Gartner BI Bake-Off 2022: AI-Powered Decision Intelligence Applied to UN Sustainability Goals

Gartner BI Bake-Offs offer prospective buyers side-by-side comparisons of analytics vendors based on scripted demos and a common data set in a controlled setting. This year’s dataset was from the United Nation’s (UN) Sustainability Development Goals (SDG). Below is the Tellius approach to analyzing UN SDG #8 (Decent Work & Economic Growth) in video and write-up form. Check it out!

What is Tellius? What are we analyzing?

What is Tellius? Tellius is an AI-Powered Decision Intelligence platform designed to help business and analytics teams expedite their journeys from data to decisions by combining the best of BI and ML/AI for rapid, data-driven driven answers to what, why, and how-type questions.

What are we analyzing? We’re going to use Tellius to analyze the United Nations Sustainable Development Goals. There are 17 of these goals, but we decided to focus on goal number eight, which is promoting inclusive growth and productivity as well as decent work opportunities.

Key Findings

• Over the past decade, global GDP growth was fairly stable, with a more extreme drop in growth towards the end of the 2010s. This drop in global GDP for the world was most heavily felt by the subregion of Latin American and the Caribbean.
• Several of the KPIs measured around promoting inclusive and sustainable economic growth and employment for all are highly correlated. An interesting relationship to explore further is the inverse correlation between financial participation and employment rate. This may indicate that countries with a high employment rate may have less knowledge-based employment and rely on skills-based employment, leading to less modern infrastructure for high financial participation.
• Yemen, amongst other low-income countries, stands out as a key country to focus on for economic growth opportunities. It has the lowest GDP growth on average and one of the lowest financial participation rates. There appears to be a lack of gender parity in the workforce, as women above the age of 16 with primary education or less are 3.3x more likely to fall into the poorest income classification.

Ad-hoc Exploration of GDP Changes in Natural Language

How did we get to these insights? We start our analysis using natural language search, a strength of the Tellius platform as the only provider that has a true natural language search experience (compared to simple keyword search). We ask the question ‘what is GDP by year for the world’ and the platform automatically interprets the question, relates it back to the data, and visualizes it — all in one step.

We can see a massive drop in 2020, when COVID-19 emerged. Comparing the United States to the world, we can see that in the past three years, the US has finally gotten back to the point of eclipsing the average world GDP growth.

As an analyst, we’re able to answer these questions and get quick visualizations. But oftentimes where traditional BI tool fall short is the ability to answer more complicated questions like why did that dip exist or where did that dip in GDP growth most significantly impact the world? 

GDP Change Root Cause Analysis via Automated Insights

In Tellius, exploring these more challenging questions is a click away via the ‘understand why’ button:

…which kicks off an automated insight, specifically a Trend insight. This is going to provide a summary view of what drove that change in GDP growth in the year 2020 compared to 2019. We can see the summary view of the actual and relative decline — in this case for GDP growth — as well as top change contributors — in this case, areas around the world — that contributed most to that growth.

In this case, the decrease in GDP in 2020 was most significant in the sub-region of Latin America and the Caribbean. If we further explore that segment, we can see that the actual drop in GDP was 10.78 compared to the overall aggregate drop of 7.66 and can also see that visualized compared to every other sub-region on the right-hand side.

All these rank orderings and insights were automatically created when a user hits that button on a search graph to generate this type of insight. From here, we can further explore the Latin America segment to look at what are the next level of contributing factors to the Latin America and Caribbean subregions specifically.

We see that the income group is a classification of a country-level relationship, compared to all the other countries in the world. In this case, upper-middle income countries had a very significant drop in GDP, in particular the country of Peru.

To quickly recap, we’ve seen all of our different data points and attributes scanned in one single click of a button, rather than having to comb through dashboards or write custom scripts — in an ad-hoc fashion using the Tellius Search and Insights capabilities.

Automating KPI Tracking and Predictive Analytics

Tellius also has a built-in feed capability to intelligently track your KPIs. In this case, if we wanted to track GDP growth on a yearly basis, simply specify those two line items in the Tellius feed system:

…then it builds out a forecast for that metric and any time actuals fall outside of the confidence intervals for that forecast, Tellius will track it, run a trends-based insight, and email the user the results. So it doesn’t even require an analyst to come to the system to ask the question. Proactive insights are delivered to you.

From here, you’ll see that by selecting details, it takes us to another example of that trend insight. But of course, in this case, customized on the time periods and the data points that we’re looking at in that feed customization layer. From here, we’re able to go from the ad hoc question answering to the proactive types of question asking. 

You can also look at Tellius to help provide future-looking types of analysis. Predictive analytics using out-of-the-box Tellius machine learning modeling capabilities.

We can create supervised ML models such as classification as well as unsupervised ML models such as clustering as well as time series regression for forecasting. All these different model types can be saved and/or embedded and the predictions are always going to be tracked back for users to compare against the actuals to iterate and improve model performance over time.

Bringing in More Data: Financial Participation

Now let’s consider financial participation and employment rates from some outside data which we easily join inside Tellius. It could be interesting to compare how these two metrics relate in various countries around the world. To do that, we just start asking questions again.

We see a really interesting opposite trend when comparing financial participation rate vs employment rate globally — the countries that have the highest financial participation often have much lower employment rates, and vice versa, likely indicating there’s a significance in what types of jobs or industries are very popular in the countries with high employment rate and low financial participation.

Now let’s add income group to that segment and break it out, on the fly:

There is a large segment of countries at the top from the high-income cluster in blue vs the bottom of the low-income countries. We can see a pretty big outlier here on the graph. In this case that low-income point here, Yemen is a country that has very low finance participation and much lower GDP growth on average compared to every other country in the world.

If we want to take this analysis and save it, we can either add it to a vizpad by selecting an existing one as well as download the CSV content or share it with other people outside the platform or even download the visualizations as a native powerpoint integration.

We’ve uncovered a couple of different insights about how these other metrics relate to GDP and growth around the world. We can more easily visualize that on a map rather than looking at that as a bar graph, just by changing the type of question our system will automatically pick up on those contents and change that visualization. 

But we saw that interesting outlier in Yemen. Let’s dive back into data and show you how to get data into Tellius and focus specifically on Yemen.

We can load data from ad hoc files like CSV or Excel or common file storage systems and data warehouses and we can leverage them in multiple sources in Tellius all in one location. In this case, we’re just going to load in an Excel file that has more information about Yemen’s financial participation metrics through survey data that the UN has provided.

We’ll load that data file and that’ll take us here to Tellius’s data preparation layer where we can see a snapshot of the data or Tellius has automatically ingested the column names, the data types, and a few other different things to help make it very easy and seamless to get data into the platform. We can also further explore the summary statistic of our data by leveraging the out-of-the-box features that we have.

Let’s explore what the breakdown is for these survey participants or who actually has a financial account in a modern system. In this case, we can see that a very low percentage of people actually have accounts in Yemen.

We can do the same thing with household income to look at the breakdown of who submitted the survey has fallen into the different segments richer or poorer around the country. 

Tellius also have a pretty expansive data preparation layer where we can complete a lot of no-code or low-code data transformations, for example, changing data types, finding and replacing different string values, upper and lower casing text, etc. All these common data preparation can be done in a point-and-click fashion in Tellius.

We also have the ability to create more advanced types of transformations using SQL and Python code, all integrated seamlessly into your pipeline to leverage the more advanced types of transformations that we might need to create and benefit from what your analysts already use today using open source languages in Python & SQL — not a tool-specific language. 

What is driving low financial inclusion in Yemen?

The last step in preparing data for Tellius is creating a business view. You can think of this like a data model. We’re defining relationships between different tables in a drag and drop process. Also, we’re able to create your own custom KPIs or calculated columns to help derive more customization from what your data offers out of the box.

Now from there, we want to focus back on that Yemen dataset. We have survey data, trying to understand what’s driving the people around the country who either don’t have an account or maybe want to focus on the financial opportunities and understand what’s driving the poorest segment of people in Yemen, other sets of characteristics that help us define that. And so to answer that type of more complex question, we’ll leverage a different type of insight called a key driver insight. We’ve already touched on a Trend insight, we have a third one called a Comparison insight, but for right now we’ll focus on a key driver insight to better understand what drives the poorest 20% segment of the population, what are the characteristics of those people based on all the attributes in our data? The output of that trend-based insight is now going to show me this result here. 

On the left-hand side, I can see what features actually contribute to answering that question. In this case, saved in the past year, the education level, gender, and age are the most important attributes to answering that question. And on the right-hand side, we’re going to look at the combination of those attributes to build out specific segments. So in this case, all of the people in this survey that have saved in the past year have not completed more than primary education are female, above the age of 16, are 3.3X TIMES MORE LIKELY to fall into that poorest income segment.

There is clearly a disparity around the opportunities that are available around the country of Yemen, definitely something worth focusing a little bit more attention on for the UN in both Yemen and other countries that have similar types of attributes. 

Sharing the results

Now, as an end user, we can very easily share this content inside the platform as a dashboard. 

But we also have the ability for an analyst to create dashboards and insights without having to really know what types of questions to ask. That’s possible through our Quickstart feature. This feature solves the blank slate problem when you start using a new tool whereby you don’t know where to start. With Quickstart, analysts simply select their dataset, the key country the dimensions and measures that they want to look at, and from there, Tellius builds out dashboards and insights without the user having to do anything. 

Tellius also has the ability to embed content into other applications outside of Tellius.

Conclusion / Next Steps

We’ve walked through the key functions of Tellius and spotted some interesting trends in the UN dataset. Aside from analyzing and informing sustainability practices, Tellius can also be used in a wide variety of other industries such as Pharmaceuticals/Life Sciences, Financial Services, Ecommerce, Healthcare, Communications, and Insurance, to name a few. Expedite your data to decision journey today with Tellius!