Fraud insights
Building an effective fraud fighting strategy requires understanding the specific drivers of fraud for your business. If you use Radar for Fraud Teams, you can access the Insights tab of the Radar page in your Dashboard to:
- Visualize the ratio of fraudulent and legitimate transactions across your payments.
- Identify combinations of Radar attributes that have material impact on your fraud rates.
- Adjust your Radar rules to effectively balance fraud prevention and legitimate customer conversion.
Configure your data set
You can specify the time period analyzed and what types of payment activity constitute fraud to further customize your results.
Specify the time period
By default, we display statistics in near real-time for the prior 30 days of transaction history. To see data for a different time period:
- Click the Date filter to open the time period editor.
- Use the dropdown to choose a relative comparator.
- Depending on the comparator you choose, set the parameters, such as in the last 1 months or between 2/26/2024 and 3/18/2024.
- Choose your local time zone or Greenwich Mean Time (UTC).
- Click Apply.
Configure your fraud definition
Click Configure to choose which types of transactions to include as fraudulent in your Insights statistics.
- All fraudulent transactions: Payments disputed for fraud, reported as early fraud warning (EFW), or refunded as fraud
- Only disputes: Any disputed payment, regardless of category
- Only fraudulent disputes: Disputed payments in the fraud category
- Only early fraud warnings: Issuer-flagged suspicious payment EFWs
Evaluate your fraud markers
Stripe analyzes all the payments for the specified time period, then presents:
- A summary of the total fraudulent and legitimate payments for the time period.
- A table of the top rule attribute values that suggest a correlation with fraud, based on the ratio of fraudulent to legitimate payments.
The summary and each attribute in the table provide the following statistics:
Statistic | Description |
---|---|
Fraud percentage | The percentage of fraudulent payments where this rule attribute was present. |
Legitimate percentage | The percentage of legitimate payments where this rule attribute was present. |
Fraud volume | The total amount of the fraudulent payments where this rule attribute was present. |
Legitimate volume | The total amount of the legitimate payments where this rule attribute was present. |
Fraud count | The total number of fraudulent payments where this rule attribute was present. |
Legitimate count | The total number of legitimate payments where this rule attribute was present. |
Personalized fraud indicator results
Use filters to discover high risk attribute combinations
You can add any of the attribute values presented as your top fraud indicators as a filter. This adjusts the table to show a new set of top attribute values that corresponded to fraud in combination with the filtered rule attribute value. Continue applying filters in this way to find the combination of rule attribute values where it’s advantageous to block those transactions.
For example, let’s say your top indicator shows that 19% of fraudulent payments included Delaware as the billing state. Blocking all payments from Delaware isn’t sensible, so filter this attribute value to see which other rule attributes corresponded most to fraud when the billing state is Delaware. You find that 42% of fraudulent payments where Delaware is the billing state also have a shipping state that isn’t Delaware. It’s still fewer than half the fraudulent payments and too aggressive to block, so you apply it as another filter. The adjustment reveals that within these filters, payments where the amount charged is greater than 500 USD corresponds to 75% of fraudulent payments.
As the example illustrates, you can filter successively to discover a set of rule attribute values that represent a material percentage of fraudulent payments when all exist in the same payment. When this combination also reflects a low incidence of legitimate payments, it might warrant creating a rule to block that combination of attribute values.
Customize filters
You can also create a filter without using the rule attributes presented in the table.
- Click the More filters button.
- Choose the rule attribute for which you want to create a filter.
- Depending on the attribute you choose, set the parameters, such as Risk score is greater than 15 or Card bin is 4242.
- Click Apply.
Create a rule
When you assemble the set of filters that represent your optimum ratio between blocking highly risky transactions without compromising legitimate payments, you can automatically create a rule to prevent payments where all the selected attributes exist simultaneously.
- Click Add block rule to slide open the rule editor.
- Check that the rule accurately reflects the attributes you filtered.
- (Optional) Augment the rule to include other attributes or your own custom metadata, such as product codes or retail locations. Try Radar Assistant to generate a rule based on your natural language prompts.
- Click Test rule.
- If necessary, correct any validation errors and retest.
- On the Review new rule page, review how this rule performs against your recent transactions to confirm whether you want to enable it.
- Click Add rule to begin applying this rule to all future transactions.
Inspect charts
Stripe provides visualizations of the attributes identified as your top drivers of fraud. Each chart shows:
- The vertical axis measures the percentage of successful payments within your specified time period.
- The horizontal axis shows possible values for the rule attribute.
The chart graphs indicate the prevalence of the rule attribute in either fraudulent or legitimate payments. This allows you to evaluate how significantly the rule attribute appears in fraudulent and legitimate payments, respectively.
Note
Graphs represent the percentage of payments by volume, not the percentage of total payments. For example, if the Billing state graph shows 6% fraudulent payment volume for Utah, it indicates that of all fraudulent payments, 6% of them had Utah as the billing state. It doesn’t indicate that 6% of all payments where Utah is the billing state are fraudulent.
Hover over any point in the chart to see additional metrics for both the fraudulent and legitimate payments graphs.
Hovering displays additional volume and count metrics.
Metric | Description |
---|---|
Percentage | Percent of payments by volume at the selected point |
Volume | Total amount charged for payments at the selected point |
Count | Number of payments at the selected point |
Change chart attributes
The charts displayed reflect the Radar rule attributes corresponding to your top fraud drivers. To generate visualizations for other attributes:
- Click Select attributes.
- Scroll through the list or enter key words in the search bar to find attributes.
- Click an attribute’s card to select or deselect it. Selected attributes display a checkmark and the button displays the total number of attributes selected.
- Click Show x attributes to generate the charts for your selected attributes.
The Radar rule attribute modal allows you to choose from more than 150 attributes.