There’s a lot of talk of "data-driven decision-making", "citizen data-scientists", etc. This isn’t new either – this has been happening for a while now. But companies that went down that route struggled hard. The main tools available in the community are Python, R and associated packages. But it is a bit much to expect business users to pick up coding. There are tools like DataRobot and H2O, but most business users we spoke to found them super intimidating.
So, Dipayan, Chandra and I wanted to build something that can help business users. Someplace where one could just select a business problem, point at the data, and set some basic context. And a machine automatically takes care of the rest. The user would then focus completely on understanding the insights, and thinking about "what's next?"
Daisho is No-Code Data Science for Business Users.
Daisho re-orients data science to a business user (or a consumer) point-of-view – leading off with problem statements instead of ML algorithms. A user first selects a problem statement (we call them recipes), and Daisho leads her through the flow – making algorithmic choices automatically based on recipe, data, and user input.
As most data scientists will tell you, the key parts to any analysis are
- Data Cleanup – otherwise its going to be Garbage-In-Garbage-Out
- Data Prep / Feature Engineering – the key will always be in the hypotheses one builds and explores.
Daisho handles both these parts! We have an extensive Data Cleanup suite which will automatically recommend the best way to cleanup the data. And feature engineering is Daisho’s core IP. Our SignalFactory automatically builds, tests and shortlists 1000s of hypotheses without any human intervention.
But Analysis isn’t enough!
Not just that: we realize that just analysis isn’t enough – it needs to be communicated, justified, and acted upon too. Daisho comes with pre-built insights which make communication easier. And an Action Engine which you can set up for predictions – real-time, batch, whichever way you want. That way, you ensure your analysis doesn’t go waste.
We start off today with Driver Analysis – both for continuous outcomes (regression), and binary outcomes (classification). You get tons of insights out of the box already here – all you do is point at the data, and choose your outcome. You can also build predictive models to take advantage of the SignalFactory, and use the Action Engine.
And we have a ton more on the roadmap: time-series analysis, forecasting, clustering, impact analysis, text mining and much more. We shall keep you updated.
You can sign up for a free trial here.