Startup aims to turn enterprise data into information

Turning data into information has become the central problem of computing. Enterprises have so many disparate and incompatible databases, it becomes hard to know what data is available and what it means. Promethium means to solve that problem.
Written by Robin Harris, Contributor

See also

    Life would be simpler if everything were in one relational database, but that's not practical. Enterprises have too much data.

    So data gets spread among different databases, some by department, others by function, often with different schemas and degrees of data checking.

    There are lots of tools for low-level checking of fields and schemas. But what if you want to understand a complicated relationship between, say, customers of a certain region, purchasing some range of product quantity, and want to understand how their service calls have changed over time?

    That will become a major data extraction and analysis problem, starting with figuring out where the data is - probably in several different group databases - and tying all the numbers together. It will take weeks, and you'll probably have to set up another database. Painful at best.

    A related problem - especially since GDPR rules kicked in the EU - is how do you make sure you aren't releasing sensitive customer data? That's the issue of governance, and it will be a growing problem for decades to come.

    Smart washing?

    Artificial Intelligence (AI) and its major subset, Machine Learning (ML) seem to be part of every startup pitch these days - smart washing - and Promethium is no exception. Their secret sauce includes ML.

    Their software extracts metadata - which is much smaller than the entire database - and puts into a data repository. Then the ML system goes to work to infer the data's context from the metadata, so, for example, the system "knows" what a customer, product, region, and service call is, so the analyst can ask the question posed above, and the software will know where to look for the answers.

    On the governance side - protecting sensitive data - the software looks for combinations of data that could be sensitive. For example, social security numbers are easily filtered out today. But what about the combination of an IP address, first name, last name, and search terms? Individually each may not be sensitive, but together they could expose the company to severe penalties under GDPR.

    Cloud or on-prem?

    Promethium works with major databases - Oracle, Teradata, SQL Server - today, with more on the way. It can be installed locally, or in the cloud, and is offered on a subscription basis.

    Target users include analysts, compliance officers, and data privacy officers. The driving requirement is to help customers handle the ever-increasing scale of data collections in an era of tightening regulation and customer sensitivity.

    The Storage Bits take

    I'm generally skeptical of AI and ML claims, but if Promethium can train their model on enough databases, they should have a valuable tool. It isn't an easy problem - if it was they wouldn't have a business - but their tool is definitely worth a look if analysis and/or compliance is in your bailiwick.

    Courteous comments welcome, of course.

    Editorial standards