Ask HN: What Data would you like to see integrated into a web analytics tool?

11 points by dageroth ↗ HN
I am currently working out what kind of public data could be imported into an analytics tool to make it more informative. For example, I have noticed that the revenue of a site usually drops if it's really hot outside, so I figured offering weatherdata would be helpful as a possible explanation for drops in traffic and revenue. Do you have any other ideas or wishes for data you'd like to see beside your website-data?

Thanks for all answers.

9 comments

[ 2.4 ms ] story [ 28.1 ms ] thread
That's a really interesting idea!

In terms of weather data how do you collate all the information: because your users could be from all over the world with all sorts of weather variations.

The only general downside I see is that some of the factors will be very website specific. Stuff like local riots, for example, might be a reason for more international hits on your local news site (that's a poor example I know because you'd imagine the site owner would figure that - but you see what I mean)

IP gives you location which gives you weather on a particular day.
You just gave me an even more interesting idea! What if you had a service which automatically found data correlations?

Create a sort of Wikipedia of timeline-based data streams. Let people have both public and private timelines. Come up with some sort of clever way to do the N^2 comparisons and find interesting correlations. Maybe even provide people a framework to conduct experiments by influencing an independent variable.

If even a few cool results sprung up out of this, you'd see loads of people pushing their data into it from all sorts of sciences. Who knows what sorts of fantastic things we can learn from the collective intelligence?

If you throw in 30 sets of numbers, you are bound to find some interesting and statistically significant correlations.

It is similar to the data mining legend that beer and diapers get bought together. They never told you how many products they analyzed. If it was 5 and they found that correlation, I'd be interested. But it was 500, then the result is pretty useless

time of day per visitor
Weather is not the only outside factor: Holidays are, national events like the Superbowl, even popular TV shows end up influencing traffic.
In addition to time of day - holidays based on visitor's location. Also, any big news stories based on the visitor's location as well as worldwide.
I want to be able to build my release log and release notes into my analytics tool. If I'm looking at a graph of page views per user and I see a spike, it would be great if I could hover over the day before that spike and see that it's when I pushed out a new feature that apparently increased engagement. It puts everything in context.