Hacker News Reader

I have a dumb question, and maybe it's because I don't understand how ebay affiliate stuff works. (Assuming Amazon-esque.)

But even if this site sends really low quality traffic, it's still likely sales that wouldn't otherwise happen, even if the ratio of visits to sales is quite low.

What does this cost ebay other than a lil more server juice? (to have a fruitless visit).

I mean, I know they're paying a cut to MCM, but they would pay that cut with good quality traffic, too.

It seems their gamble on false positives would not be as risky with other set-ups (paying out for downloads, trials, etc).

(If anybody wants to school me on incorrect ebay assumptions, that would be OK. I'm assuming it's basically like Amazon.)

[update: Reread. The author says he gets commission based on a variety of things including "clicks, new accounts, bids, and wins." It seems that this whole dynamic could be clear for both sides if ebay only paid out on things that ended up getting bought.]

Affiliates typically work on a CPA basis: cost per action. eBay is a little trickier, but it boils down to that. They get paid based on conversions, not based on traffic.

Most people who sell via affiliates have what is called a backend. (BCC does not.). It means, loosely speaking, any way for the company to increase LTV [edit: life time value, total economic value extracted from customer over the entirety of your relationship with them] after the initial sale, which is typically when the affiliate gets compensated. eBay LTVs are dominated by he backend: I bought my first box of Warhammer figures for $20 and earned eBay a buck back in 2000. Over the last ten years, I have been worth north of $2000 to eBay.

It is very difficult to know with certainty that the LTV of the kid buying the Warhammer figures is going to hit four figures because in a few short years he is going to move to Japan then drop off your radar for a few years then sell a hundred thousand dollars of bingo card software using his Paypal account.

So you need to guesstimate LTV for customers, within 30 or 60 or so days of acquisition, to pay your affiliates. If you underpay, they send you less traffic and you lose. You probably pay your affiliates more than the instant revenue. If eBay figured my LTV was $1 and paid the affiliate 50 cents, they would have sent me to another site instead. If Netflix paid only the first month's revenue, they'd lose signups to other sites.

If you overpay, you hemmorage money.

So you get really, really sophisticated about measuring and predicting LTV. You understand your business better than any affiliate can hope to. And when the numbers predict red ink? You drop them like a bad habit.

What is LTV?

life-time value (of a user)

Can eBay pay the affiliates based on the realized LTV? i.e. eBay pays 50 cents when you're worth $20, and as you increase your value they keep on paying the affiliates. After all these years, you're worth north of $2000 to eBay, and eBay has paid out a total of $50 to affiliates along the way.

I can see what the practicality might be like, but it seems much simpler and fair than estimating LTV and cutting off affiliates if you were incorrect.

That is not technologically or organizationally impossible, but there are many reasons to prefer paying based on estimated rather than realized LTV. One reason is that many affiliates incur non-zero traffic acquisition costs immediately, or close to immediately, because they use AdWords or similar methods. (Or, alternatively, because their SEO enjoys being paid money on a regular basis. I certainly like getting my salary from the bingo business regardless of whether the business sells a lot or a little in a given month!)

eBay currently extends them money far in advance of eBay receiving it, with the understanding that if there is an upside eBay will get most of it. If eBay pays them only after eBay gets money, that will cause many of them to go cashflow negative, and since they are not billion dollar publicly listed companies with easy access to capital at low rates, that leads to the firm either failing or exiting the eBay relationship. Neither of these outcomes benefit eBay.

There are numerous other risks this exposes affiliates to -- eBay retroactively changes the rules, eBay becomes unable to pay out the money, eBay has extrinsic changes in their business which lower payments after affiliates have incurred upfront costs sending them traffic, etc. All of these risks make eBay a less attractive partner to do business with.

You can't always have everything you want. There is a tradeoff here: eBay will pay you substantial amounts of money, up front, absorbing most of the financing and execution risks. However, in return for this, you agree to having your business relationship governed by what will appear to be black magic.

Very interesting example. This is definitely the reality, but perhaps you can teach an old beast new tricks.

1. In the hypothetical situation you provided, it seems that if your CPA is based on trial signups, the cost structure doesn't scale with the demand and value you get from using the SEO charging per trial. If you generate revenue on a conversion, why do you have a CPA based on a trial registration? Doesn't eBay pay per conversion? Perhaps this is a common practice. Can you renegotiate to a more reasonable cost structure, maybe get it right from the start or allow some flexibility if you and the SEO aren't aware of how the costs will play out?

2. If eBay is unhappy about the LTV per CPA, why not agree on a cost structure that more readily scales with varying LTV rather than blame the affiliate for not meeting a particular threshold?

3. There is an opportunity cost with firing an affiliate due to poor traffic quality for a month rather than keeping them consistently there. You might fire your SEO for one month's bad performance in a heavy-handed fashion, but what if Christmas, not just any old Christmas, turns out to be a legendary one. Bingocardcreator is the new hot fad for Christmas family games night and conversion rates are through the roof. If there was a cost structure that hedges odd months like October, (3-month moving average of CPA for example) your SEO would be around consistently, ready for a key opportunity.

I see parallels here to most successful farmers that use futures to hedge their fuel costs and stabilize their revenue in varying weather and environmental conditions. It won't save you from an SEO not being economical, but it will make the costs less uncertain.

4. Helping an SEO affiliate optimize on an aspect that may or may not be under their control may be a big task, but already, providing a conversion table like you did is probably more than what was probably provided by eBay.

5. It is not economical to work with people to optimize their sites to hopefully be sending quality traffic again, since that is hard, non-obvious Sites don't have to do a lot, just be transparent and manage expectations above the bare minimum. Affiliates like everyone else don't like uncertainty mixed into their return on investment. Even if the nature is an uncertain one, you can still manage expectations to keep that fact transparent, and keep a cost structure flexible but clear enough to support.

6. might be against the interests or intentions of the site owner, and consumes resources which would be better spent optimizing the customer's own sites. Brilliant insight, this is probably a factor as well, but I care to believe that eBay still handled this case in a sloppy fashion.

The more voodoo you do to maximize the alignment of the affiliate program and eBay's long term interests -- and eBay does do the voodoo that you do so well -- the less easy it will be to understand for anyone who is on the outside looking in. My transparent, easy-to-implement, easy-to-understand hypothetical $0.25-per-trial affiliate structure is harmful to the developer's economic interests and harmful to the desire of affiliates to continue working with me.

After we switch to any of your suggestions with dynamically scaling payouts -- which eBay did a long, long time ago -- we enter a world where the payouts, seen from a vantage point outside the system, resemble black magic. That is exactly this site's complaint. They have been judged by the black magic and they have been found wanting.

There is no way to explain to them what they've done wrong because it is an emergent behavior of the interaction been Google's rankings (black magic), their own site and strategy, eBay's business practices (black magic), the underlying economy (black magic), etc. There might very well not be any person in eBay who can explain, given any amount of resources to investigate, what this account did "wrong". But they can justify, with incredible predictive capability, that systemwide the black magic generates higher average LTVs for lower average traffic acquisition costs.

Any sufficiently advanced conversion optimization system is AI. (Blog post coming up on this.)

Thank you Patrick. I'd like to do more research on the topic and await your blog post. As this topic pertains to eBay, it is extremely interesting.

As you mention some other ways are not impossible, but have serious problems. I agree, but at the very least, I think the status quo is not optimos. If you are starting up an affiliate scheme, it would be interesting to explore all options even ones that seem more complex or unconventional, because frankly, the ultimate black magic is a spontaneous killing of a business with no warning signs or even the converse. Either eBay wins (by getting some extremely high-LVT traffic) or the affiliate wins then loses (by getting killed). There has to be a way to a middleground.

wow, very insightful response.