Analytics App Store + Power User Curve + Amplitude ML Engine + More (PAN #36)
Edition #36 - September 7, 2020 Originally sent via Mailchimp
Good morning product analytics friends 👋
I’m taking you on a trip down memory lane today. And that’s all just an excuse for me to dust off my crystal ball and talk about the analytic’s App Store layer.
With that, on with the 36th edition of the Product Analytics newsletter!
What has been my highlight?
The Analytics App Store
Lantrns.co by Yours Truly
Again, shameless plug for my own content. But to be fair, I don’t really publish often on my blog and so I hope you’ll forgive me :)
So what’s this one about? I’m taking a philosophical crack at how I see the analytical landscape evolving and what the future might hold in terms of innovations.
I’m old enough to have seen and participated in the first days of web development and how it gradually became a commodity to be able to build web apps. I’ve witnessed first hand how websites were cool just because you could display animated disco balls on a webpage, or create dynamic menus by manipulating a browser’s DOM with on-click events. But when CMS such as Wordpress and frameworks such as Ruby on Rails gained popularity, that became accessible to everyone and web apps could be built easily, efficiently and cheaply.
Value shifted to how you added value on top of that layer. Frameworks had themes and plugins. And the Lean startup philosophy thought us that not all ideas should be built, but that you should build MVPs and only evolve the ones that gained traction.
My argument is that modern BI architectures are almost like the CMS in the analytics landscape. The tools, frameworks, best practices are evolving and it’s getting cheaper to own your data in cloud data warehouses. There aren’t any big barrier to entry to building your own BI stack and we can see troves of practitioners and companies joining the party.
Commoditization is not a bad thing, to the contrary. It’s like how iPhones made mobile phones a commodity (as opposed to how owning Blackberries was exciting in itself) and that trigged innovation in the form of mobile apps.
My question is how will innovation manifest itself in the analytics landscape now that we’re witnessing a commoditization of modern BI architectures. What is going to be the “App Store” equivalent for modern analytics?
Agree or disagree with that trajectory? And how do you see innovation happening?
Growing your product with the help of data.
Jack Dorsey on Twitter’s Mistakes
Spotify.com by @nytimes
Since 2016’s elections, social medias have been under the spotlight because of how they might have been gamed in order to influence the election’s results. Twitter acknowledges that and this conversation on New York Time’s “The Daily” podcast is a great insight into product ownership and how all decisions impacts how users interact with the product.
In a sense, behaviour is a reflection of what you consider important. There are incentives that is “gamified” within a product and that supercharges behaviours. For Twitter, the likes and retweets are important signals that drives content creation and distribution.
The conversation is refreshing as Jack is very transparent on earlier decisions that have had deep ramifications. The point is that you should think early about what your gamification engines and what you measure, as this correlates directly to the behaviours they trigger.
Factory operations to transform data into analytics.
Dagster: The Data Orchestrator
Medium.com by @schrockn
I might have already talked about Dagster in the past, but even if I did, this is a thorough introduction that’s worth your time if you’re not that familiar with it.
Here’s an interesting quote from it:
“We think data engineering today is in a similar position as web frontend engineering was a decade ago: wrestling with a novel and complex domain, dramatically under-tooled, and often regarded with unjustified disdain by practitioners in better-understood domains with more mature tooling, like systems programming.”
I don’t see how that’s true in the analytics landscape and that’s pretty much the core of my argument in the Analytics App Store above. There is tooling that enables dataops practices such as the ones present in systems programming. That said, there’s always room for improvement and could be that Dagster is filling a need for broader data apps.
I classify this one in the « need to further explore » category. If you do have something to share that might better illustrate use cases of Dagster with dbt and/or Great Explorations for example, I’d be really interested in learning more.
Deriving insights from your product’s data.
The Power User Curve
AndrewChen.co by @andrewchen
You might be familiar with DAU (Daily Active Users) and MAU (Monthly Active Users) as metrics for user engagement. But as argued here, it’s really the power users of a product that brings it to life (influencers, power sellers, power riders, etc). For those we need another type of metric, hence the L30.
“It’s a histogram of users’ engagement by the total number of days they were active in a month, from 1 day out of the month to all 30 (or 28, or 31) days”.
Lots of great ideas on how you can use that metric to look at the evolution of engagement by cohorts, by features, new releases and other segmentations. Also, where D(M)AUs only gives you a single metric, you actually get to see a nice curve that provides richer insight in engagement.
What’s happening the product analytics market.
Predictive Analytics with Amplitude
Amplitude.com by Cindy Rogers
And finally this is just something that hit my radar lately and I still have to dig into. Although I do argue that it is better to own your data and come up with your own product analytics, tools like Amplitude are pushing the envelope and that’s always welcomed.
We’ve talked in the past about how machine learning can be part of your data warehouse transformation process, or at least how it could become even more part of it. Nova is the machine learning engine for Amplitude.
This isn’t a new feature from Amplitude, but an enhancement of what they already have. And that post goes into much details about how they are injecting ML into their analytics.
I’d love to get ML-driven fields in my warehouses, such as predicted churn and LTV for product users. Could the ideas from Nova be ported to modern BI architectures?