Comment by downandout | Hacker News Reader

Comment by downandout | Hacker News Reader

The tl;dr version is this:

Discover Card, which spends ~$15M/yr advertising with Amazon, wanted to give a $10 gift card to Kindle users that changed their default Amazon 1-Click purchase settings to use a Discover card. Instead, Amazon gave the gift cards to everyone that used Discover for a 1-click digital purchase, the vast majority of whom already had Discover as their default 1-Click purchase card. Discover's $500K budget was predictably drained in rapid fashion, and they barely got any of the actions they had agreed to pay for. The author of this letter was encouraged to hide this fact, pitch it as an overwhelming success of the campaign, and to ask Discover to expand the budget. He was fired after complaining about being uncomfortable with participating in obvious fraud against their 2nd largest advertiser, and is now suing Amazon.

The failures here occurred in every department. First, at a fundamental technical level, I don't understand how this could happen in the first place if it wasn't intentional. This was a simple CPA campaign. When someone changed their default card to Discover, they got a gift card. So it begins with their "ad execution team". Second, the moment the problem was discovered, they should have simply credited the campaign such that they were only charged for the actions they agreed and intended to pay for. Third, any employee actively involved in encouraging fraud, let alone fraud against their 2nd largest advertiser, should be fired. Their engineering, marketing, legal, and HR teams all failed miserably on this one.

I don't envision myself ever having a need to run a CPA campaign through Amazon, but based on this I would stay away from them as much as possible. They had to have multiple internal discussions about whether or not they should commit a crime against a multi-million dollar advertiser. That's certainly enough to scare me away.

Is there any evidence Discover agreed to that? As far as I can see he is the only one claiming the sole purpose of the campaign was to convert 1-click settings. That is his editorializing. Everywhere else it's mentioned, e.g. in Amazon's response[1], the campaign was not tied to that goal specifically. He even notes this himself:

"The promotion was structured in a way where anyone with a Kindle, who used their Discover card to buy a digital good (e.g. mp3 or movie), would get a $10 Amazon Gift Card."

Amazon in its reply noted that they reviewed the progress of the campaign with the customer and Discover agreed to proceed with some minor adjustments and capped to the original budget.

[1] http://www.businessinsider.com/amazon-employee-lawsuit-kivin...

I was just condensing what I read in his letter. It does appear though that their intention though was to incentivize people to change their 1-click settings. It wouldn't make much sense to give gift cards to people that were already doing what you wanted them to do.

> Amazon in its reply noted that they reviewed the progress of the campaign with the customer and Discover agreed to proceed with some minor adjustments and capped to the original budget.

He appears to be alleging that Discover made this decision based upon false or very creatively spun data. During a meeting about this, a Senior VP said: "Are we hiding something? This doesn't feel right". The reply was "In this case we're hiding that it doesn't perform well". Instead of showing them that their $500,000 got them ~100 additional purchases made with Discover cards (an absurd CPA of $5,000), someone suggested that "We can show indexed sales (on device + on site) vs. a control group that didn't see the ad". That sounds intentionally deceitful (not to mention quite evil). They tried to hide the ill effects of their colossal mistakes.

> "The promotion was structured in a way where anyone with a Kindle, who used their Discover card to buy a digital good (e.g. mp3 or movie), would get a $10 Amazon Gift Card."

That's how it worked in practice obviously, but solely based upon his letter, it appears that this was not Discover's intention. Some of the emails he quoted in his letter also appear to be pretty damning evidence that they knew that this was wrong, and essentially didn't care.

Well it's not only solely based upon his letter, it's solely based upon his editorialization. The email evidence we have shows that Discover was aware of how the promotion was set up. Kivin is relying on an appeal to logic to support this notion that there's no purpose outside of 1-click conversions, but if you think about it, there are other benefits. Plenty of cards run promotions and rewards that apply to current customers, even though the aim is ultimately to bring more people over. And that's exactly how it looks this one was set up.

They gave additional $10 gift cards on each purchase. So not only was this not based on conversion, it wasn't even based on uniques. Even if Discover okay'd the design of the ad campaign it seems pretty clear that everybody dropped the ball on ensuring it reached it's intended targets.

Not 'solely'. I think it's fairly unambiguous even from Kivin's superiors that they knew that the effect of the promotion was unexpected and unintended, and that there were "numerous concerns" from (as far as I could see) Amazon Marketing, Accounting, and Business Development about "what the hell happened, how did this go live with no or near no oversight from these business units - we need to tell the customer /what went wrong/" - their emphasis.

If that was the case though, that it was completely upfront, I don't see why discussing this with Discover would be such a big issue.

I seems at very least Amazon were scared that Discover could insist that it has been implemented wrong if all the facts were laid out on the table.

If you read the email from Jill Losser (ad buyer at Discover) on page 10 you can see she is pretty pissed off about the problem and also uncertain on the details of what happened because Amazon managers have been holding back data on exactly where the discount monies went. When you cite the line about 'the promotion was structured in a way...' you're taking his description about how it operated in practice while ignoring his statement on how it was supposed to operate (laid out in the next two sentences). I agree that this is editorializing on his part and it would have been better, if possible, to include the text of Discover's ad buy request - but when you look at the email from the Discover ad buyer it's pretty clear that she's unhappy about the way the promotion has turned out.

The best interpretation you can put on this is that Discover may have failed to be sufficiently specific in the terms of its ad buy and that Amazon misinterpreted this lack of specificity to mean that Discover wanted to give $10 to any Amazon customer using a Discover card as their 1-click option, rather than incentive customers who were not doing so to change that behavior.

Mind, I'm not commenting on the overall merits of Varghese's letter but on the likelihood that he accurately represented Discover's expectation.

> she is pretty pissed off about the problem

The problem she was pissed off about was lack of visibility into the performance data of the program, not its structure. There is no mention of a problem with how it is structured; in fact, there is a reference to another promotion, "free holiday shipping", which very conceivably would be structured the same way. Offer a promo to Discover card users, promote Discover card without going so far as to tie it exclusively to conversions.

The Amazon's subsequent email, cited by BI, notes that they resolved this visibility problem with Discover and they approved continuing the program within its original budget. That really goes against the notion that Discover had a different impression about how it was supposed to operate.

I disagree. In the first paragraph she says 'in the first two weeks we have over-delivered on the impressions allocated for the entire campaign'. Varghese argues that the issue was not one of excess impressions, but promotional monies being used up without any impressions being delivered at all, ie to Kindle e-reader users who already had Discover cards as a default and got a $10 rebate for just using their e-reader as they normally would, without ever having been served an ad in the first place.

There's explicit discussion of new vs existing defaulted Discover cards in her request for data, and a request for information on 'how [Discover] will be made good', which would not be meaningful if they did not consider any funds misallocated. In fact that's mentioned twice, the second time saying that they will still 'need to be made whole'. You don't ask to be made whole unless you've suffered some sort of economic loss, such as not getting what you thought you'd paid for.

Of course we can't draw conclusions on the basis of cherry-picked emails, but this one does clearly suggest that Discover felt itself to have been short-changed in some fashion besides a lack of analytics information.

there is a reference to another promotion, "free holiday shipping", which very conceivably would be structured the same way

Conceivably, but not necessarily, and even if it was structured the same way that doesn't mean Discover should have had any expectation about it. If was was in the habit of ordering apple pie from you and one day added an additional order for pumpkin pie, I would not be happy just to receive an additional apple pie - not because I had lost my taste for it, but because of the failure to fulfill my order for something different.

The Amazon's subsequent email, cited by BI, notes that they resolved this visibility problem with Discover and they approved continuing the program within its original budget.

That is itself a bone of contention - Varghese is suggesting Paul Kotas shares responsibility with Munira Rahemtulla for the whole situation and helped her obfuscate the issue. So without endorsing Vargheve's position, that email could be entirely consistent with it.

> There's explicit discussion of new vs existing defaulted Discover cards in her request for data... which would not be meaningful if theydid not consider any funds misallocated.

That's not really true. Cards run promotions like that all the time where the benefit is offered to all cardholders. One goal may be to drive adoption, but it also drives other goals like retention and brand value. It's false to assume that the only reason they'd be pissed about lack of data on campaign performance is that they had only the one specific campaign goal in mind.

Selectively editing my quote so as to exclude the primary point (about Discover's insistence on being made whole) forces me to doubt either your reading comprehension or your honesty. Whichever is to blame, I see no point to continuing this conversation.

Re: reading comprehension, being "made good" was specifically in reference to the distribution of Fire vs e-reader impressions -- which was in fact made good by subsequently restricting the campaign to Fire (see BI story). That's 100% consistent with Amazon's SVP statement on the matter.

There is absolutely nothing in there that states that by "made whole" Discover meant the campaign should solely target 1-click conversions. That is pure speculation and runs counter to all the email evidence.

thanks. So he didn't get caned for disrespecting the chain of command about some stupid latency issue

No, that issue seemed to get cleared up successfully. He used that to foreshadow issues with his manager and general escalation communication.

That's right he didn't. The OP didn't quite do the TL;DR right, really the letter goes something like this:

Part 1 (How I made my manager hate me): I insisted that we have better performance and everyone told me to f* off, so I went above my manager and now she hates me. (This is reasonable of her)

Par 2 She asks me to commit a crime and I raise it with HR, etc.

Reply on news.ycombinator.com