Hacker News Reader

So... what's the flaw in the data gathering that leads analytics people to believe this? I got a ludicrous amount of real estate advertising over the first year after I bought my house which was when my "whew, glad I never have to do that again" feelings were strongest. Is it just that they extrapolate from consumables?

(About the only time I'm in the market for "another one of those" is if the first one was so low quality that I returned instead of putting up with it - or if I sampled a few things to see which one was good and then need enough more to finish the project.)

Not claiming that the ads don't have some influence, but they pretty much can't result in another sale...

I think the advertising industry has a very long tradition of relying on and trusting bad data if it's the only data they have. As long as everyone plays along and believes that Nielsen ratings or circulation numbers or click counts are accurate, you can have a more or less functional market for advertising spots. And there's obviously demand for more detailed data, as long as advertisers can believe that it is accurate and can make their ads more effective.

When the analytics produce an obviously wrong conclusion (such as saying you should be shown more car ads immediately after completing the purchase of a car), everyone involved has a vested interest in believing that the analytics must be right in some fashion. Doing otherwise would mean taking on a big risk, straying from the herd with a different advertising strategy that's guaranteed to take the blame for any drop in sales in the near future.

> When the analytics produce an obviously wrong conclusion (such as saying you should be shown more car ads immediately after completing the purchase of a car)

Sorry to say: Probably another wrong intuition.

My husband was in an accident that totaled his car not too long ago. So we bought a new one. Then after having it for a few weeks, I realized that I liked many of the new features it had so I started looking for one, too. True story.

The first purchase of a car in our family caused (and, more importantly for this discussion, was predictive of) the purchase of a second car. I’m sure this does not only happen after accidents.

Sorry to say: Probably another wrong intuition.

I have bought multiple cars in my lifetime. But I have *never* bought one right after another. This is not intuition, it's a fact.

No one is suggesting this can't happen --- but is it likely? Is it the norm?

"Personalized" ads assume it is --- and erroneously so in my case.

Start looking at cars online and you'll be bombarded with car ads for months after you've made a purchase. That ad money was just wasted on me --- and I'm sure on many others as well.

"Context sensitive" ads make no such erroneous assumption --- and they still have you covered in any case. You are shown ads only as long as you continue to express an interest in the subject.

There is no flaw. If you go to any retailer where sales can be tied to accounts, and you run a few SQL queries on their sales history database, you can generate a table of counts of transaction_n and transaction_n+1.

You will find that the most probable (the mode of the distribution) thing to buy in transaction n+1 is either a substitute for, a complement of, or identical to transaction n.

isn't there a difference between buying a gallon of milk every week because that's part of your standard household equipment versus an item you buy once every couple of years?

i think no one would disagree that there are things they buy very frequently, almost on a schedule. (then again, with these items people are probably very accustomed to buying the same brand of milk every week, so ads don't seem reasonable here as well).

The incentive to show post-purchase ads is: 1. people didn't like nth purchase so they're looking for a substitute. 2. reinforce purchase behaviors / reduce buyer's remorse.

Ads work at human psychology; they're not fully logical. Though I'm sure there's inefficiency in marketing but if post-purchase ads weren't ROI positive, I doubt the market would be paying for them.

Like I said before, this phenomenon is broadly applicable across almost everything. I’m hedging with “almost” because I haven’t personally seen data for every category of product, but I believe that it applies universally.

In a sibling comment I replied to someone with an objection that it doesn’t apply to cars… with a personal anecdote that it does, at least sometimes, apply to even cars.

Even governments do it: Having previous purchased a nuclear submarine is highly predictive of future nuclear submarine purchases. Likewise, you probably have not purchased a nuclear submarine, which makes me think you probably won’t buy one this year. ;-)

Does that apply across all product categories?

I can totally believe it for something like bananas or shirts. If I just bought some there’s a good chance I’ll soon buy more.

But vacuum cleaners? Cars? Who’s out there buying those more than once every couple of years at most?

They're A/B testing all the time, and it wouldn't take long at all for them to figure that out.

How? There is no way to A/B test if B doesn't exist.

I'm talking about "B" being a practical ad network alternative to Google's effective monopoly. An offering that provides everything Google does --- except the "personalization". Instead of keying on the person's history , key on the context --- what they are currently searching for or what they are currently looking at.

As far as I know, this really doesn't exist in a generalized, competitive form.

But it would be relatively easy and cost effective to implement. Just bring back the <META name="keywords"> tag and key on it instead of the person. Think about all the time and money Google spends to invade people's privacy --- and simply eliminate it.

The ad networks want advertisers to *believe* that the "black box" they are offering is market based and cost effective but they have no real, practical way to confirm this.

They're called search-based targeted ads. That's where Google's AdWords started from, and you can still do that if you wish. But personalized ads work a magnitude better.

What advertising really needs is not more from the same monopoly but rather an alternative to it.

The fact that Google choose to steer advertisers away from Adwords by overcharging for them doesn't mean that they can't be made just as effective.

You're claiming that everyone in the industry knows personalized ads are better, and not that they're just using the most available or most popular solution, or the solution their superiors are most comfortable with.

How do they know this? Are there papers or something?

Across... well, basically every category of product... the product which you are most likely to purchase next is the same (or a substitute for or a complement to) the last product you purchased.

Thanks for summarizing the logical fallacy of "personalized ads".

Maybe what you say applies within the narrow context of a single brick and mortar retail establishment --- clothing for example.

If I buy a shirt at a *clothing* store, when I visit this same *clothing* store again (in a few weeks or months) I may be inclined to buy another one. The mere fact that I took the time to visit again suggests it is likely.

But just after buying a shirt, how likely am I to buy another one when I visit a pet supply store across town?

The brick and mortar world is naturally divided and separated by context. When you walk into a pet supply store, you *never* get hit with ads for clothing.

This is not true with on-line advertising which is totally divorced from any rational context and as a result, the ads are much less effective than they could be.

And the ad networks don't really care if the ads are effective or not. In fact, they probably prefer it not be to encourage advertisers to spend more on advertising in a "black box" system that they fully control.

Last attribution doesn't work. Retargeting is a joke and the most valuable spenders have ad blockers so you can't even reach them through display ads. The data is manipulated everywhere as long as you can hand wave towards ROAS.

So basically contextual ads suck.