Be a startup Metric King | Hacker News Reader

Be a startup Metric King

klinger.io

57 pointsmmenafra23 comments

Open articleView on HN

Author here: Please don't growth hack my titles for your submissions. Thx ;)

Do you keep anywhere a quick list of various metrics for various types of cases, for a startup?

Awesome question.

Imho the core metrics are quantified product assumptions.

Some related to user interaction, later (hopefully) some related to businessmodel assumptions.

Depending on phase the focus tends to shift from retention to growth.

Alistar and Ben made an awesome chart for their Lean Analytics book (recommended) http://i.imgur.com/0VBP68T.png

These slides by me also might help: http://www.slideshare.net/andreasklinger/startup-metrics-a-l...

hth

Awesome post, I think you should write about your Google Analytics setup. I recently switched to Universal Analytics, and it's been quite a mess.

Also, I highly recommend checking out Chartio (www.chartio.com) - since it's pulling everything from the DB, you can track and measure anything imaginable (http://chartio.com/docs/charts/transform/formulas)

My stack is GA for acquisition, Mixpanel for product analytics, and Chartio for the stuff in between. It works well, but there's always room to improve.

100% agree on chart.io i tend to recommend it when people want a more complex dashboard setup

most startups i recommend to keep with simple dashboards

What issue(s) are you having with GA right now? I've done some advanced configurations with GA and maybe can help.

In a nutshell, we fire a server-side event when a user is created. What happens is that we're using oauth to authenticate, which then screws up the referring data. It attributes all signups to LinkedIn/Facebook, which is wrong. On classic analytics, everything worked fine. Universal Analytics tracks differently, so it's been a bit of a headache.

Without having a full view of what's happening, my hunch would be to look into these things:

1. Channel grouping (modify how referrers are defined)

2. Attribution model (eg, change to first interaction)

3. Referrer override (specify custom URL to be credited with referral)

4. Custom page path (on page that fires event, pass custom URL path to GA instead of actual URL)

5. Referral exclusion (ignore temporary stray into oAuth domain and continue user's same session)

Hope that helps. If you could use a hand in getting this fixed quickly then get in touch (contact info in profile). Same goes for anyone else with tricky Analytics situations!

Hi, Sorry really liked your article wanted to share it with everybody. No harm in that. piz

no worries. You seem to be better with titles than me anyway :)

The article, and its original title, are structured and "theoretic" (not quite the word I want, but fairly close): it is very similar to college-level reading material. The replacement title honestly sounds more like "silly blog post from someone who hasn't thought much about the problem yet". I thereby clicked this link because I was interested in seeing the comments, not the article; it was only after I saw the author say the title was modified that I bothered looking over the article. I guess a comment to make is that the original author also likely wanted to share his content with everybody, and probably already attempted to choose a title that he felt portrayed his content in a light that would make that possible. Sometimes, titles should be more context-specific, but the kind of audience this post was targeting is probably itself similar to Hacker News. Meanwhile, due to Hacker News having an official policy against the modification of titles (for better or for worse), if it isn't clear that the title was re-targetted (maybe pointing to a specific paragraph or idea), the reader will expect the title of the actual post to be equivalent to the one submitted, and make judgements accordingly.

Yeah i understand you. Didn't want to make the post sound silly, but I guess that's your opinion on the subject. The only thing am confused about is that I see NH post all the time that the title and the article title are not the same, but i'll guess from now on every time i share third party content i'll stick with the original title or something that relates directly to the article.

cheers

Hi andreas, very cool post. I agree 100% that one should leverage his own database first thing: those are numbers easy to truck and on which you can rely on.

Since I am looking for to undertake a medium(ish) OS project in Flask and D3.js I wanted to ask: do you think there might be a need for a good simple app that startups can run on their own servers and can use to segment their database?

I know there is a bunch of analytics platforms out there. I work in finance, believe me I know. But I was wondering if there is any particular pain point for startups tracking their db numbers that has not been fixed yet.

Totally. There is one tool missing:

A simple way to create own (database depended) dashboards. Most of the OSS tools (eg http://fnordmetric.io/ etc) focus on time-based graphs. Cool stuff.

But personally i would rather look for a OSS tool that distributes also report files/recipes that startups can share. Eg. Retention Cohort, AARRR etc.

If it would also have a simple way to add graphs even better.

I was actually working on this kind of tool but stepped above my personal developer (skill) comfort-zone. If you want more feedback feel free to ping me andreas%klinger.io

I found this a great read and immensely useful, but the styling to pull out words/phrases in bold felt really distracting, and above all unnecessary.

Though it might be Chrome on Win 7, which makes fonts generally look awful.

Thanks for the feedback

I need to get a new theme. Really unhappy with this one (and it's fonts)

Unfortunately, it is only/nearly only web metrics (he mentions accounting somewhat down the page).

No discussion of productivity metrics, burn rates, etc.

Agree.

FYI: I mention it in the top as well.

I'm surprised more folks don't store data about every page view/action in their database. If you have a lot of data, postgres_fdw makes it easy to scale. Analytics becomes much simpler if all your data is in a single system.

Agree

But to be fair: In my exp the timestamps on models are usually already sufficient. In most cases you want to monitor state changes.

Not sure what you mean.

Say you are an e-commerce business. If you store each page view in your system, it's simple to write a sql query that details where all your visitors are coming from, the path they take to order something, how long it takes them to order something, which users haven't ordered anything or visited the site in the past two weeks, etc.

Adding a timestamp to a table doesn't really help here.

What about intercom.io ?

i usually recommend it in my workshops/talks/consulting

the post was already getting to big

i might will do a dedicated post on health dashboards - where they fit in well

Reply on news.ycombinator.com