5
5 Comments

Indie Hackers, what are you using to collect user data?

I've seen Segment and Sherlock as two options, but I'm curious who here has implemented these and how they like them.

Sherlock seems interesting in particular due to its ability to track user engagement as a single score that changes over time, so you can figure out who on the free plan you could offer an upgrade to, and who you should reach out to and offer pre-emptive support.

It's a little pricey though, what are your thoughts?

(link: https://www.sherlockscore.com/)

on March 13, 2020
  1. 1

    Hey Kevin, great questions. I have a few thoughts in parallel:

    1. Segment and Sherlock are kind of apples and oranges. Segment is like a data pipeline that could feed Sherlock, or Google Analytics, or Amplitude, or any combination of analytics tools. Segment itself doesn't have a UI to analyze data collected from usage, but it's a great way to implement analytics generally, especially if (as it sounds like) you haven't settled on what tool(s) you like.

    2. As an analytics practitioner day to day, I'm skeptical of Sherlock's USP, which is assigning point values to user behaviors. I believe there's a problem with having to assign a point value to each action.

    • If viewing a page is 1 point, how many points should a trial signup earn? 5? 20? 100? Once you've chosen that number, what should be the point value for an email list signup?
    • Are all pages of equal value? A view of each of the Pricing page and the Privacy Policy are equally worth 1 point?
    • If a user has a problem and hits Refresh over and over, do they rack up points and begin to look like highly engaged users?

    Personally, I would rather see actions as they are and respond to them in kind. Let me see an error event and apologize to the user. Let me see a trial signup and kick off an email sequence. Maybe I'll allow a branching path in the email sequence if they do/don't use Key Feature A that correlates highly with higher retention (as measured by months paid).

    1. I see little reason for a new IHer, early in their product cycle, to pay out the ear for analytics. Plenty of good tools have very permissive free tiers. My recommended stack is: Segment + Google Analytics + Mixpanel.

    (Plug time: I'm writing a book that's a step-by-step walkthrough of how to set all that up, building on my 8 years in billion-dollar analytics working at places like PlayStation and Rakuten. Want to be notified about the book? https://signup.analyticsforindies.com )

    1. Analytics tools increasingly fit into a few 'genres', if you will. "CDPs" are geared toward personalized communication. "Data pipelines" like Segment, and to a certain extent Avo, are developer-friendly. One 'genre' I am seeing is those that market to product managers and use copy like "product analytics" (looking at you, Amplitude) or Sherlock's "product engagement" to appeal to them. Product Managers imply that a company is a certain size, and they're usually being harangued by management to "be more data-driven", and they can usually get some spend approved. So it's a logical angle for an analytics startup to make.
    1. 1

      Thanks for the detailed write up! I’d like to pick your brain a bit further.

      It seems to me like there’s value in both types of analytics. Sherlock’s model seems valuable for two main reasons: reaching out to engaged users to upsell, and reaching out to disengaged users to offer pre-emotive support / receive feedback. It seems to me like error management is an important but separate concern.

      Signing up for your book!

      1. 1

        Certainly! (And thanks for the signup!)

        I wouldn't say so much that there are two types of analytics, but rather a lot of use cases that are well served by it.

        Having information on hand to reach out to users, either highly engaged or disengaged ones, is a common use case and isn't unique to Sherlock. Off the top of my head, Mixpanel, Intercom, and Adobe all support it. (Of those, Mixpanel has a free tier, Intercom has SMB pricing, and Adobe is pure enterprise.)

        Tracking errors for the sake of performance/uptime is a separate concern, you're right. But if you do track them, then you're also able to have that as a signal in your communications to users. If a certain user encountered a high volume of errors, you can know it, and you can take action to reduce their likelihood to churn. The more you track, the more you have at your disposal to use when understanding (and speaking to) your users.

        ps - in the event that this thread falls off, please feel free to email me. It's in my profile.

        1. 1

          So, just to put all the cards on the table, what I really want is a simple SDK I can plug into my front-end / back-end that I can do in a day, with a low monthly cost, and end up with some useful information at the end. I’m thinking usefathom.com but for user engagement. Is there anything like that on the market, or am I just over-estimating the complexity of setting up and using Segment -> MixPanel, for example?

          1. 1

            All good! I think what you're looking for is very common. (And the frequency with which the question comes up here inspired the book.)

            The work to implement Segment and then point it at any number of destinations (Mixpanel or otherwise) is the same as implementing Fathom or Google Analytics or Amplitude or whatever, but it has the added advantage of allowing you to forward to multiple destinations in the future (so you can try out new tools without technical heavy lifting, or leverage strong suits of each).

            My recommended stack at the moment is Segment -> Mixpanel + Google Analytics. Feel free to install the basic Segment snippet and try that out - you'll be up and running in a couple hours.