Leave a comment
Get the GH Bookmarklet

Ask GH

I have been using Mixpanel, but it's becoming so expensive. My (brazilian) startup has about 1M users with ~8M pageviews/month (just to have a high level idea). The problem is that I'm generating so many datapoints that I'm paying more then $1,500/month, even after removing a lot of events just to save money. But I still need to track more events so I can extract ideas to solve my problems. Is it normal to spend so much (at this stage) on data analytics? Or should I switch to something else?

What are your thoughts on an open source solution like this: https://github.com/paulasmuth/fnordmetric
So far, what I'm going to do is to calculate what is cheaper: hire a part time engineer to build and maintain an internal data analytics system or continue using Mixpanel.


  • CC

    Chris Conrey

    over 7 years ago #

    No, dear god no.

    Building your own analytics tool is next to useless until you're someone like Amazon (who has their own internal data that never leaves).

    A better question is how to solve the same problem with existing tools like Mixpanel/GA/etc.

    Also, is building that internal tool going to significantly improve your company's/clients' successes and pay for itself.

    PS - it'll be more than a part time engineer to do correctly.

    • CC

      Chris Conrey

      over 7 years ago #

      also - do you need that fine detail or can you sample it to get a statistically significant sample that is worth making decisions from?

      There seems like there should be several other ways to solve this problem rather than building a tool.

      • AS

        Andre Simoes

        over 7 years ago #

        Thanks for your feedback!
        Chris, I already solve my problems with these tools (Mixpanel/GA) and I need that fine details to take some decisions. The problem is that is becoming expensive and I don't know if it's normal to pay more then $1.500 on this stage.
        There were already several events that I needed to track but I didn't because I assumed that was going to blow my datapoints up (and $).

        But is good to know that maybe I'm wrong thinking of build my custom data analytics. Maybe I should just pay it and keep trying to keep costs low.

  • SE

    Sean Ellis

    over 7 years ago #

    Andre, in the past it did make sense to build your own analytics tool (we did at LogMeIn). But it was a huge pain, which is why I became an advisor to KISSmetrics. What previously took months to develop now takes minutes to implement. Since that time Mix Panel and others have emerged to also fill the need and in fact Google Analytics has become a very viable free alternative.

    At $1500/month, you need to ask yourself the value of the tracking you are able to accomplish. If it's not worth $1500 and you've already cut back on events, then I would recommend pushing GA to its limit. It's way more powerful than most people realize.

  • CW

    Cory Watilo

    over 7 years ago #

    Not to give Preact a plug (my company), but it's relevant. =]

    We are actually setting out to solve this exact problem. Most analytics providers charge per event and you are expected to know what is important to log. However, we've found that the most important and actionable data comes from sequences of events that are seemingly unimportant, and usually not logged. That's why we don't charge on an event-level, but charge by the number of customers you have or the number of seats of our software you're using.

    Our ultimate goal is to provide actionable insights about how people are using your software - not only in aggregate, but also on the individual-level; rather than you having to hire a data scientist yourself, we want to be your "outsourced" data science team.

    Here's a quick little walkthrough showing the kinds of insights you can get from logging obscure events, without having to pay for them: http://blog.preact.io/post/50100060994/preact-ceo-co-founder-christopher-gooley-showing

    Hope this helps!

  • GS

    Gaurav Sharma

    over 7 years ago #

    It is not such a bad idea to roll your own custom data tracking. However, don't do it just to save money because maintaining it would most likely cost more. Here are the reasons why/when I would think about implementing custom analytics.

    1. Current tools are not good enough or you have outgrown them.
    2. Automate testing based on analytics. Etsy does it well http://codeascraft.com/2011/02/15/measure-anything-measure-everything/
    3. Personalization - The site would like to personalize the content for each user. The time and effort it will take to integrate your codebase back with Mixpanel/KISSMetrics API would be more than rolling up your own metrics tracking.

    This being said, IMO the best way to use these custom solutions is to augment the learnings from existing analytics systems (GA, etc.)

  • JL

    Jeremy Levy

    almost 7 years ago #

    I’m a little late here but I thought it made sense to chime in.

    In most circumstances you definitely don’t want to build it yourself. As other people have mentioned it’s a non-trivial and a massive commitment. I know this from experience, I faced the exact same issue a number of years ago at a previous company I founded. Where the price and the power of the tools in market didn’t fit our needs, and we didn’t want to sample data because it limited the usefulness of the platform.

    I would suggest taking a look at Indicative (a company I founded to address exactly these problems). We are focused on being the easiest and most intuitive platform to help companies understand and visualize their business. We are a cloud based, self-service platform, simple, yet incredibly powerful. Priced to be accessible and affordable to all businesses.

    We built the platform with product managers and marketers in mind. Meaning it doesn’t require any technical experience in data science, databases or otherwise. It allows you to easily perform things like real-time ad-hoc data exploration, track user behavior to improve conversion rates throughout a product, compare customer behavior over time to uncover drivers of engagement and retention and customize and examine all key performance indicators at a glance.

    Happy to answer any questions and hope this helps.