Tiny point-and-test scanner makes analysis easy for little firms
The small Hamilton-based company uses AI and spectrometry to help textile, hops, medical, oil and many other companies.


Quick facts:

  • Big data-enabled materials testing
  •  Founder George Hill and his wife Susan Hill own 38.4% each
  • Turnover undisclosed
  • Set up in 2010
  • Based in Hamilton

Imagine pointing a little black box at a shirt to see if it really is 100% cotton, discover if a plastic container is truly biodegradable or instantly know the quality of manuka, hops or cannabis oil.

It’s the sort of thing only large laboratories should be capable of, right? Well, Hamilton-based Sagitto says it has a solution to connect machine learning with near-infrared spectroscopy (NIR) technology to help anyone test the materials of almost anything.

Designed to be entirely controlled by an iPhone or iPad, the company’s miniature NIR scanner uses digital light processing (both systems based on technology developed by their partners at US company Texas Instruments) to measure the variability of raw materials in each sample.

Some examples of its use include composition and authenticity of essential oils; alpha and beta acid composition, HSI, oil and moisture content of hops; composition of plastic objects; fibre composition of textiles; thermoplastic sheathing (TPS) in electric wiring; and vanillin concentrations in vanilla extracts, according to the company’s website.

Sagitto's machine at work.
A Sagitto machine at work.

Experimenting constantly

Founder George Hill says his small team of data scientists are constantly working on experiments to deliver new solutions and possible applications for more and different types of materials.

“We’re working with textiles, hops, manuka oil, cannabis, plastics, kava, hops and many others. It always amazes me how many different applications keep coming out of the woodwork.

“This is putting a laboratory in the hands of our customers. The device gathers and uses the spectral information to get what we call ‘training data.’ This data and the models generated are both owned by the customer – we just help them get the best from it with machine learning,” he says.

The technology allows companies to monetise their “underutilised data” by selling access to these models on a per analysis basis. Artificial intelligence is the core and, even though the data sets are generally small, and it takes a lot of time to build the models, Sagitto’s solution and hardware are far cheaper than similar offerings from large pharmaceutical businesses.

For instance, each of its spectrometers cost $US1950 but come pre-loaded with the latest machine learning models on a service plan ranging from $950 to $3500 per month on an hourly usage basis. However, Hill says other large companies charge about $US20,000 just for the hardware, without any of the data, AI or services.

Mr Hill says although a Callaghan Innovation R&D career grant allowed it to employ a data science graduate earlier this year and he encourages the government to keep funding this capability, most of the company’s investment has come from family and friends.

“The company is majority owned by myself and my wife Susan, with 10 friends and family shareholders who have been with the company since its inception in 2010 and who between them have invested about $150,000.

“About 60% of Sagitto’s revenue is from exports but this is expected to climb to over 80% by the end of the 2019 financial year,” he says.

Where it all fits
While Sagitto’s small team of three is constantly looking for new sectors and materials to build new prediction and analysis models, the present applications are gaining some traction.

“We’ve probably sold about 20 so far because often there is quite a long gestation cycle. We’ve been working with Hop Products Australia for over a year and it's only now that together we’re able to see the opportunity. And we’ve been working with NZ Manuka oil for two years,” Hill says.

The “opportunity” with analysing hops products lies in the growing – but still small – global craft brewing industry. Traditionally, most hops were sold to big breweries that owned their own laboratories. But today in the US, there are about 6500 craft breweries, about 2500 in the UK, and a few hundred in both New Zealand and Australia.

“These brewers aren’t big enough to have their own laboratory, so customers have to rely on what the label says. But hops are a natural product that degrades over time, which means these brewers are working blind and don’t really know what they’re using. This technology allows them to test the hops almost instantly to see what its real quality is.”

The world is also looking closer at what happens to textiles at the end of their lives. In some European countries, for instance, it is illegal to dump textiles, which creates a serious problem for large retailers, so more investment is flowing into recycling to turn textiles back into fibre.

“All of these processes are great but they require companies to know exactly what they’re feeding into those recycling systems. If a process is designed for recycling polyester cotton, you won’t want to feed it wool. That’s where we can help with an application to determine exactly whether it is 80% cotton or 95% polyester, for instance,” Hill says.

Sagitto managing director George Hill.

Is it biodegradable?

Another application is in determining the biodegradability of a particular plastic or its overall type. Similarly, with electrical cabling.

“If it complies, that’s good, but a lot of this electrical cabling is made in countries you can’t always trust to make it the same. We’re able to let our customers test every batch to ensure the cable is of high quality and will conform to standards.”

Although the company began about eight years ago, the technology and market have only matured in the past few years. The machine learning existed, as did the software, but the miniaturisation hadn’t happened. And since Microsoft and Google have each publicised the concept of machine learning, the path is opening for Sagitto to push its solution further afield.

“Even three years ago, if I told people we were using machine learning, I’d get a blank look. Now, the pitch has gotten easier. So, a lot has happened in the last two or three years. We have access now to cloud computing, which we didn’t have before. The vision was always the right one but just having all those pieces fall into place has been the challenge,” Hill says.

“The margins are very good, so we need to get volumes up. The hard part is getting the first customer for the application and, once this can be replicated, then it can become profitable. The payout comes down the line. We’re won’t be a unicorn, that’s not the aspiration. But we’re happy to ride this and see where it goes.

- Nathan Smith

Share this article