Fraud detection software does not have a single sticker price. What you pay depends on how many transactions you process, whether you want a plug-and-play tool or a custom model, and how much risk you are willing to carry yourself. Most retail store owners are surprised to find that entry-level tools cost less than a monthly Shopify subscription, while enterprise-grade systems can run as much as a part-time employee.
This article breaks down exactly how pricing works, what the cost tiers look like for small and mid-size merchants, and whether there is a cheaper path if your fraud rate is already low.
How do fraud detection tools for retail stores typically charge?
The pricing model matters as much as the dollar amount. Most tools use one of three structures, and choosing the wrong one for your volume can cost you more than the fraud itself.
The most common structure is a percentage of transaction value, typically 0.1% to 0.5% of each order flagged or reviewed. A store doing $50,000 a month in sales would pay $50 to $250 per month under this model, regardless of how many transactions the system actually catches. Riskified and Signifyd both use this approach, and they also absorb the chargeback cost if they approve an order that turns out to be fraud. That guarantee is baked into the price.
The second structure is a flat monthly fee with a transaction cap. Tools like Kount and Fraud.net charge $150 to $500 per month for up to a set number of orders, typically 500 to 5,000. Once you exceed the cap, you pay per transaction on top. This model is predictable for stores with stable monthly volume.
The third structure is pay-per-decision, where you pay a small fee each time the system scores a transaction, usually $0.02 to $0.10 per check. This is common with API-based tools that integrate directly into your checkout flow. For a store processing 2,000 orders per month, that works out to $40 to $200 per month. The cost scales exactly with volume, which is useful when your sales spike seasonally.
A Western fraud consultancy building a custom detection model charges $5,000 to $15,000 upfront, plus $2,000 to $5,000 per month for ongoing tuning and monitoring. That overhead is designed for enterprise retailers processing millions of dollars per month. For a store doing $500,000 or less annually, it rarely makes financial sense.
What are the main cost tiers for small and mid-size merchants?
Think of the market in three bands, separated by annual revenue and the sophistication of fraud risk you face.
Stores doing under $500,000 a year in online sales typically see fraud rates of 0.1% to 0.4% of revenue. At that scale, a plug-and-play tool at $50 to $150 per month covers most of what you need. Shopify Fraud Protect, for example, costs nothing extra if you are already on Shopify Payments, and it covers chargebacks on orders it marks as protected. For stores on other platforms, Stripe Radar's basic tier is included in Stripe's standard processing fee, which makes the marginal cost of AI fraud screening effectively zero for the first layer of protection.
| Store Size | Annual Revenue | Typical Fraud Rate | Recommended Tool Type | Monthly Cost |
|---|---|---|---|---|
| Small | Under $500K | 0.1%–0.4% | Built-in platform tool or entry-level SaaS | $0–$150 |
| Mid-size | $500K–$5M | 0.3%–0.8% | Dedicated fraud platform | $150–$500 |
| Growing | $5M–$20M | 0.5%–1.2% | Dedicated platform + custom rules | $400–$800/mo + setup |
| Enterprise | $20M+ | Varies | Custom model or managed service | $2,000+/mo |
Mid-size stores in the $500,000 to $5 million range often run into the limits of platform-native tools. The built-in filters catch obvious fraud but miss the more specific patterns tied to a particular product category or customer geography. At this level, dedicated tools like Kount, Sift, or Fraud.net start making sense. Budget $150 to $500 per month for a solid off-the-shelf solution that includes manual review queues, device fingerprinting, and velocity checks.
Stores between $5 million and $20 million in annual revenue often have a fraud problem specific enough that generic models start missing things. A sporting goods retailer loses inventory to account takeover fraud. A beauty brand sees organized refund abuse. At this point, a custom rules layer on top of a commercial platform adds real value. Western agencies charge $5,000 to $15,000 to build that layer. A global engineering team like Timespade builds the same custom model for $3,000 to $6,000 with no ongoing markup once it is deployed.
How does transaction volume affect the monthly bill?
Volume is the primary cost lever for most pricing models, but not always in the direction you expect.
For percentage-based tools, higher revenue means a higher bill even if your fraud rate stays constant. A store growing from $50,000 to $200,000 in monthly sales will see its Riskified or Signifyd invoice roughly quadruple, from about $100 to about $400 per month, assuming no change in the fraud rate. The tool is doing proportionally more work, but the unit economics for the merchant get worse as the business scales. That is the tradeoff for the chargeback guarantee.
For flat-fee and per-decision tools, the picture is different. Fraud.net's $299/month plan covers up to 5,000 transactions. If you are processing 1,000 orders a month, you are paying for capacity you do not need. If you grow to 4,500 orders, the per-order cost drops from $0.30 to $0.07. The model rewards volume growth.
According to Juniper Research's 2021 data, global ecommerce fraud losses were already approaching $20 billion annually, with small and mid-size merchants absorbing a disproportionate share because they lack dedicated fraud teams. The same research found that merchants using dedicated AI fraud tools reduced chargebacks by 35% to 65% compared to manual review alone.
The practical implication: if you are processing more than 500 transactions per month, a per-decision API is almost always cheaper than a percentage-based platform. Below that threshold, the percentage model makes sense because the minimum monthly fee on a flat-rate tool often exceeds what you would pay on a percentage basis.
| Monthly Transactions | Best Pricing Model | Estimated Monthly Cost |
|---|---|---|
| Under 200 | Platform-native (free or % of sales) | $0–$80 |
| 200–1,000 | Per-decision API | $20–$100 |
| 1,000–5,000 | Flat monthly fee | $150–$400 |
| 5,000–20,000 | Flat fee + custom rules | $300–$800 |
| 20,000+ | Custom model or enterprise contract | $1,000+/mo |
Are there affordable entry points for stores with low fraud rates?
Yes, and most merchants do not take advantage of them.
If your chargeback rate is below 0.5% of transactions and you are not in a high-risk product category (electronics, gaming, luxury goods, gift cards), you probably do not need a dedicated fraud platform at all. Stripe Radar and Shopify Fraud Protect together cover the vast majority of common attack patterns at no incremental cost beyond what you already pay to process payments.
The real cost of fraud for low-risk stores is not the tool. It is the false positives. A 2021 Javelin Strategy study found that US merchants declined $331 billion in legitimate orders due to fraud fears, more than thirteen times the value of actual fraud losses that year. A misconfigured rule that blocks any order from a new IP address might stop one fraudulent transaction and reject twenty real customers. That is the hidden cost that most fraud tool comparisons ignore.
For stores with low fraud rates, the right starting point is this: turn on whatever your payment processor includes, set conservative velocity rules (flag any customer placing more than three orders in an hour, for example), and review declined orders manually once a week. That workflow costs nothing and catches most fraud at low volumes.
When you start seeing specific patterns, such as a cluster of refund requests from the same state, or gift card purchases followed immediately by disputes, that is when a dedicated tool earns its fee. At that point, Sift's Fraud Prevention suite starts at $500 per month with a 90-day free trial for new merchants, and Kount offers a free tier for stores processing under 500 transactions monthly.
If the fraud pattern is specific enough that generic tools keep missing it, a custom detection model makes sense. Timespade builds fraud detection systems as part of its Predictive AI practice. A custom model trained on your own transaction history, tuned to your specific product category and customer geography, costs $3,000 to $6,000 to build. Western agencies charge three to four times that for the same scope. Once deployed, there is no per-transaction fee on the model itself, only the underlying infrastructure, which runs about $50 to $150 per month for most mid-size stores.
The right fraud detection spend is whatever covers your actual risk without generating enough false positives to outweigh the savings. For most stores under $2 million in annual revenue, that number is between $0 and $200 per month. For stores in the $2 million to $10 million range with measurable fraud exposure, $300 to $600 per month is a reasonable budget before considering custom solutions.
