Data science as a service, or DSaaS, has been out of the news for a while, but it has recently resurfaced as a hot subject. AWS and Change Healthcare announced a collaboration in March to include DSaaS for health analytics.
Comviva, a mobility solutions company, launched its own DSaaS offering to assist telecom providers with marketing the same month. In February, data services company Calligo purchased data analytics company Decisive Data to extend its DSaaS offerings.
With
Professional services gaining traction, it's critical for businesses to understand whether they should invest in the service and why they should.
What is data science as a business, and how does it work?
Data science as a service, according to Anand Rao, partner and global AI chief at PwC, is the outsourcing of data science activities to an external provider.
He said, "The client provides the data, and the DSaaS provides the client with the insights from the data."
He clarified that it's especially useful for temporary jobs, peak workloads, and standardised tasks like running research on monthly or quarterly reports.
One of the reasons you don't hear much about data science as a service these days is that the word can refer to a variety of items. It ranges from analytics tools integrated into common SaaS systems like Salesforce to specialist vendors that sell prebuilt models for unique business apps that they can configure and manage for customers to regular consulting deployments.
Regardless of the flavour, DSaaS provides a lot of value to businesses, whether or not they have an in-house data science team.
DSaaS as AI-as-a-Service (AI-as-a-Service
According to Kjell Carlsson, principal analyst at Forrester Research, more and more business applications have analytics and AI functionality built right in.
You're less likely to use data science as a service if you have a lot of different systems to link.rnPrincipal analyst at Forrester Research, Kjell Carlsson
He believes that a company's own data science team might create something more personalised than these apps.
"However, getting those models into the hands of the end user is extremely difficult," he said. "Those SaaS business applications are already in use by the end user, and the model is already embedded in their workflow."
According to Carlsson, the advantages of getting an embedded model that's already there and simple to use outweigh the possible benefits of creating a bespoke model in several instances.
If the data is spread across several systems, however, the size of the integration challenge can necessitate an in-house effort.
"The more systems you have to link, the less likely you are to use data science as a service," Carlsson said.
Platforms for business intelligence based on DSaaS
Augmented BI and augmented analytics are becoming more popular in general-purpose business intelligence suites provided as a service. Tableau Online, for example, provides cloud-based self-service analytics. Power BI is available as a service from Microsoft, and IBM's analytics tools are also available as a service.
Dave Costenaro, the chief data officer at Ability, an AI-powered support desk, uses DSaaS options including Tableau for analytics and AWS for data storage.
"Data processing and storage infrastructure can be conveniently outsourced to a variety of cloud database vendors," he said.
As AI development platforms, DSaaS
According to Carlsson, vendors such as C3.ai provide components and prebuilt AI modules that businesses can use to create their own predictive applications.
Even if an organisation has an internal data science team, using an external data science platform makes sense because it allows for more flexibility in scaling models up and down as required and easily spinning up test environments. It can also save money on capital expenditures or licencing fees, and the provider is in charge of keeping the infrastructure up to date.
According to Hugh Burgin, U.S. data and AI leader of the Microsoft Services Group at EY, using a DSaaS platform will also provide access to the vendor's proprietary data science algorithms.
However, there are several drawbacks.
"It can also feel like a black box in a hosted model offered by a provider, where the company has little impact and insight into how the data operations and data science functions," he said. "Executive buy-in on using external resources could be a problem for certain businesses."
DSaaS as a consultancy service
Professional services firms are also happy to build AI models from the ground up. Many businesses with in-house data science departments may choose to hire a consulting company to help them with specialised projects.
"Even if you have a data scientist on staff," Carlsson said, "a data scientist who knows how to create vision models is uncommon." "It's incredibly rare to find anyone who can build a speech recognition model."
Furthermore, constructing the model necessitates labelled training data, which an organisation can lack or obtain only with great difficulty.
A typical scenario is an organisation that requires intelligent text extraction to deal with workflows involving a large number of scanned documents. Incorporating such records, extracting data into a format that the organisation can use, and performing text analytics are not skills that most businesses have in-house.
Consultants are often often used by businesses for one-time projects.
"Data science as a service will provide a fast hit to solve a quick business issue," said Chandana Gopal, IDC's future of intelligence research chief. "When you don't have the expertise in-house, you're able to achieve market results quickly."
Outsourcing can also make sense if an organisation is unsure whether or not a project can succeed. It's pointless to hire a team of data scientists just to discover that the data doesn't produce the desired outcome.
"Once you know there's a known advantage to using data scientists," Gopal said, "you could recruit an in-house team."