Ecommerce teams often place fraud prevention outside the conversion discussion. Marketing brings shoppers to checkout, product improves the flow and a risk team blocks suspicious orders. That division creates a blind spot: an overly cautious fraud model can reject legitimate customers, while a permissive model can inflate revenue today and leak margin through chargebacks, promotion abuse and returns tomorrow.
A new case released by Riskified says Australian retailer Kogan achieved approval rates above 98% in higher-risk and higher-value transaction categories and identified $1.5 million in annual savings from better management of fraud and policy abuse. Because the figures come from a vendor case study, they should be treated as reported outcomes rather than an independent experiment. The operating lesson is still useful: the risk decision at checkout is a revenue decision.
Approval rate is not the opposite of fraud control
A low fraud rate can look excellent when the system simply rejects more uncertain orders. The hidden cost appears in false declines: legitimate buyers who are blocked, asked for excessive verification or quietly abandon after a payment failure. Their lost lifetime value rarely appears in the fraud team’s dashboard.
The better objective is profitable approval. That means approving as many legitimate orders as possible while keeping fraud, chargebacks and abuse below sustainable thresholds. It also means distinguishing a criminal transaction from a loyal customer using a promotion, a serial return abuser or a payment authorization failure. Those problems need different responses.
What the Kogan case changes in the measurement model
Kogan reportedly combined identity, device and behavioral signals to understand repeat policy abuse, promotion misuse and customer-level profitability. Automated dispute work also reduced manual effort. The important shift is from evaluating a single transaction to evaluating the person and pattern behind it.
For a marketing leader, this connects acquisition quality to checkout economics. A campaign may deliver a strong platform return while attracting a disproportionate share of bonus abuse, first-order fraud or low-value disputes. Conversely, a high-value audience may look weaker because conservative rules reject unfamiliar but legitimate buyers. Channel reporting needs post-transaction risk data to reveal either effect.
Build a risk-adjusted conversion scorecard
Track gross approval rate, authorization rate, false-decline estimates, chargeback rate, policy-abuse loss, manual review rate and decision latency. Add contribution margin after refunds, disputes and incentives. Segment the view by acquisition channel, new versus returning customer, order value, device, geography and promotion.
No single metric should dominate. Raising approval rate is not a victory if chargebacks exceed network thresholds. Lowering fraud loss is not a victory if good customers disappear. The scorecard should show the net value created by each risk decision and the customer friction required to achieve it.
A practical experiment for ecommerce teams
1. Find the uncertainty zones
Identify segments with unusually high decline rates, manual reviews or payment retries. High-value first-time buyers, cross-border orders and promotional peaks often reveal rules that are too broad. Compare them with confirmed fraud rather than assuming every decline was correct.
2. Separate fraud from policy abuse
Stolen payment credentials, repeated coupon use, wardrobing and subscription disputes are different behaviors. Blocking all of them at checkout can punish legitimate customers. Use limits, eligibility rules, return controls or post-purchase review where those responses are more proportionate.
3. Run a controlled approval test
Relax or replace one rule for a defined segment and monitor approved revenue, fraud, disputes and support contacts through the full settlement window. Do not judge after the first day: some losses emerge weeks later. Predefine the maximum acceptable downside.
4. Close the loop with acquisition
Send risk-adjusted value back to marketing analysis. If a campaign generates many approved but abusive orders, its real customer acquisition cost is higher than the ad platform reports. If a channel brings legitimate customers who are disproportionately declined, the checkout model is suppressing marketing performance.
Questions to ask a risk vendor
- How is a false decline estimated and audited?
- Which losses are covered, and which forms of policy abuse are excluded?
- Can decisions be explained at customer and rule level?
- How quickly do models adapt during promotions and seasonal peaks?
- Can performance be segmented by campaign, channel and customer value?
- What happens to approval quality if the vendor’s service is unavailable?
The CMO takeaway
Checkout risk is part of the promise made by marketing. A customer acquired at significant cost should not be rejected by an invisible rule that no growth team reviews. At the same time, a conversion that later becomes fraud or abuse is not growth.
Use the Kogan result as a prompt to put risk, payments, product and acquisition around one scorecard. The goal is neither maximum approval nor minimum fraud. It is maximum durable margin with the least necessary customer friction.
Sources
- Business Wire: Kogan increases approval rates and improves customer experience with Riskified
- Riskified: Ecommerce customer success library
