cost is removed, and whether an incentive or subsidy is doing the work. The cost of trust, support, payment operations and regulation belongs in the analysis.

Connected everyday utility is a plausible advantage. Copying the interface, number of features or promotional spending would miss the underlying customer problem. Do not infer a small firm's attainable audience, margins or regulatory permissions from Kaspi's reported scale. Evidence of engagement also does not isolate AI's contribution.

Scenario 1: develop one second service around an existing customer need. Needs: clear ownership and a shared identity with appropriate permission. Risk: low adoption or a fragmented handoff.

Scenario 2: integrate a partner's service before owning it. Needs: service-level rules and economical referral terms. Risk: dependence on partner quality. These are applications of the model, not disclosed historical alternatives.

AI can help selection, service triage or merchant workflows when data and permissions support it. A recommendation is useful only if its quality and subsequent customer outcome improve. The historical ecosystem figures are not an AI impact estimate. Protect human escalation and avoid collecting extra personal data just because a model could use them.

Retail: connect selection, delivery or repair for the same customer.

B2B: link quotation, order status and replenishment.

Service: test booking plus preparation or follow-up. Begin with one adjacent task and a simple operational agreement rather than a full platform.

Hypothesis: offering one useful second service after completion of the first task will increase repeat engagement because it removes an adjacent friction. Days 1–5: identify the task and baseline. Days 6–10: prepare a manual connected offer. Days 11–18: compare eligible customer cohorts. Days 19–21: review adoption and full costs.

Primary metric: adoption of the second useful task or a defined return rate. Guardrails: complaints, consent, margin and support time. Owner: customer or product lead. Resources: one partner or operator and a cohort log. Scale if adoption and economics meet pre-agreed criteria; iterate if relevance is clear but the handoff fails; stop on harm or weak demand. Observe longer if the natural repurchase interval exceeds 21 days.

This analysis uses the listed public publications and accessible research descriptions. Key definitions and limitations appear beside the figures. Corporate estimates remain company-reported; interpretations, alternative scenarios and experiments are RA LAB editorial work. A full independent audit of every publication was not performed, and we do not claim to have read restricted full texts. The research does not imply affiliation with the companies.

Compare Decathlon's connected retail journey with this ecosystem model. Then choose one adjacent job in the Business Problem Solver and formulate a bounded experiment.