Metalytics Blog

Metalytics Blog

This post serves as an overview of why we exist and our vision for a better model of data product delivery.

Ready to create impactful Data Products that consumers love? Join the Waitlist

Data is important

The best orgs in the world use data to set goals and evaluate decisions. They make it their competitive advantage.

If you can’t measure it, you can’t improve it. If you’re not improving, you’re dead.

Metrics trees, also known as KPI trees, show the relationship between the key metrics in a business, from the overall goal down to all the metrics that influence it.

We’ve found that choosing the right metrics and focusing on the team rather than the individual can motivate the entire team to consistently deliver great results.

Data drives accountability, transparency, and trust. Data cuts through opinions, charisma, and bias. It is the great equalizer, and the best chance we have at improving our teams, our organizations, and our world.

Data must be of high quality to be useful...

When data is unreliable, analytics efforts are at best ineffective, and at worst actively hurting the business.

An organization can have the best ML models, the greatest developer experience, and a streamlined production process– but throw garbage data into this pristine system, and you will still get garbage results.

...but implementing data quality is hard work...

There's a wide variety of tools in the quality toolkit, and each one takes focus and time to implement well:

By the way, that's not just true of quality in data - all reliable, fault-tolerant, high-quality software is expensive.

...and quality does not necessarily imply value.

The most sophisticated quality measures in the world cannot guarantee that a dataset solves a real problem or delivers business value.

It’s important to “build the thing right” and engineer for quality, but the best teams are also digging deep to ensure they are “building the right thing”.

Failing to solve the right problem can sap 50% or more of a data team’s time...

Pre-Production Analytics

Pre-Production Analytics

Unused Models

Unused Models

Post-Production Complexity

Post-Production Complexity

I’ve had every single one of these problems for the last ten years…these things are the bane of my existence on a weekly basis

...so the best teams are bringing a Product mindset to data.

There is a better way to build and run a data organization: run it as if you were building a Data Product and all of your colleagues are your customers.

Not enough effort has been made to understand the people using data products, what they want to get done, and the broader context in which they operate.

Always ask your stakeholders what they actually look at to make decisions. You might think and hope it's the gorgeous looker dashboard based on beautifully kimball-model DBT, but oftentimes it's not...it's on us as data leaders to look at what people are actually using to make decisions

Good Data Product Tooling should…

Automate 80% of data product discovery

  • Which are the most complex queries and analyses running on top of my production models?
    • Who is running them? Let's talk to them and find out why.
  • Which queries are using pre-production or lower quality data?
    • How often is that data reaching stakeholders? What can we learn from consumers or pre-production data?
  • What % of analyst and data science work is using production, high quality data?
    • How is this number trending over time?
    • This is the number one metric for understanding the progress and velocity of a data modernization initiative.
  • Which models have low or no adoption?
    • Do we need to educate consumers? Adjust the models? Deprecate them?

Automate 80% of data product delivery

  • What steps can I take to move complexity out of notebooks and sql scripts and into quality, monitored pipelines?
    • What would an initial draft of that model look like?
    • What existing or similar models might serve this use case?
  • What pre-production queries are good candidates to migrate into high-quality production models?
    • What would an initial draft of that model look like?
    • What existing or similar models might serve this use case?
  • Which models can potentially be pruned automatically?

Ready to create impactful Data Products that consumers love?

Join the Waitlist

← Back to home