Roll Your Own Analytics • Hacker News Board

Re: Roll Your Own Analytics

#31

Post by dalbasal » Thu, Mar 14, 2019, 3:32 PM UTC

epicide wrote: post #19

kotrunga wrote: post #18

I like the idea.... but this doesn't make sense. I'm all for not using Google Analytics- I don't use it on my site. But why would you then use Google Sheets to hold the info? That's ridiculous. That's like telling people how to make an apple pie without sugar, and the last step before baking is dumping in a pound of sugar! If you want an easy data store, use something like SQLite. There are plenty of options that are…

The article isn't about removing all traces of Google from your site/stack. It's about removing 3rd party tracking from your site. Specifically, it focuses on removing Google Analytics. Using Google Sheets doesn't add 3rd party tracking back in. Even if you don't trust Google to not take a look at your spreadsheet, it's unlikely anybody or anything at Google would treat your spreadsheet like analytics data. All that…

I actually liked it. It seems like a good place on the spectrum between ideologically obsessed and practical. Google sheets is convenient.

Re: Roll Your Own Analytics

#32

Post by dalbasal » Thu, Mar 14, 2019, 3:34 PM UTC

pcmaffey wrote: post #21

kotrunga wrote: post #18

I like the idea.... but this doesn't make sense. I'm all for not using Google Analytics- I don't use it on my site. But why would you then use Google Sheets to hold the info? That's ridiculous. That's like telling people how to make an apple pie without sugar, and the last step before baking is dumping in a pound of sugar! If you want an easy data store, use something like SQLite. There are plenty of options that are…

For sure, there are a bunch of reasons not to use Google Sheets. For this, it's just an easy, free, and temp way to get a new site up without needing to setup a backend. It's not really a hard requirement for me to get off Google though, just to avoid using their ad-tech. For privacy concerns, using Google Sheets isn't all that different from using MySQL on Google Cloud, etc.

I think it's a sound, practical decision.

Re: Roll Your Own Analytics

#33

Post by news_to_me » Thu, Mar 14, 2019, 3:34 PM UTC

This is great! I’m actually building my own analytics site for similar reasons[1] and to learn Rails. It's still in early development, but I invite anyone to join and test it out in the meantime.

[1] https://analytics.zjm.me

Re: Roll Your Own Analytics

#34

Post by lukethomas » Thu, Mar 14, 2019, 3:45 PM UTC

At the past 3 places I've worked, we've setup Snowplow Analytics (https://snowplowanalytics.com/) and would strongly recommend it over GA, Segment, and other third-party systems.

If you're looking for a first-party system, Snowplow is an amazing setup.

Re: Roll Your Own Analytics

#35

Post by jayd16 » Thu, Mar 14, 2019, 3:53 PM UTC

dalbasal wrote: post #29

Working backwards, I start by defining what data I care about tracking. This is actually one of the biggest ancillary benefits of rolling your own analytics. I get to perfectly fit my analytics to my application. This could easily make the whole thing worthwhile. Ultimately, the number one ingredient for useful analytics is figuring out what you want to know, and how you are going to use that knowledge. It's usually…

Its not very scientific though. You're starting from a point of assuming you already know what you want instead of collecting data and analyzing it to see what correlates to what.

Re: Roll Your Own Analytics

#36

Post by pcmaffey » Thu, Mar 14, 2019, 4:01 PM UTC

cyrusmg wrote: post #25

I wish pages like this one had RSS feed. I would like to follow future articles, but I do not want it to end up in my email inbox.

Ok here it is: https://www.pcmaffey.com/rss.xml

thanks for the tip

Re: Roll Your Own Analytics

#37

Post by kn8 » Thu, Mar 14, 2019, 4:03 PM UTC

amartya916 wrote: post #28

fbelzile wrote: post #24

Very cool. For those that want to de-Google-ify their website analytics in a more practical sense, consider using a self-hosted Matomo (formerly Piwik) instance: https://matomo.org/ I find it has 80% of the features GA has, it's GPL v3+ and only takes a few minutes to setup.

Depending on what kind of analytics you want, Fathom is quite nice too. I recently set it up on my servers (about a month ago) and its been working well. It’s open source as well, with a public Trello board that speaks to future dev goals. https://github.com/usefathom/fathom

And here's my write up of how you can quickly host your own instance of Fathom using dokku: https://www.kn8.lt/blog/hosting-your-own-analytics-with-fath...

Re: Roll Your Own Analytics

#38

Post by dalbasal » Thu, Mar 14, 2019, 4:06 PM UTC

jayd16 wrote: post #35

dalbasal wrote: post #29

Working backwards, I start by defining what data I care about tracking. This is actually one of the biggest ancillary benefits of rolling your own analytics. I get to perfectly fit my analytics to my application. This could easily make the whole thing worthwhile. Ultimately, the number one ingredient for useful analytics is figuring out what you want to know, and how you are going to use that knowledge. It's usually…

Its not very scientific though. You're starting from a point of assuming you already know what you want instead of collecting data and analyzing it to see what correlates to what.

I would argue it's more scientific (even though the goal generally isn't science). Hypothesize, design measurements, test, conclude.

Modern science often has the opposite issue. Economists (for example) have a lot of data. They can automate regression analysis to find correlations, then fit the theory (in economics, this is often an intricate model of an economic system) to the "result."

The problem is, 99% certainty doesn't mean anything if effectively test 1000 hypotheses. 10 correlations will occur randomly.

Re: Roll Your Own Analytics

#39

Post by jayd16 » Thu, Mar 14, 2019, 4:17 PM UTC

dalbasal wrote: post #38

jayd16 wrote: post #35

Earlier quoted context omitted.

Its not very scientific though. You're starting from a point of assuming you already know what you want instead of collecting data and analyzing it to see what correlates to what.

I would argue it's more scientific (even though the goal generally isn't science). Hypothesize, design measurements, test, conclude. Modern science often has the opposite issue. Economists (for example) have a lot of data. They can automate regression analysis to find correlations, then fit the theory (in economics, this is often an intricate model of an economic system) to the "result." The problem is, 99% certainty…

More data does not make you a worse data scientist. You can run a bad study at any amount of data. The only difference is a new hypothesis will have historical data to look at. More data not require you create some over fitted model.

More data is strictly better.

Re: Roll Your Own Analytics

#40

Post by tasubotadas » Thu, Mar 14, 2019, 4:21 PM UTC

fbelzile wrote: post #24

Very cool. For those that want to de-Google-ify their website analytics in a more practical sense, consider using a self-hosted Matomo (formerly Piwik) instance: https://matomo.org/ I find it has 80% of the features GA has, it's GPL v3+ and only takes a few minutes to setup.

>and only takes a few minutes to setup.

The usual trap. That's why we have managed DBs.