WebbMason Analytics has developed and implemented multiple analytics platforms and we have put most of the products reviewed in the Gartner Magic Quadrant for Data Science and Machine Learning Platforms and the Magic Quadrant for Analytics and Business Intelligence Platforms into production. From this experience, we know choosing the right technology for your advanced analytics and data science projects isn’t easy. Researching the options can make even the most technically competent person feel like they’ve fallen down the rabbit hole.
To get some clarity, many executives turn to Gartner Magic Quadrant reports. The problem is, the two most recent reports covering the advanced analytics and data science sector, have made choosing the right technology even more baffling owing to overlapping category definitions and a failure to consider platforms holistically.
There’s a problem in the definitions
Gartner offers a very broad-brush definition of analytics and business intelligence platforms.
“Modern analytics and business intelligence (BI) platforms are characterized by easy-to-use tools that support the full analytic workflow – from data preparation and ingestion to visual exploration and insight generation.”
The way the Gartner report defines data science and machine learning is problematic too.
“A data science and ML platform supports various skilled data scientists in multiple tasks across the data and analytics pipeline. These range from data ingestion, data preparation, interactive exploration and visualization and feature engineering to advanced modeling, testing and deployment.”
We find these definitions too vague to be useful. They overlap in too many ways to enable buyers to clearly differentiate between business intelligence, analytics and data science. For example, the Magic Quadrant for Analytics and Business Intelligence Platforms has criteria that states:
“Advanced analytics for citizen data scientists: Enables users to easily access advanced analytics capabilities that are self-contained within the platform, through menu-driven options or through the import and integration of externally developed models.”
But, similar criteria can be found in the data science and machine learning Magic Quadrant.
From our perspective, business intelligence technologies help businesses analyze and report on things that have already happened.
Analytics technologies are very different. They analyze data to explain what happened and why it occurred.
Data science technologies enable data modeling to determine why something occurred and to predict if it will happen again.
While each of these three capabilities is important, independently each forms just one layer of a highly complex stack of processes and tools necessary to support the full data and analytics pipeline.
Misleading qualification for a Magic Quadrant
The technologies reviewed are not apples for apples. This means that if you are looking for a clear indication of which technology you should purchase, this report isn’t very helpful. Some of the products offer comprehensive functionality and support multiple tasks across the data and analytics pipeline. Others are niche solutions that will need to be integrated with other technologies to form a larger analytics platform.
For example, let’s look at the comparison between Alteryx, Dataiku and KNIME in the Magic Quadrant for Data Science and Machine Learning Platforms report.
Both Alteryx and Dataiku support the end-to-end development of data and analytics, whereas KNIME only supports model development. While each of these three technologies is valuable, they support the analytics pipeline in very different ways. In our opinion, the Magic Quadrant fails at articulating this to buyers.
Despite offering a more niche solution, the report assesses KNIME to have a greater completeness of vision and it gives this product a significantly higher rating than the other two products. While KNIME may or may not outperform Dataiku and Alteryx, this score is misleading as the product does not have the ability to support the entire end-to-end data and analytics pipeline.
Is there a better way to evaluate advanced analytics technology?
In our experience, it is easier to understand where each technology fits if you take a holistic view of an analytics platform and its goals. We recommend looking at the continuum of technology needed for the end-to-end data and analytics workflow rather than trying to lock products into arbitrary definitions.
The technologies you need for your analytics platform are based on your company’s goals, legacy systems and the products necessary to solve your business challenges.
The platform architects at WebbMason Analytics organize technologies under five, high-level categories:
- ETL (Extract, Transform and Load)
- Data wrangling
- Data science and machine learning
- Multimodal workbenches
- Data visualization
But, even within our approach, technology products can live in multiple categories. For example, Tableau’s new data preparation solution places Tableau in both data visualization as well as data wrangling.
Organizing technologies into these categories allows us to recognize that some technologies support more tasks in the end-to-end data and analytics pipelines than others. This approach makes it easier for buyers to assess not only a technology’s category, but also the complete technology set that is required to support the end-to-end data and analytics pipeline. We believe this approach provides more clarity to buyers than the vague and overlapping definitions of the analytics and business intelligence Magic Quadrant and the data science and machine learning Magic Quadrant.
Choosing the right combination of technologies is never easy in the complex fields of advanced analytics. We think rather than getting tangled up in the difference between business intelligence, analytics and data science, you should start by looking at how each technology maps to the end-to-end analytics platform. Only then is it helpful to leverage Gartner’s evaluation of the capabilities of a product.