Phoebe Gates and Sophia Kianni’s AI shopping startup Phia is facing scrutiny over allegations that its browser extension attributed online purchases to the company even when Phia had not actually referred the shopper.
The controversy centres on tracking cookies, small pieces of data commonly used in affiliate marketing to identify which service referred a customer to a retailer. Phia earns commissions when shoppers it refers go on to make purchases. The allegation is that some of its features could place Phia’s tracking information during checkout without the company having meaningfully generated the sale.
The financial impact after those features were switched off has become one of the most striking details in the dispute. Phia’s average daily revenue fell from about $80,000 to between $10,000 and $28,000, according to an internal revenue chart reported by Bloomberg.
Phia has disputed the suggestion that the decline can be attributed entirely to the removal of those features. A company spokesperson said the startup had also disabled most of its monetisation efforts during the period.
How Phia’s shopping extension generated commissions
Phia is an AI-powered shopping assistant co-founded by Gates and Kianni, who were roommates at Stanford. Its browser extension is designed to compare prices, surface resale options and suggest alternatives while people shop online.
Affiliate commissions form part of its business model. In a typical transaction, a retailer can pay a shopping platform if that platform referred the customer who ultimately completed the purchase. Cookies or referral codes help determine which service receives credit.
Affiliate-marketing researcher Ben Edelman and Capital One Shopping, said Phia’s extension could open background tabs and insert referral information during the checkout process.
One feature described as a “passive trigger” could reportedly place Phia cookies on websites that a user had previously visited. Another could insert a cookie when a shopper clicked on a checkout page carrying a Phia pop-up. In some instances, simply closing the pop-up could reportedly activate the process.
That matters because attribution determines who gets paid. If Phia received credit for a purchase it had not actually generated, the company could collect an affiliate commission that otherwise would not have been assigned to it.
Internal messages put the timeline under scrutiny
Phia initially said that incorrect attribution had resulted from a recent change to its code.
After Bloomberg contacted the company in July, a spokesperson said Phia had learned within the previous 24 hours that a recent release was causing incorrect attribution for some users. The company said its team identified the problem, mitigated it and resolved it.
Later reporting complicated that explanation.
Internal communications reported by Bloomberg indicated that Gates and Kianni had discussed automatic cookie placement months earlier. A December 18 message attributed to Gates said she was “worried this is an issue across the board” and asked whether automatic cookie drops were operating across sites with coupons so Phia could “confirm we are monetising on all [gross merchandise value].”
Bloomberg’s subsequent reporting said the founders were aware for at least seven months of features that could insert tracking cookies into the checkout process.
Those reports do not establish criminal wrongdoing, and no one has been charged with a crime in connection with the matter.
Revenue fell sharply after the features were removed
The numbers reported from inside Phia have added another layer to the controversy.
After the disputed features were disabled, Phia’s average daily revenue dropped from roughly $80,000 to a range of $10,000 to $28,000, according to an internal chart reviewed by Bloomberg.
A separate internal Slack message from a Phia data scientist reportedly indicated that the features accounted for more than half of the merchandise value the company said it generated in June.
Phia challenged that interpretation. Its spokesperson said the revenue decline was partly explained by the company turning off most of its monetisation activity at the time and said the information attributed to the data scientist was incorrect.
The distinction is important. The reported revenue figures show what happened after the features were disabled, but Phia disputes claims about how much of the fall should be linked directly to the disputed attribution practices.
Impact.com suspended Phia and payments are being reversed
The fallout has extended to Phia’s affiliate infrastructure.
Impact.com, which handles commissions, told Bloomberg that it suspended Phia from its platform following the earlier report. Money that had been scheduled for payment to Phia was also being reallocated, according to the reporting.
Phia says the features responsible for misattributions were removed on July 7.
The company has also begun reviewing transactions and reversing payments linked to incorrectly attributed purchases.
“We are reviewing every transaction, we are fully committed to and have already begun issuing all transaction reversals to brand partners as a result of any misattribution, and we are hiring a head of compliance to make sure something like this never happens again,” a Phia spokesperson said.
A fast-growing startup faces a test over affiliate attribution
The controversy arrives as Phia has attracted substantial investor attention. The company has raised more than $40 million, with investors named in the reporting including Kleiner Perkins, Hailey Bieber, Kris Jenner, Kim Kardashian, Karlie Kloss and Sydney Sweeney.
What began publicly as an issue Phia described as incorrect attribution caused by a recent code release has consequently widened into questions about how long the underlying features had existed, how they affected affiliate commissions and how much of the startup’s reported commerce activity was connected to them.
For Phia, the next phase is now centred on reviewing transactions, returning commissions linked to misattribution and strengthening compliance.
The clearest measurable change so far remains the one visible in its reported revenue: after the disputed cookie features were disabled, average daily revenue moved from about $80,000 to a range as low as $10,000.
Also Read: Butterfly Learnings Raises Rs 65 Cr in Pre-Series B Led by Inflexor Ventures
