Nubank's operational costs have been increasing faster than revenue. How would you approach this?

Question

Accepted Answer

Confirm the objective. Let me restate the objective. Operational costs are growing faster than revenue, so margins are compressing. My goal is to diagnose which cost categories are driving this, find the root causes, and recommend ways to reduce or control costs without disrupting what works. Before I structure this, are we trying to hit a specific margin target, or to understand the drivers first and then set targets?. The interviewer says, Understand the drivers first, then propose solutions to reverse the trend. Clarifying questions, five batches. Cost equals COGS plus OpEx. Decompose both and find where costs are growing faster than they should.. Batch 1, scale and timing. What is the current cost-to-revenue ratio versus 12 months ago? Is the increase sudden or gradual? And did anything change in the business at the same time, headcount expansion, new-market entry, a product launch, or a vendor change?. The interviewer says, Cost-to-revenue was 65% a year ago, now 82%. Gradual increase over 9 months. Nubank expanded into 5 new markets and headcount grew 45% in that period.. Batch 2, decomposition. I want to split costs into COGS, the direct cost of delivering the product per customer, and OpEx, the overhead to run the business. Which is growing faster? And within each, what are the biggest line items, people, infrastructure, acquisition, compliance?. The interviewer says, Both are growing but OpEx faster. People costs are the biggest driver, 60% of the total cost base. Customer support costs have doubled. Compliance and legal have grown three times in the new markets.. Batch 3, segmentation. On support costs doubling, is that a higher ticket rate per customer or just more customers? Are tickets disproportionately from new markets or existing ones? And on compliance, are the costs one-time setup or recurring?. The interviewer says, Support ticket rate per customer is up 40%, not just volume. New markets drive 65% of the additional tickets. Compliance costs are roughly 40% one-time and 60% recurring, mainly local regulatory reporting.. Batch 4, unit economics. What is the cost per customer in new markets versus established ones? And do we have a payback-period model for the expansion, how long until each new market is revenue-positive? I want to know if this is investment, temporary, or structural, permanent.. The interviewer says, Cost per customer in new markets is 2.3 times established markets. No formal payback model exists. Expansion was driven by growth targets, not unit economics.. Batch 5, constraints and impact. Are there constraints on headcount reduction, local employment law or hiring commitments? And what is the cost of not acting, if costs continue at this rate, when does the business hit a margin floor that forces external action?. The interviewer says, No major employment-law constraints. At the current trajectory margins compress to single digits within 6 months. Leadership is treating this as high priority. Structure the problem, MECE. Profit equals revenue minus costs, and revenue is growing, so this is a cost-structure problem. Three buckets. Bucket A, direct delivery costs: cost per customer, support, infrastructure. Bucket B, expansion overhead: compliance, legal, local operations driven by new markets. Bucket C, organisational efficiency: headcount-to-revenue and automation. Buckets B and C are the primary suspects. A. Direct delivery costs (COGS) Support ticket rate per customer, a product-quality issue? Infrastructure cost per transaction B. Expansion overhead (primary suspect) Compliance and regulatory reporting, 60% recurring New-market setup costs not yet amortised No unit-economics model for new markets C. Organisational efficiency (primary suspect) Headcount grew 45%, faster than revenue Support automation, self-serve deflection rate Process duplication across markets Analysis and root cause. Bucket A, direct costs: the 40% higher ticket rate in new markets is a product issue, not just volume. New-market users hit more friction, language, document types, local payment methods, and escalate to support. That is a solvable product problem, not a permanent structural cost.. Bucket B, expansion overhead: compliance is 60% recurring and local regulatory reporting is non-negotiable. But with no payback model, some of these 5 markets may never reach cost-positive at current scale. This is the most important strategic finding, cost was incurred without validating revenue potential.. Bucket C, organisational efficiency: 45% headcount growth has outpaced revenue, much of it support roles concentrated in new-market operations. A significant share of support volume is deflectable through automation, established markets likely run on lower support-to-customer ratios. Solutions, feasible and creative. Immediate, 0 to 4 weeks, low cost: support deflection via self-serve. Build localised FAQs and in-app guided flows for the top 10 ticket types in new markets, targeting 30% deflection. Each percentage point of deflection cuts the support headcount requirement by about 0.5 FTE at scale.. Short-term, 1 to 3 months, zero cost, analytical: new-market unit-economics audit. Build a payback-period model for each of the 5 markets, identifying which are on a path to unit-economics-positive and which are not. For markets with no viable path, consider pausing active investment while maintaining presence.. Long-term, 3 to 6 months, medium cost: shared services plus compliance automation. Centralise compliance reporting across new markets into a shared-services function and automate report generation where rules allow. Eliminates duplicated compliance headcount across 5 markets running parallel teams.. One creative process fix: a unit-economics gate on future expansion. Before any new-market launch, require a payback-period model and a cost-per-customer projection against established-market benchmarks. It prevents the problem recurring and costs nothing, it is a process change, not a product change. Measure success. Measured in layers: Leading indicator, week 4: support-ticket deflection rate. 25 to 30% deflection in new markets within 4 weeks of the self-serve launch. Primary KPI, month 3: cost-to-revenue ratio. 82% down to below 72% in 90 days, reversing the trend and buying runway. Long-term health, month 6: cost per customer, new versus established. New-market ratio falls from 2.3 times to below 1.5 times established markets. Guardrail, ongoing: customer-satisfaction score. Must not decline, cost reduction should not come at the expense of service quality. Bottom line. Nubank's cost problem is not a spending problem, it is an expansion-without-unit-economics problem. Five markets were entered without validating whether they would ever be cost-positive. The immediate fix is support deflection, cheap, fast, and aimed at the highest-cost symptom. The strategic fix is a market-by-market audit plus a permanent unit-economics gate on future expansion. Done right, this reverses the cost-to-revenue trend within 90 days and prevents the same mistake in the next 5 markets.

Root cause	Evidence	Priority
Product friction in new markets driving a high support rate	40% higher ticket rate, 65% from new markets	Highest
Expansion without a unit-economics model	2.3 times cost per customer, no payback period, some markets may be structurally unprofitable	Highest
Low support automation, manual handling of deflectable tickets	Support costs doubled, automation opportunity exists	Medium