Alteryx Extends Its Reach
Alteryx Server, a self-service platform for data preparation, is now available via Microsoft's Azure cloud service. Analysts and data scientists now have a way to deploy, manage, and govern their Alteryx practices in Azure.
Alteryx Server, a self-service platform for data preparation, is now available via Microsoft's Azure cloud service. Alteryx already offers Designer, its self-service design environment, via Azure. The availability of Alteryx Server on Azure gives analysts and data scientists (and the organizations for which they work) a means to deploy, manage, and govern their self-service practices in the Azure environment. It also extends Alteryx' reach: Alteryx is already available for Amazon Web Services (AWS), after all.
Alteryx Server on Azure works exactly like Alteryx Server on premises. Analysts, data scientists, and other users build data workflows in Alteryx Designer and publish them to Alteryx Server, which takes care of deploying and scheduling them, as well as running and monitoring them. (Did they run as scheduled? Complete as scheduled? Produce valid outputs?) Existing customers can transfer unused Alteryx Server licenses to Azure.
Alteryx Server and Self-Service Analysis
Alteryx Server is what transforms Alteryx from an end-user-oriented product to a platform for self-service data analysis, argues vice president of product marketing Bob Laurent.
"The hardcore data [transformation] work is actually done in an Alteryx workflow inside of our Alteryx Server product. A data scientist would build out a workflow in Alteryx Designer [and] push it to Alteryx Server, and then Alteryx Server would run this workflow every night and push it out into [for example] a Tableau file," he explains, stressing that customers are increasingly using Alteryx Server to support operational -- i.e., repeatable, ongoing -- use cases.
"What makes Alteryx a platform as opposed to a tool is that there's an SDK, there's Alteryx Server. This allows people to build things on top."
Alteryx -- a 20-year-old company -- is enjoying new relevance thanks to the popularity of self-service data analysis and visualization offerings such as Tableau. Tableau-toting analysts and data scientists snapped up self-service data prep offerings from Alteryx, Paxata, Trifacta, and others to augment Tableau's limited data blending facilities.
The upshot is that Alteryx has become a popular option for users of self-service tools such as Sense (from Qlik), Spotfire (from TIBCO), and even Microsoft's venerable Excel spreadsheet. Data scientists and other would-be discoverers use Alteryx to engineer data so it can be consumed in their preferred self-service data analysis environments.
One knock against self-service is that it's difficult to scale -- so far, the use of self-service discovery, data preparation, and analytics tools hasn't translated into especially high levels of reuse and repeatability. It's one thing to support a small cadre of self-service discovery or self-service data prep users; most such tools were, after all, designed for analysts, data scientists, and other power users. It's something else to scale a self-service discovery or data prep practice to hundreds, thousands, or tens of thousands of information consumers.
Alteryx itself is something of a case study in the growing pains of self-service. The company says Alteryx Server provides the underpinnings for an enterprise data analysis platform. This is true to the extent that Alteryx Server gives consumers a central context in which to publish and operationalize workflows such that they can be made available to other consumers or called/invoked (via the Alteryx SDK) by third-party systems, applications, and services.
At this point, however, it isn't possible to score (i.e., run) an Alteryx model in, for example, an Oracle or Teradata database engine. Nor is it possible to export workflows as Java, C, or C++ code that can be consumed by third-party systems, applications, or services.
Using Alteryx Server, however, it is possible to export the output of an Alteryx workflow -- i.e., the data set itself -- to Oracle or Teradata. In this way, a prepared data set or the results of a predictive analysis can be operationalized, Laurent says. "The basic way you would consume this would be to consume the actual data itself," he explains. "The other thing you can do is you can package it up as a model and then publish it into the Alteryx Analytics Gallery ... where it would be available to everybody else inside the organization.
"When you're publishing it that way, you're not publishing it in code; you're publishing it as an Alteryx macro so that other Alteryx users can install it and use it in their workflows."
Alteryx is "looking at" options such as Java or C/C++ code export, Laurent says. Although he didn't say it, support for code export is probably a matter of when, not if. To its credit, Alteryx seems reasonably responsive to the needs of its user base: in response to demands for better R support, for example, it introduced R Tool, a facility data scientists can use to run R code in the Alteryx environment. (It's said to be considering a similar facility for Python.)
As Alteryx (like Tableau) attempts to expand beyond its existing enterprise beachheads (e.g., use among business analysts, statisticians, and data scientists), new user constituencies, such as enterprise IT organizations, will demand more and varied export facilities, deployment or run-time options, and so on. There's every reason to believe it will work to accommodate them.
Not Just a Data Prep Platform
In addition to data prep, Alteryx also provides an advanced analytics platform for performing data analysis.
Laurent cites the Alteryx Analytics Gallery, an online resource for data preparation workflows, predictive models, analytical functions, algorithms, and other assets. Consumers can browse the Analytics Gallery to select prebuilt workflows for supervised machine learning techniques, including algorithms for prediction, classification, affinity, and regression. (Alteryx also supports R and Python algorithms via prebuilt tools, which are exposed as shortcut icons.)
The company has a partnership with machine learning specialist DataRobot, too. This gives joint customers easy access to best-in-class machine learning technology from within Alteryx itself. "There's a new DataRobot tool inside of Alteryx that allows you to make [the preparation of a data set for data analysis] a whole lot easier, so you can take the result of a data preparation and blending step and then put your data into DataRobot, pick your target variable, and it will use the DataRobot API to ... automate a lot of that," he explains.
The Alteryx-DataRobot partnership is an example of what could be called data analysis mutualism: on its own, DataRobot provides few built-in data prep tools. Alteryx does incorporate machine learning tools and concepts -- decision trees, among others -- and also uses machine learning algorithms when necessary to assist with analyzing, preparing, and transforming data for analysis. However, this usage is tactical, not strategic; at this point, machine learning is not core to Alteryx's strategy.
Alteryx Here, There, and Everywhere
Alteryx is already available for Amazon's AWS cloud platform -- even down to the same bring-your-own-license option it supports with Microsoft's Azure platform.
In many ways, the cloud is an ideal environment for both data preparation and analysis: a not-insignificant proportion of the information that analysts and data scientists work with lives in AWS, Azure, and other cloud platforms, as well as in the social Web. The elastic, RESTful cloud is as good a place as any to consolidate and explore data, as well as to prepare and engineer it for analysis.