3 Effective Ways NLP Supports Business Analytics

Hey there,

Today, we’ll explore how Natural Language Processing (NLP) is challenging the status quo of how we do data analytics.

If you've ever struggled to make sense of a huge spreadsheet or complex dataset, you're not alone. In fact, most business decision makers are overwhelmed by the sheer volume and complexity of modern data.

We're going to look at three ways NLP can help you overcome this problem.

Ready? Let’s go!

Note: The following article is adapted and reposted with permission from my original contribution to the AtScale blog. You can read the full version here.

The Need for NLP in Data Analytics

Today's businesses aren't just swimming in data - they're drowning in it. For most business users, consuming analytics often feels like trying to drink from a firehose. The situation isn't helped by the fact that data literacy remains a challenge for most organizations.

Honestly, there's no magic button that will turn all your complex data into easy-to-understand insights overnight-even (or more accurately, especially) with AI.

But that doesn't mean AI can't have an impact. While AI can't cure the cause of the problem, it can help make things a little better - much like a painkiller.

Data often comes in the form of numbers, codes, and formats that are perfect for computers, but a real headache for humans. That’s where NLP, a major subset of AI, comes in as a translator between decision makers and their data.

Let’s look at some examples:

Example 1: Answer Questions Instead of Reading Reports

Being able to simply ask a question to your data and get the answer right away, without having to look at a complex dashboard, has been the dream of data analyst for years. With the help of modern NLP, this dream has come much closer to reality. The change is happening in several stages:

Stage 1: Keyword-Based Search

This is the starting point. Here, you need to type in a specific question using the exact words that are present in your data set.

For example, you might ask: "What were the total sales in the US in 2019 and 2020?" To get an answer, your data model must include a "sales" table, a "location" field with the value "US", and a "year" column containing the values 2019 and 2020.

While this method works, it can be quite rigid and limiting.

Stage 2: Question & Answer-Based Search

In this next stage, NLP understands the context of your question, giving you more accurate and meaningful results. You're no longer restricted to the exact wording used in the data model.

So if you ask: "How much money did we make in the US in 2020 compared to 2019?", NLP can map your question to the right data fields and understand you're asking about sales revenue.

A great example of this technique is the Q&A tool in Power BI, which allows users to ask questions and create custom visualizations using natural language.

Stage 3: Conversational Analytics

This is the next frontier in NLP-based data analysis. Tools like Generative AI will help you in this stage to make the interaction with your data begins to feel more like a conversation than submitting a query.

You can ask follow-up questions based on the answers you get, and the system will understand the context of your conversation.

While we're still some way off from fully realizing this stage, we're making solid progress. There are many startups eagerly trying to make this work – Defog.ai, Avian.io, or OpenAI’s very own Code Interpreter for ChatGPT to just name a few.

Example 2: Text to Tables

The biggest portion of data growth isn’t happening in tidy, tabular data. Instead, it’s "unstructured", coming in the form of emails, documents, reports, social media posts, etc.

Traditional databases and spreadsheets struggle to make sense of all this unstructured data.

NLP can help to translate this messiness into a more structured form.

For example, take this earnings call transcript from a publicly traded company. If we want to keep track of all the data and information mentioned in this call, we could either painstakingly go through the transcript by hand, or we could use NLP to do the heavy lifting for us.

Feed the earnings call transcript into an NLP system like ChatGPT, ask it to extract the relevant, numerical information from the report for Q4, and output it in tabular form.

In seconds, you'll have a table containing all the key figures mentioned in the call, ready for analysis:

While this technology is still young and may need some double-checking, it's impressive how quickly and accurately it can extract information from an unstructured document. As the technology matures and becomes more widely available, business users will be able to process unstructured data at scale and make it readily available for downstream analysis.

Example 3: Auto-Explaining Data Visuals

A big part of data analysis involves deciphering charts to understand their meaning. While this is the daily bread and butter for data analysts, many business users still struggle to interpret anything more complex than a bar or pie (sigh) chart.

In these cases, business users need some context, a bit of commentary, or a narrative to make the information "click".

In many companies, data analysts handcraft PowerPoint slides containing numerous visuals and accompanying explanations to provide this clarity.

So, where does AI come in?

Modern NLP technology can automatically create captions or annotations for a given visualization and its underlying data. These annotations can highlight key takeaways, point out trends, and be edited to fit different audiences.

For example, Power BI features an AI-powered tool called Smart Narrative, and Tableau has a similar feature called Data Stories.

To try this out, open a sample Power BI report, create some charts and drag the "Smart Narrative" box onto the canvas. The tool then generates automated annotations for the entire report page.

These auto-captions adjust dynamically as the underlying data changes. For example, when you filter the data to only show data from a certain year, the text will automatically update to reflect insights based on the new data.

This is a fantastic way to make reports more accessible to business users and to communicate the most important insights from a report — all with much less manual work.

Sure, these features aren't perfect, but as they improve and we become more adept at formatting data for AI interpretation, we can expect to see tools like these continue to mature.

Conclusion

As we've seen, NLP technologies offer a powerful way to simplify and speed up business analytics workflows.

Data complexity won't disappear at the push of a button. But with the help of NLP, we're getting closer to making data more accessible, understandable, and ultimately actionable for everyone in the organization.

Many tools are readily available in popular data platforms, and new startups are popping up by the day. Keep an eye on these developments and take your own small steps.

The way you work with data could change forever.

And with that, this newsletter takes a 2-week summer break!

I hope you enjoyed today's issue.

Enjoy your summer and see you back in August!

Best,
Tobias

Want to learn more? Here are 3 ways I could help:

1. Read my book: Improve your AI/ML skills and apply them to real-world use cases with AI-Powered Business Intelligence (O'Reilly).

2. Book a meeting: Let's get to know each other in a coffee chat (use code FREEFLOW to waive the fees).

3. Follow me: I regularly share free content on LinkedIn and Twitter.