but can you rephrase that in terms of revenue? and tbh 10% is right about the amount of pain you 'd expect to inflict to monetize your traffic.
Gwern's Law: Ads cost you a tenth of your users
11–20 of 323 posts
Re: Gwern's Law: Ads cost you a tenth of your users
#12Not having ads costs you 100% of your members if it's the only way to monetize your content (not all content is suited to be supported by paywalls or Patreon).
Re: Gwern's Law: Ads cost you a tenth of your users
#13Only if they are done intrusively. Nobody cares about the ads on CrossMe, for example...but Jewels Jungle, the interstitials may aggravate a small number.
I have troubles with the concept of a "non-intrusive" ad - If you don't notice it to any degree, it's not doing its job, and if you DO notice it to any degree, it's intrusive and disruptive (to a matching degree) I think you mean "Nobody cares ENOUGH" rather than "Nobody cares", and that's a big distinction.
I think "intrusiveness" is pretty subjective and has a lot to do with the content of the ad.
If HN ran native ads for reverse mortgage companies instead of job ads for YC startups you'd feel very different about them even though the format is the same.
Re: Gwern's Law: Ads cost you a tenth of your users
#14That would make sense. Ads are designed to be visible and clicked on, but on the other hand you can't go too far without annoying the users. It's a fine line to walk so it's not surprising you'd lose some users as a result. The unblockable ads they had on Facebook at some point is what finally pushed me to close my account.
but on the other hand you can't go too far without annoying the users
Back in the ancient history before Google and search engines, a web site I built was submitted for inclusion in DMOZ. At the time, all web sites were hand-approved by an editor who also wrote a description of the site.
The site was included, but the description noted that it was "covered with too many ads" or something similar.
The site had one 468x60 static image banner ad, and one 236x60 static image ad.
My how our standards have changed.
Re: Gwern's Law: Ads cost you a tenth of your users
#15I assumed this would be the case and removed ads from my web game about a year ago. But didn't see a difference with traffic or retention (granted, I didn't A/B test it! I removed them 100%). Which was disappointing. The only difference I noticed is that I lost that revenue (5 figures).
But probably depends on volume and/or type of ads?
[EDIT] Clarified re: traffic
Re: Gwern's Law: Ads cost you a tenth of your users
#16That would make sense. Ads are designed to be visible and clicked on, but on the other hand you can't go too far without annoying the users. It's a fine line to walk so it's not surprising you'd lose some users as a result. The unblockable ads they had on Facebook at some point is what finally pushed me to close my account.
Also, at the risk of stating the obvious: clicking on the ads takes you off the site in question. Some of the people who would have been clicking around your site more are now at someone else's site.
Click rates (CTR) are usually quite low, if you get 2% you can open the champagne. Or at the very least, keep it at the ready and look at your conversion rate.
Re: Gwern's Law: Ads cost you a tenth of your users
#17I assumed this would be the case and removed ads from my web game about a year ago. But didn't see a difference with traffic or retention (granted, I didn't A/B test it! I removed them 100%). Which was disappointing. The only difference I noticed is that I lost that revenue (5 figures). But probably depends on volume and/or type of ads? [EDIT] Clarified re: traffic
'Revenue'!='users'. EDIT: and 10% is, while important, a pretty subtle difference compared to the enormous variability of traffic. Traffic is a time-series which drifts and has many potentially large jumps. If you removed ads only once, how would you ever notice? The effect is very noticeable if you look at samples with _n_ in the millions, or if you remove/add scores of times and model it statistically, though.
Re: Gwern's Law: Ads cost you a tenth of your users
#18Only if they are done intrusively. Nobody cares about the ads on CrossMe, for example...but Jewels Jungle, the interstitials may aggravate a small number.
I have troubles with the concept of a "non-intrusive" ad - If you don't notice it to any degree, it's not doing its job, and if you DO notice it to any degree, it's intrusive and disruptive (to a matching degree) I think you mean "Nobody cares ENOUGH" rather than "Nobody cares", and that's a big distinction.
Intrusiveness depends on the ad's relevance and value:
An ad for a product you don't care about is disruptive.
A 30% off promotion for a rarely discounted product you've wanted for a few years is amazing.
Re: Gwern's Law: Ads cost you a tenth of your users
#19but can you rephrase that in terms of revenue? and tbh 10% is right about the amount of pain you 'd expect to inflict to monetize your traffic.
No, because that will depend on the purpose of a site, what alternative revenue streams there, the site owners' utility preferences between traffic and revenue, what sort of ads there are, what sort of clickthrough rate they have, what the advertising rate is, and so on. At best, you can make a directional generalization: given that people systematically appear to underestimate the effects of ads (they just don't know), there is probably a large but unknown number of people who have overshot the optimal number of ads by an unknown amount, which will be different for each person.
In contrast, the binary variable ads/no-ads appears to have a (surprisingly) consistent effect on general metrics of users/usage, as measured by 5 very different large-scale studies.
Re: Gwern's Law: Ads cost you a tenth of your users
#20I assumed this would be the case and removed ads from my web game about a year ago. But didn't see a difference with traffic or retention (granted, I didn't A/B test it! I removed them 100%). Which was disappointing. The only difference I noticed is that I lost that revenue (5 figures). But probably depends on volume and/or type of ads? [EDIT] Clarified re: traffic
I mean, he's not making a prediction on revenue. Gwern is making a statement about traffic. Did your traffic increase ~10%?
From the Pandora paper, (which showed that as more ads were played during pandora sessions, subscriptions were boosted but free users left) it sounds like there's an "optimal advertisement" rate for revenue. Enough to not push away users, but not too much to deter users altogether.