Analytics over the last few years has taken off. But with this rapid growth, the nature of analytics has changed, and as a result, we desperately need new approaches to doing our work. In this article, we'll discuss the idea of analytics frameworks, then discuss why such models fall short. Finally, we'll explore what you might focus on instead to be a better data analyst.

What is an analytics framework?

If you don't know, then count yourself lucky. But analytics frameworks are structured means for thinking about and doing data analysis. They tend to be prevalent in academic institutions, scrambling the develop analytics programs to capitalize on the rising prevalence of analytics needs in industry.

Why do analytics frameworks fall short?

As data quality, analytics engineering, data engineering become proper domains, analytics has changed from a full stack domain into an interfacial discipline -- one that requires tighter business alignment, stronger business acumen, and convincing delivery. Consequently, frameworks built even just a few years ago often emphasize the wrong things. While these frameworks help expose you to some of the ways companies are able to solve problems around working with and exploiting data, they're often overly dogmatic and overly exhaustive, diverting attention from what really matters in modern times: leveraging data towards business value.

In any case, if you've tried to put any frameworks into practice, you'll know that it's rarely a drop-in solution. They're certainly descriptive and perhaps useful to form your own opinions, but the belief that they are all you need will only degrade your ability to think from first principles, and cast you in your stakeholders' eyes as an overly rigid thinker, unable to connect what they're doing to what really matters.

A popular framework, for instance, is the CRISP-DM framework (Cross Industry Standard Process for Data Mining). This framework outlines 6 steps you should always take when starting a data science or analytics project:

1.  Business understanding – What does the business need?
2.  Data understanding – What data do we have / need? Is it clean?
3.  Data preparation – How do we organize the data for modeling?
4.  Modeling – What modeling techniques should we apply?
5.  Evaluation – Which model best meets the business objectives?
6.  Deployment – How do stakeholders access the results?

Source: https://www.datascience-pm.com/crisp-dm-2/

But in being so formulaic, in pursuing exhaustiveness here, the framework overemphasizes the obvious, diverting attention from what matters. "Data understanding" of course is a requirement for doing analytics work . Of course, you need to prepare the data. Of course, you should think about the techniques to use. And of course, you should think about how the results are going to be used and whether they address business needs. But there's a muddling of technical execution guidelines and business-related interfacial best practices. You end up reading this list, coming away with little more than a sense that analytics requires thoughtfulness every step of the way.

At the end of the day, the framework can be broken into three main points:

1. Business value: you should ensure your work is providing business value.
2. Technical correctness: you should be careful in doing the analysis, to ensure your work is done correctly.
3. Accessibility: you should ensure your work is accessible and usable by others.

What should we do instead?

So what to do instead? If you haven't entered the analytics world yet, don't worry too much -- you'll figure the shortcomings of what you've learned, how to apply what you've learned, etc. soon enough. That said, of those three major buckets of focus, I'd encourage you to think about business value and accessibility more often than technical correctness.

As analysts, we're inclined by nature to want to explore data. And as a result, no one needs to tell you that you need to clean data, that you need to careful consider your approach, that there are gotchas that you should be careful of when working with data. But in the dumpster diving, it's common to lose sight of what you're doing. What's the objective? And how will your work be used? How will you ensure that your work has maximal business value? At the end of the day, maximizing business value comes down to two things: ensure your work actually addresses business concerns, and ensure you and your team have a system for ensuring that your work gets seen.

How Hyperquery can help

This is a corporate blog, after all, so let me leave you with a quick pitch. Hyperquery is a data notebook not only meant to make the technical parts of your work delightful, but nudge you towards best practices around the delivery and alignment aspects of your work.

We get there in two ways:
1. Accessibility: we are a notebook, but we make make the writing and collaboration experience as delightful as the technical aspects, ensuring that stakeholders are as comfortable here as you are. We've observed time and time again that with Hyperquery, stakeholder adoption is unprecedented, meaning your work will be more tightly coupled with the needs of the business, from conception to delivery.
2. Centralization: unlike other BI or notebooking solutions, we ensure everything is discoverable after creation. With a hierarchical nesting structure, page linking, blazingly fast search, and a homepage that surfaces recent work, we make it easy for you to not only organize your work, but easily find the work of others.

Use Hyperquery for free at hyperquery.ai.