Growing volume of machine generated, not human created, data confronts the AI industry
Growing volume of machine generated, not human created, data confronts the AI industry
TIM LEEMASTER
Inzata Analytics COO Christopher Rafter
Inzata Analytics, the Tampa, Florida-based AI firm, is seeing a growing volume of data from the machines we all use that carry different characteristics than data produced by humans, says Chief Operating Officer Christopher Rafter.
“It’s not well structured and evolving into 100 times the size of human generated data,” Rafter says. “It looks different and accounts for about 15-20% of the problems that we see now.”
The five-year-old software firm allows companies to organize, blend and model raw data using AI.
Whether its Wi-Fi, a router or a digital camera data is getting logged somewhere, particularly with the growth of the Internet of Things, Rafter says.
“It’s a new type of data – not necessarily better or worse but we’re noticing that it is increasing.”
The company, which has close to twenty clients and is targeting 100 this year, starts with a look at how a client structures and organizes data. Often data systems are used differently by different people so a restructuring of the data begins by finding the most authoritative, or “believable” but not necessarily 100% convincing, source for each data point internally.
Companies can also have data spread across as many as 10 to 15 different systems. “There’s a lot of ‘cave exploring’," Rafter says but that lays the foundation for a new unitary data set.
Passes through to identify missing and duplicated data are also required and predictive modelling can be used to generate missing data.
It can often get quite complicated, particularly when a company has made acquisitions where operational integration, not data, is the focus. “Systems are mushed together [and] attention is not paid to data quality,” according to Rafter.
An electronics distributor Inzata worked with, for example, had 24 different records for a single sale of double AA batteries, Rafter says.
With that firm it took a full year to get an operational data set. Other cases have taken as little as two to three weeks. “Volume is the biggest determiner of the average time it takes,” Rafter says.
The company cottoned on to its current direction about three years ago when Inzata worked out that clients were really only accessing and analyzing about 10% of the data they had available.
“It’s like pumping crude oil and not driving it to the refinery,” Rafter says.
Inzata plans to double its headcount this year to about 40 by adding four or five technology staff and 15 to 16 sales and marketing employees.
Its main shareholder is fellow Floridian firm DSM Technology Consultants, a cloud computing company. “Friends and family” investors make up the rest. At some point after the company hits 100 clients a fund raise may be in order, Rafter says.
The company became cash flow positive last year but missed a previously set revenue target for 2019 of USD 5m.
“It’s such a competitive marketplace for analytics,” Rafter says. “We have a good conversion rate when we get in front of people but it’s hard to get their attention.”
Rafter expects to hit USD 5m in revenue by the middle of this year.