Hacker News new | past | comments | ask | show | jobs | submit login

Any clever advertiser doesn't judge the success of a campaign by its number of clicks, but by its number of conversions (and ultimately, revenue). What's become a new pain in the advertising space are spam/ghost referrals, messing your analytics only for the sole purpose of making you visit their websites, also adding up to the stack of fake audience of your site (which the article fails to mention).



The really crazy thing is that tracking from click to conversion is still not a truly solved problem!

e.g. When a person clicks, and then closes the browser and/or views other web sites, only to come back as direct traffic later and convert. The funnel is very hard to track and very easy to lose.

Not that I do this full time (I don't), but I've yet to really see a solution that lets me truly match up a person who clicks an ad and then converts.

At best, I can roughly correlate ads to a bump in revenue.


Your example is actually pretty easy to track. That's why sites drop cookies and have look back logs.

Harder to track is when someone clicks on a link on mobile, and then switches to desktop to complete the transaction. Unless everyone is logged in, you have no chance of tracking the funnel.


I know nothing about advertising, but isn't this method thwarted by users (likely small in number) that either do not allow cookies or dump cookies after a session? The latter type would only affect tracking if the transaction was completed in another session. I would think that IP tracking is somewhat useless due to shared public address among users.

Really what I'm asking is that is it trivially possible to track a user that does not store cookies?


Yes, dumping cookies makes it much harder to track lookback conversions. Some networks will use fingerprinting (IP, user-agent-- panopticlick style stuff) to supplement their data, but it's much less reliable and a determined user, like someone blocking or dumping cookies, is going to defeat it.


Boris bought iframes on doubleclick (among other places). That's how he gets legitimate users' cookies.

Assumptions about video are that it not be clicked, so post-click analysis just isn't done.


In some cases it's not even the same person doing the purchase, which adds another layer of complexity. Especially common with services that might be purchased by a business. Lately I'm getting a bunch of seemingly targeted PaaS and IaaS ads. I click on them occasionally. If one turns up a service I find interesting, there's a good chance the sale it leads to will be initiated on a different computer, by a different person.

This even happens pretty regularly for consumer stuff. I've suggested travel-related things to my parents, for example, which have led to direct sales: I saw an ad for a sale at $hotel_chain, remembered that my parents are taking a trip in October, tell them about it, and they book. It's pretty hard to track back their purchase to the ad that was shown to me. (Not entirely impossible, though, e.g. giving out different coupon codes is one method that print advertising uses.)


In Google Analytics you can get rid of most spam referrals by creating a filter that only passes traffic that landed on your actual hostnames.


The whole problem is really just the age old issue that if you track the wrong metric, people will optimise for something you don't want.

It's like paying programmers per line of code written and then getting bloated and shitty code.

You need to be doing CPA (Cost per Acquisition) deals not CPM or CPC.


CPM can drive your costs down on a highly targeted niche. Let's say you want to target a female tech journalist living in SF and using Slack, you'd rather use CPM than CPC as you're pretty sure there are only 20 people that match the criteria and even though you'd pay 50 USD per 1000 impressions, you'd only pay 5 dollars to ensure each person in this particular niche sees your ad 5 times!


CPA is hard work.

I can count on one hand the number of people in this industry that like hard work.

It's not just the wrong metric, it's a fundamental misunderstanding of technology: for example the "viewable impression" as being anything other than software that is telling other software that it's on the screen...


The (myspace) vehicle the article is being sold as CPM pre-roll video, and not per click as the expectation is that people don't click video ads when they're watching an ad before the youtube clip that they came to see.


The deep pockets in online advertising aren't able to easily track the effectiveness of their ads.

They're big brands, like Target, Old Navy, Tide, movies, etc.

They don't even care if people click their ads. They just want the impressions.


The advertisers in this case are buying video from Myspace.

There's no expectation that people click on video.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: