Chase, Amex And Citi Pay You To Shop—But A Discount Doesn’t Mean The Merchant Made More Money

Citi launched a business on September 23 built around knowing where you spend money and selling merchants the opportunity to influence where you spend it next. Their new Citi Commerce Media uses 6.5 billion annual transactions across more than 700 spending categories. Citi says they can identify emerging purchase intentions and put relevant offers and ads in front of customers. Their planned acquisition of rewards platform Kard is supposed to help. You may be familiar with merchant-funded offers. Sometimes that’s pitched as a card benefit, like a monthly Uber or Lyft credit. The brand wants to get in front of a valuable group of customers. That’s great for the brand for awhile. They acquire customers through discounting. Some of the customers stay with them. Eventually the brand may decide they’ve acquired all the customers they ever will, there’s no reason to keep giving discounts to the ones who’ll stick with them. So they don’t renew their contract (or the card issuer needs to fund more of the offer). This is one reason that cards using merchant offers to justify annual fees need a pipeline of fresh offers. But more targeted offers are supposed to move the needle with incremental purchases. Someone who buys Chipotle might be convinced by an offer to try Qdoba. Qdoba will pay a lot to reach that customer, and pays the card issuer and funds a discount or rebate. That’s Amex Offers, Citi Offers, Chase Offers, etc. Getting money back on purchases you’ll make anyway is great. That’s not what the merchant is after. They want incremental sales, and the best customers to reach are ones who haven’t decided whom to buy from yet or who are already buying from a competitor. So layered on top of targeting based on transaction levels are companies that sell ‘customer profiles’ that are supposed to divine purchase intent. But a new study using 96 million credit card transactions offers a reason to be skeptical of how much additional value comes from these profiles. Your Bank Wants To Become An Advertising Business Amex Offers, Chase Offers, and Citi Merchant Offers and similar card-linked offers are parts of businesses selling merchants access to customers: Chase Media Solutions uses purchase history to help merchants reach new customers, bring back lapsed ones, and encourage existing customers to spend more. American Express’s Amex Advance uses machine learning to predict when customers will be in the market for products and services, including targeting through Amex Offers. Cardlytics supplies purchase-based advertising through banking channels. A merchant can seek customers who spend in its category, including at competitors. Citi’s new platform adds advertising across its website, app, and paid media, alongside its broader offers business. The bank sees a purchase and can connect an offer with a transaction which is far better than counting clicks off of ads. But that doesn’t mean the offer caused the purchase. The Customer, Bank And Merchant All Have Different Interests Hotels may offer a $20 rebate on a $200 stay. If you were already going to stay there, using the offer saves you $20. You might move the charge from your primary card to the one displaying the offer. The card issuer gains spending and interchange revenue (and, they hope, APR). The hotel may discounting a booking it was going to receive anyway. I’m probably not changing where I stay over $20 off $200. However, You can imagine switching hotels for a bigger offer, or that some consumers do switch because of this one. The offer wins business for the property and takes it from someone else. You book direct instead of through an online travel agency, so that the merchant of record triggers the offer. The hotel saves on distribution costs even without selling an incremental room. You take an additional trip or add a night to hit the minimum amount that triggers the offer. That’s generated new demand. You move a stay forward to meet the offer deadline. That month’s numbers look better, but it could trade off with revenue they’d get later. The bank wants its card at the front of your wallet, more spending, and revenue from the advertising relationship. The merchant wants additional profit after paying for the discount and campaign. Those goals overlap, but they aren’t the same. As a customer, you care about getting a better deal. As I wrote recently, tools like Cardpointers help you maximize the value you get out of these offers without it changing your purchase patterns – for instance opting into all the available offers without even realizing what’s going to get you a benefit, so that you’re pickup up the cash and not changing your behavior even subtly (it can also help you really chase maximum value). More Detailed Profiles Though Don’t Seem To Help? A new Korean credit-card study examined 13 quarters of transactions, with 229,586 customers. Researchers measured how customers moved between spending categories and how their shopping patterns changed over time. The test was whether those measures improved predictions beyond information already available from spending history. And it only seemed to get an additional 0.0009—0.09 percentage points better. That’s data from a single bank card product, it doesn’t know that predictive models fail broadly. But it suggests that adding elaborate measures of changing customer “personas” don’t get much beyond an existing model that builds off actual transactions. Basic offer targeting based on recent purchases works, the question is whether profiling works better. There Is Evidence That Offers Change Behavior, But Also That Advertisers Overpay An eBay experiment found that ads on searches for the company’s own brand produced no measurable short-term benefit. Customers were generally headed there anyway. Ads on other search terms worked better for new and infrequent customers, while frequent customers accounted for much of the spending without the same benefit. On the other hand, a 2019 airline ancillary pricing experiment showed that using customer context increased revenue per offer by about 10% over manually set rules. Random discounts increased purchases but reduced revenue. Choosing the offer intelligently helped. Cardlytics explicitly offers test-versus-control measurement which is directionally needed. You want comparable customers randomly assigned to receive an offer and not, with results measured after discount costs and over enough time to catch purchases merely shifted forward. Then compare the expensive profiling against ordinary recent-spending targeting. Topics on this page