Hacker News Reader

Comment by ot | Hacker News Reader

> Google said this change is to allow the search company to serve the search results faster on more searches

I'm surprised that this makes enough of an impact: in order to find the top 10 results the engine needs to retrieve and rank a much larger number anyway (I'd guess at least 100 make it to the final stages of the funnel), and that is where virtually all the cost is. That initial set is probably held in some cache so that subsequent page loads don't re-do the search from scratch.

So either this is a small win in frontend efficiency, or continuous scrolling is fast enough that a large fraction of users goes past the initial set?

View on HN

I suspect the problem is that the most expensive to serve users go past the initial set.

Yet I believe most of those users are the lowest value users - many of them will be bots and web scrapers, whilst others might be real humans doing long tail queries that have no relevant ads to serve.

I'd pay extra both to serve and to receive incredibly niche ads that would only show up for long tail queries.

That was the original promise of all the ad tracking, wasn't it? Outrageously accurate ad targeting? Reach the weirdos in your tribe wherever they are?

Now it just feels like drowning in a sea of things a bunch of giant corporations have decided everyone ought to want, and have been programming people to want en masse, rather than helping people find the unique, interesting little things they would actually want were it not for all the programming.

> most of those users are the lowest value users

Yes they are. Because they want results not SEO "content" sites.

Caching the retrieved results for future follow-up page 2 queries is nontrivial. It’s also not clear if it’s worth it, the number of requests going to page 2 are in the lower percentages at best, maybe even <1%.

So what you typically do instead is to just issue a new query and request 20 instead of 10 results, drop the first 10, and voila, there is your page 2.

I do not know if this is how google websearch does it though.

Not sure they cache, at least not the SERPs. The SERPs vary by user, and even browser.

Open a private window in Chrome, FF and Edge. Go to Google. Enter the same search term and it's likely the SERPs will be different.

> So either this is a small win in frontend efficiency, or continuous scrolling is fast enough that a large fraction of users goes past the initial set?

Yeah, it doesn't really make sense. Infinite scroll is pages tacked onto each other in the frontend - the backend should be doing the same amount of work for page X, except less overall because it's not returning the header/footer. I don't think frontend efficiency works either because the first page of results should be the same - or, possibly, the tiniest bit more work because now they have a footer to return as well.

Reply on news.ycombinator.com