Black Friday Chargebacks: AI Fraud Screening in Burnaby

Black Friday is the week a small online shop sees its best sales and, a few weeks later, its worst surprises. Chargebacks don't land on the day. They trickle in through December and January, after the stock has shipped and the money's been spent.

If you sell online from Burnaby or anywhere in Metro Vancouver, the good news is that most of the AI fraud screening you need is probably already built into your payment processor. The catch is that it only covers part of the problem. Here's what the official documentation says it does, what it costs as of October 2026, and where you still need a person.

Why chargebacks hurt more than the lost sale

A chargeback isn't a refund. When a cardholder disputes a payment, Stripe "debits the disputed amount, plus a dispute fee" from your account, and the outcome sits "at the sole discretion of the account owner's bank," according to Stripe's guide to how disputes work.

The timeline is long. The same page says card networks typically let cardholders dispute within 120 days of the payment, you usually get 7 to 21 days to respond once a chargeback lands, and the whole cycle can run two to three months.

Then there are the fees. As of the June 17, 2025 update on Stripe's dispute pricing page, a Canadian business pays CA$15 when a dispute is received and another CA$15 if it chooses to counter (checked October 2026). Stripe returns the countered fee if you win. It says it never returns the received fee unless your contract states otherwise.

For wider context, the Competition Bureau reported that Canadians lost over $704 million to fraud in 2025 and that only 5% to 10% of frauds are reported. Those numbers are mostly consumer scams, not merchant chargebacks, so read them as background rather than a measure of your own risk.

What AI fraud screening actually does

Two platforms cover most small shops around here.

Shopify. Shopify's fraud analysis rates online credit card orders as low, medium or high risk of a chargeback due to fraud. Shopify says the recommendations "are powered by machine learning algorithms that are trained on historical transactions across all Shopify stores." Indicators include AVS and CVV checks, IP and location details, device and network patterns, and whether the buyer tried more than one card. One line matters more than the rest: "Shopify does not cover charge reversals from banks." Ship a high risk order and the loss is yours.

Stripe Radar. Stripe says Radar evaluates transactions in real time "using AI algorithms to assess the risk of fraud." As of October 2026 it comes in four plans: Lite, Standard, Plus and Pro. Stripe's plan comparison claims businesses block 42% more fraud on Standard than on Lite, on average. That's a vendor figure you can't check independently. Custom rules, manual reviews, numeric risk scores and adaptive 3D Secure need Plus or above, and the paid plans charge a fee for each screened transaction.

On Stripe's Canadian Radar pricing page, checked in October 2026, Lite is included with standard payments pricing, while Standard, Plus and Pro start at CA$14, CA$20 and CA$26 per month. The per transaction rate wasn't listed on that page, so confirm it in your dashboard before you upgrade.

It's worth knowing how Stripe itself frames the output. Its risk settings documentation says scores run from 0 to 99, that a Normal level "doesn't guarantee that the payment is legitimate," and that Elevated "doesn't mean the payment is fraudulent." It's a probability, not a verdict.

The pattern to watch: card testing

Card testing is when someone runs stolen card numbers through your checkout to see which ones still work. Stripe's card testing guide lists the symptoms: a spike in failed or blocked payments, a spike in 402 errors, and "suspicious payments with low transaction amounts, often with nonsensical customer names and emails."

The damage goes beyond the few charges that succeed. Stripe warns that a high decline rate can raise declines on legitimate payments "even after card testing ceases," adds fees, and can land you in card monitoring programs. It also says "simple firewall rules or filters based on a single heuristic such as IP addresses are usually not sufficient" on their own.

A practical setup for the next five weeks

• Turn on what you already have. Confirm Shopify fraud analysis or Radar is active. Stripe says "the more data your integration provides, the more successful card testing prevention can be," so a recommended hosted checkout that passes IP address, email, name and billing address beats a bare custom form.
• Hold medium and high risk orders before fulfilment. Shopify's own suggestion for medium risk is to "Confirm with the customer before fulfilling." Don't ship first and ask later.
• Use 3D Secure on risky payments, not every payment. Liability for fraud disputes on authenticated payments typically shifts to the card issuer, but Stripe's 3D Secure documentation says plainly that it "isn't guaranteed." It also doesn't cover other dispute reasons, like product not received.
• Watch early fraud warnings. Stripe says about 40% of Visa and Mastercard early fraud warnings become fraud disputes if you don't refund. Its analysis suggests refunding when the charge is roughly at or below your dispute fee, and calls it "likely not worthwhile" on charges more than 35 percent above that fee. With a CA$15 fee, that's about CA$20. That's our arithmetic, not a Stripe number.
• Answer every inquiry. Stripe warns that ignoring one can escalate to a "likely unwinnable" chargeback, and on 3D Secure payments a no reply chargeback can invalidate the liability shift.
• Add rate limits and CAPTCHA if you run a custom checkout. Stripe gives the example of limiting how many new customers a single IP address can create in one day.

Where an AI employee fits

The scoring model belongs to your processor. You can't retrain it, and building your own makes no sense for a shop with a few thousand orders. What a small team usually lacks is someone watching the queue at 11 p.m. on Black Friday weekend.

That's the gap an AI assistant can fill. It can watch for early fraud warnings and new disputes, put held orders in front of a human with the indicators summarized, email medium risk buyers to confirm a shipping address, and pull tracking numbers, delivery confirmations and order notes into a draft evidence packet before the response window closes.

A person still approves every refund, cancellation and dispute response. Language models can get facts wrong, so anything going to a bank gets checked line by line.

Here's an illustrative example, not a client result. A shop that receives 10 fraud disputes after Black Friday pays CA$150 in received fees alone at Stripe's current Canadian rate. That's before the lost goods, the shipping, and another CA$15 for each dispute it decides to fight.

Where this doesn't apply

In store and certain order types. Shopify gives no fraud recommendation for POS orders, gift card purchases, free orders, subscription renewals or B2B orders. If most of your sales are tap payments at a counter in New Westminster, online fraud screening isn't your main lever.

Friendly fraud and service disputes. Fraud scores and 3D Secure target stolen cards. Disputes like product not received follow the standard process, and you win those with good records and customer service, not a model.

Overblocking. Every threshold trades fraud for revenue. Stripe's risk settings range from "Maximize protection" to "Maximize revenue," and tightening hard during your busiest week can turn away real buyers. Backtest before you change anything.

You're already fine. If you use a hosted checkout with the defaults on and disputes are rare, an extra automation layer may cost more than it saves. Don't automate a problem you don't have.

Privacy. Screening means collecting IP, device and address data. BC's privacy law applies "to all organizations and to all personal information held by organizations unless PIPA says that it does not apply," and limits collection to reasonable purposes, judged by "what a reasonable person would think is appropriate," per the OIPC BC guide to PIPA. Tell customers why you collect it and keep only what you need.

Sources

• Stripe. "How disputes work." Stripe Documentation, accessed October 2026. docs.stripe.com/disputes/how-disputes-work
• Stripe. "June 2025 pricing updates for disputes." Stripe Support, 2025, accessed October 2026. support.stripe.com
• Stripe. "Radar." Stripe Documentation, accessed October 2026. docs.stripe.com/radar
• Stripe. "How Radar works." Stripe Documentation, accessed October 2026. docs.stripe.com/radar/how-radar-works
• Stripe. "Risk setting and risk controls." Stripe Documentation, accessed October 2026. docs.stripe.com/radar/risk-settings
• Stripe. "Radar pricing (Canada)." Stripe, accessed October 2026. stripe.com/en-ca/radar/pricing
• Stripe. "Protect yourself from card testing." Stripe Documentation, accessed October 2026. docs.stripe.com/disputes/prevention/card-testing
• Stripe. "Authenticate with 3D Secure." Stripe Documentation, accessed October 2026. docs.stripe.com/payments/3d-secure/authentication-flow
• Shopify. "Fraud analysis." Shopify Help Center, accessed October 2026. help.shopify.com
• Competition Bureau Canada. "Fraud Prevention Month to bring hidden crime into the spotlight." Government of Canada, March 2026. canada.ca
• Office of the Information and Privacy Commissioner for British Columbia. "A Guide to B.C.'s Personal Information Protection Act for Businesses and Organizations." October 2015. oipc.bc.ca

Want a second set of eyes on your checkout and dispute workflow before the rush? Book a free call with Autana. We're based in Burnaby and New Westminster, and we'll tell you straight if your processor's defaults are already enough.