Want to know if this job is worth applying to?
Алматы, Қазақстан
Onsite
Higgsfield AI is the fastest-scaling generative AI company in history, hitting $1B in annual revenue run rate, 30M+ users worldwide, 6M+ generations per day, and powering 390 of Fortune 500 brands.
We're building at the absolute frontier of AI-powered video creation and next-generation creative tools. Joining Higgsfield means becoming part of a high-impact team shaping the future of AI-native experiences, at a company that isn't just moving fast, but rewriting what fast looks like.
We sell subscriptions and credit packs to consumers in over a hundred countries, and today almost all of it moves as cross-border card payments in USD. That has three consequences, and this role owns all three.
We are declined more than we should be. In some of our largest growth markets roughly one attempt in three fails, and a meaningful share of those are not "no money" — they are cross-border blocks, card limits, stale credentials on renewal, and issuers who do not recognise us.
We pay more than we should. Our blended cost of acceptance is the single largest line below compute, and the majority of the network component sits on transactions our processor classes as cross-border.
We are not on the rails our customers use. In two of our fastest-growing markets, the dominant local payment method is one we do not offer at all.
You own the measurement and the argument behind fixing all of it.
You make sure:
We know our true authorization rate — at the unit the business actually cares about, not the one that is easy to compute
Every basis point of payment cost has a name, a geography and an owner
A decline is a diagnosis, not a statistic
When we propose a new rail, a new entity or a new processor arrangement, the number underneath it holds up to Stripe's own analysis and to our board's
Authorization & decline analysis
Own authorization rate end to end: by country, issuer, BIN, card brand and funding type, plan price point, first payment vs renewal, and payment method.
Build and maintain the decline taxonomy — what each decline code actually means, which are recoverable, which are terminal, and which are us rather than them.
Measure at the right unit. An invoice that succeeds on the third retry is a success, not two failures and a success, and reporting it the other way has made teams optimise the wrong thing.
Separate soft declines we can recover from hard declines we cannot, and put a number on the recoverable pool.
Cost of acceptance
Own the blended take rate and its decomposition: interchange, scheme fees, processor markup, cross-border and currency-conversion components, disputes, and the fixed per-transaction pieces.
Find where the cost actually sits. Averages hide it — the cost is concentrated by corridor, by card type and by whether the transaction is domestic.
Build the evidence for commercial conversations with our processor, and keep it current enough to reuse.
Local payment methods, entities and routing
Size the case for new rails market by market: what share of local commerce runs on them, what our current card performance costs us, and what we would realistically recover.
Model merchant-of-record and local-acquiring arrangements against the alternative of our own entities, honestly — including the fee we would pay for the privilege.
Design and read the experiments that prove it, with the product analytics team. A new payment method is an A/B test, not a launch.
Renewals, dunning and involuntary churn
Own the retry and dunning strategy as a measurable system: when to retry, how often, through which rail, and when to stop.
Quantify involuntary churn separately from voluntary. They look identical in a cancellation count and they need completely different fixes.
Track the health of stored credentials — network tokens, card-account-updater coverage, expiring cards — because in a subscription business that is silent revenue leakage.
Disputes, refunds and reconciliation
Own dispute and chargeback rates against network thresholds, with the lag handled properly — disputes arrive weeks after the charge and a same-period denominator understates you.
Reconcile processor data to the revenue ledger and explain the differences rather than plugging them.
Work with the antifraud team where disputes are abuse rather than genuine dissatisfaction.
SQL and Python on raw processor and billing data in BigQuery. We use AI heavily — resourcefulness beats syntax.
Partner with Finance (the numbers get booked), Product (checkout and paywall), Engineering (the billing integration) and our processor's team directly.
2+ years in payments analytics, payment operations, billing, or risk at a merchant, PSP, acquirer, or fintech. This is a domain role — general analytics experience without payments does not transfer quickly enough.
Fluency in the actual mechanics: authorization vs capture vs settlement, decline codes, 3DS and SCA, network tokens and account updater, interchange and scheme fees, chargeback lifecycle, MCC, BIN, cross-border and DCC.
Strong SQL, without help. Most of this job is joining sessions to intents to charges to invoices to refunds across imperfect keys and getting the denominator right.
Python at working level for analysis.
The instinct to check the unit of measurement before quoting a rate — per attempt, per invoice, per user, per session are four different numbers and only one of them answers the question.
Comfort with subscription billing specifically: renewals, proration, plan changes, failed payments, dunning, involuntary churn.
Commercial judgment: you can build the case for a rail or a rate and defend it against a vendor who does this for a living.
Clear written and spoken English, B2+.
Backgrounds that often do well:
Payment operations or payments analytics at a consumer subscription, marketplace or gaming company
Analysts from PSPs, acquirers, orchestration platforms or card networks
Billing / revenue operations analysts who owned dunning and involuntary churn
Risk or chargeback analysts who moved toward the acceptance side
This role is not a fit if you:
Want to own checkout UX or pricing — that is Product Analyst, and it is open
Want to build fraud models — that is the Antifraud data scientist, and it is open
Would report an authorization rate without saying what the denominator was
Would take a processor's own analysis of their own performance at face value
Need someone else to prepare the data
Want predictable 9–5 workdays
First interview (30 min)
Business case (60 min)
Take-home with real-shaped data (7 days, ~10–12 hours of work)
Team interview (60 min)
Paid on-site trial (1 month)
Competitive salary in USD
Equity: participation in the company's stock option program
On-site role in Almaty, Kazakhstan
Relocation support (flight + temporary housing)
Flat structure, high autonomy, fast career growth