Learnings from one week with Gyana


David Kell

Sept. 24, 2021 • 5 min

A week ago, we launched Gyana 2.0. It’s a rebuild of our tool, powered by our community.

When you’re focused on building a product, it’s difficult to experience it as a user. (I recently explained this as the difference between testing a car in a garage versus driving it on the open road.)

The original drive to build Gyana came from frustrations using data in our business. It was important to me that we could solve our own problems, before we try to convince everyone else!

And so, for the last week, I’ve been solving real Gyana problems with Gyana. And I’m happy (relieved?) to say that, after many iterations, we’ve built a tool that I’ve always wanted to use. (So if nothing else, we have one happy customer!).

We’ve used it to understand market segmentation, visualize Gyana 1.0 behaviour, track company KPIs and analyse Gyana 2.0.

Not only has it provided a replacement for the DIY stack we’d otherwise have built (and spent $$$ on), it’s unlocked new ways of working that I never expected. It’s really cool, and I’m quite excited to talk about it!

So without further ado, here’s what we’ve learned from 1 week with Gyana.

1. One tool >> many tools

Every B2B SaaS tool reaches a point where they build a reporting or analytics feature. In our business, we regularly spend hours on dashboards in Intercom, Amplitude and Google Analytics.

I’d never appreciated what a mental overhead it is to switch between all these different interfaces all the time. The filters are implemented differently. The metrics have different meanings. The charts look different. It’s hard to compare them in a consistent way. It breaks your brain.

No doubt, for product analytics, Amplitude has a better interface than Gyana. But I have to learn that interface, along with 100 others, and they’re all subtley different. In Gyana, I need to know one UX and I can do anything with every dataset. So naturally (and this was a surprise to me) I just preferred just pulling my Amplitude data into Gyana, and doing everything there.

I can’t overstate what a relief this was, like an underlying stress I had that I didn’t even know was there. You just learn one abstraction (“all my data is a bunch of tables”) and it’s easy to work with.

One tool >> many tools

Tracking user behaviour with Intercom, Amplitude and GA in one dashboard

On top of that, the consistency of a single dashboard is great too. I structure them around specific questions, pull in the supporting data and get my answers.

2. Throw away your modelling layer

The idea for Gyana, as we currently know it, came from the experience of trying to be data-driven in a small startup.

BI tools assume your data is structured in a certain way - either a single table or “snowflake” schema, as it’s called. The process of building this is “data modelling” (or more recently “analytics engineering”), and it’s a full-time job. Usually done by the “IT team”. No good for a startup!

That’s why we built workflows. Workflows are designed to replace data modelling, in a visual interface. And they work.

But what’s interesting is that workflows are so quick (and dare I say it, fun) to build, that it’s easier for me to do my analysis end to end, in one go. The data modelling happens “just in time”.

Throw away your modelling layer

Just in time data modelling for Typeform responses.

That’s not to say that, over time, you wouldn’t want to build a library of re-usable workflows and put someone in charge of this (note to self: we should add features for this). So the section header is a bit over the top!

Rather than throwaway, perhaps we say “democratise” or “decentralise”. Modelling is an emergent thing, driven by people solving problems (and not the other way around). This is powerful.

The best thing I can say about workflows is that you don’t think about data modelling at all. While this pre-dates me, I imagine it’s similar to the feeling of using the first spreadsheet software and not realise you are programming.

3. Data integration is painless

Since you can’t do data analytics without data (!), integrations are a boring yet critical ingredient in Gyana. Our job is to make them as painless as possible.

For any data source, I typically start with a CSV import (or possibly a Google Sheet). Every tool nowadays has a CSV export option, and it takes ~30 seconds to get it analysis ready in Gyana. It’s a low friction way to validate an idea. And when I was ready to do more with that source, it’s time for a connector.

You can think of connectors as pipes connecting your data sources with Gyana, continually (and securely) syncing the latest data. They take time to setup initially (while you’re waiting, use a CSV!), but quickly pay off in saving manual work.

Back in 2020, we made the (then controversial, but increasingly sensible) decision to partner with Fivetran. Building reliable data integrations is the kind of thing that looks easy but is fiendishly difficult to do reliably and securely, and Fivetran has built a product that is trusted by the largest enterprises with 150+ prebuilt connectors. We think of them as the “AWS of data integrations”.

After our experiences of the 1.0, we put a lot of work into making the experience to setup and manage connectors really seamless, and for me that really paid off. Everything is smooth now, and getting redirected to Fivetran is a positive experience. It’s like when a website uses Stripe checkout. You know this is going to be fast, reliable and secure.

The other worry we had, was that the data from our connectors is too “raw” for a no code tool. We even thought about doing data modelling (bad idea: see above) to make it simpler.

But surprisingly, it’s not a problem at all. In most analyses, it was quick to find the 1-3 tables I needed and combine in a workflow. I didn’t need to understand the entire schema.

Integrations are reliable enough now that I’m excited to add more, just to see what I can find out. We’ll have all 150+ from Fivetran in the next or month. That will be fun.

Data integration is painless

Our mission control for data integrations - gotta catch ‘em all!

What comes next?

Gyana powers all our internal analytics. It’s already played key roles in our pricing decisions, product roadmap and user segmentation.

I’m really happy to see it works end to end. Are there features I’d like to add? Yes! Immediately after doing this, I wrote up a GitHub issue a task list of 50 items. (Sidenote: keep an eye on our public feedback portal if you’re interested in what’s coming next.)

But it’s now good enough that there’s no other tool I’d personally rather use. If you’d like to give it a go, join our waitlist and you’ll be on our free tier by the end of 2021.