It’s time for some Old Gartner’s nostalgia again.
- Late in February, Gartner published their Magic Quadrants for Data Science and Machine Learning, and Business Intelligence. Marketing folk look forward to these annual publications, rather as country folk anticipate the Old Farmer’s Almanac appearing in their Christmas stockings.
- We in the country enjoy the Almanac’s broad brush predictions and obscure methodology: we derive our weather forecasts from a secret formula that was devised by the founder of this Almanac, Robert B. Thomas, in 1792. Thomas believed that weather on Earth was influenced by sunspots …” We appreciate their quirky, if somewhat dated advice: use buttermilk to whiten your table linens.
- Old Gartner’s affords similar pleasures with an ache of nostalgia. Instead of sunspots we have strategic planning assumptions and inclusion criteria which, if you listen to vendors, are equally obscure. And Old Gartner’s also includes helpful advice echoing a previous century’s concerns: complex data modeling such as multifact table models must be created either outside of Tableau in a data warehouse or via self-service data preparation partners. Those were the days!
- But really, I am not here to pick apart Gartner’s work. The ever-incisive Jen Underwood has her own doubts, more precisely expressed than mine. Yet I’m sure Jen agrees with me that the Gartner analysts we know are sedulous in their work and put great effort into it.
- What caught my eye this year was a short, but significant, comment from DataRobot, whose automated machine learning platform is nicely instrumented and practical. You can read for yourself, Why We Aren’t Included in the 2018 Gartner Magic Quadrant for Data Science and Machine-Learning Platforms but here is the important phrase: Most Magic Quadrants help you identify one supreme leader for purchase, but in this case, pitting complementary tools against each other is not consistent with how data science pipelines are built, and fuels vendor marketing hype, confusing buyers.
- The Quadrants are certainly not intended to identify clear leaders, although that’s often how they are used. Yet DataRobot’s main objection is sound. In the field of data science, as practiced today, tools are extensively compounded to enhance each others’ capabilities.
- Data scientists are omnivorous, choosing algorithms over platforms and techniques over tools. This is an important distinction between data scientists and business intelligence analysts.
- In BI, some analysts are greatly committed to their tools of choice: Tableau Zen Masters, Qlik Luminaries and so on; their view of business problems and analytic techniques are refracted by the applications they use and a fascination with their surfaces.
- Data scientists, in contrast, are not bound by a data platform, their tools, or the algorithms they run. It really doesn’t make sense to say I’m an Alteryx data scientist in the way one may say, I’m a Tableau Zen Master.
- At one time, you might indeed see yourself as primarily a SAS or SPSS user. But that has passed with the success of R, the cloud machine learning platforms and the seemingly endless Apache projects for handling data. The pool of data science tools is floored with many-colored pebbles.
- Today, if a researcher at Microsoft finds Google’s TensorFlow compelling for a project, they need not look on in watchful hunger – that’s where they will do their work. If a data scientist in an enterprise with RapidMiner widely installed feels it is better to run and test ensemble models with Data Robot rather than AutoModel, they will do routinely.
- To be fair, in both Business Intelligence and Data Science, Old Gartner’s is not aimed at such users, but rather at decision makers whose view of the world is a little out of time, and perhaps includes knowing not only that MicroStrategy is a challenger, but that March 7th is the best day to make sauerkraut and to plant below-ground crops.
Whether for data science or analytics, there is no one-size-fits-all solution for your business. Different teams have varying capabilities, diverse needs and ultimately different objectives. The TreeHive Analytics Gameplan process can help you to identify well-scoped, achievable steps towards your analytic ambitions. You can read more about it here.