Looker is all about helping businesses get more value through the smarter use of data. In a lot of cases, that’s about putting the right data in the right decision-maker’s hands at the right time.
But in addition to business users who need access to actionable intel, many companies have invested in data scientists who use cutting edge algorithms to extract meaning from large amounts of data. And these people need access to the right data, too.
You’d think they’d have a stable full of amazing tools for their work, but the reality is that while the tools for building machine learning models are fantastic (and always getting better), the tools for the other parts of a data scientist’s work -- preparing the data and then operationalizing their findings -- are typically lacking.
So that’s why we’re announcing a slew of new capabilities in Looker to make those parts of a data scientist’s job--data prep, data wrangling, data cleansing, data visualization, data presentation, and data action--easier.
If you’re a data scientist, this should make you happy. After all, wouldn’t you rather spend your time doing the interesting work of building and tuning models rather than recleaning the same data set?
And if you’re somebody who employs data scientists, this should make you really happy. Would you rather your data scientists’ (expensive) hours go to doing data custodial work? Or would you rather they be able to access clean, correct data from across your organization quickly and easily?
Data scientists at places like Avant and Stack Overflow are already getting huge value from Looker, as their stories below illustrate.
Helping developers go further
Julia Silge is the lead data scientist at Stack Overflow, the most popular developer knowledge-share in the world. The platform hosts 50 million visitors a month, and these visitors ask a question on the platform every 12 seconds. It’s Julia’s responsibility, as lead Data Scientist, to surface insights from this massive amount of information and help her team to better prioritize where to spend their energy.
Concretely, one challenge Stack Overflow faced was building a sustainable Talent Solutions business on top of their “free” community platform. “We needed to understand what makes clients more likely to have a good experience, come back, and spend with us again,” she explained. To do this, Julia had to gather and stitch together multiple types of data from various sources, test out machine learning statistical models on that dataset, and put the predictive model output in relevant context for her stakeholders to better make their business decisions.
The work was hugely impactful: Stack Overflow nearly doubled their renewal rate in only one year. And Looker’s scalable and efficient architecture played a key role by helping Julia shrink the most time-consuming and manual part of her job from many hours of work down to minutes, allowing her to focus more on solving high-value data science problems.
Enabling opportunity through better access to capital
Avant is constantly striving to better match its customers with the best offering. And as a company that has democratized analytics for every employee, Avant has become one of our go-to examples for what an instrumented workplace and universal data access can achieve.
But the next step in their data culture’s evolution was to democratize access to advanced analytic techniques. To do this, they introduced the machine-learning-as-a-service rocketship known as DataRobot. And the combination of Looker and DataRobot has driven tangible and exciting outcomes for Avant and their data culture:
Our advanced analytics workflows are at least 5x more efficient with Looker. Before, it would take a data scientist a whole week to get the answer they need... preparing the data, feeding it into the predictive model, and outputting the results for stakeholders. Now, thanks to Looker, that process takes less than a day. When we combine Looker with DataRobot, our analysts can easily run 10 models and choose the winner all in one day and quickly move on to the next problem.
~ Charles W, VP Product, Avant
New capabilities to make data science more efficient and valuable
One of the biggest goals of our new tools aimed at data scientists is making sure that it’s dead simple for them to integrate Looker into their existing workflows. That’s why we’re focused on integrating into their existing best-of-breed data science tools with some exciting new technology partnerships. Our TensorFlow integration allows data scientist professionals to more easily take advantage of Google-scale machine learning on their governed and reliable Looker datasets. For companies that need more support, our partnership with Big Squid helps the “citizen data scientist” run their Looker data through the sophisticated analytical techniques of a pro, in only a few clicks.
We’re also rolling out four new capabilities that help data teams augment Looker analytics with artificial intelligence and machine learning on large datasets: Streaming Results, an R SDK , Python Connections and Merged Results. Now, you can more seamlessly send your entire set of data directly to your data science tool of choice (e.g. RStudio, Jupyter Notebooks, etc.) for advanced statistical analysis. Then, as you iterate and optimize your statistical model, you can operationalize outputs in relevant, actionable context alongside other business metrics, federating results from any connected database all in the front-end of Looker.
With these developments, both professional and citizen data scientists can focus their time on higher value thinking for greater impact.
As organizations build out and evolve their use of data, Looker is focused on helping more people make more impact in more ways. Efficient, high value data workflows mean tighter feedback loops for testing, learning and iterating on models, accelerating the value organizations derive. Data cultures like those at Stack Overflow and Avant are at the leading edge, and we’re excited to see more customers take full advantage of Looker in their data science workflows going forward.