Purpl Scientific

Wondering if anyone has any input on this device. They say they test within +/- 2% when compared to laboratory HPLC analysis results. Sounds interesting but curious as to the science behind it.

Seems like it only does flower or did I misinterpret the video?

I was demoed a unit similar to this once. It was accurate for flower, but totally failed concentrates

uses near infared & and internet connection.
the FAQ claims “concentrates are coming”.
unlike the more expensive machines, which do all the math locally, this thing seems to store the training data on the web. as such it might be more trainable than some of the other offerings.

1 Like

Guess we have to let the IT team keep working before we buy. Whoever gets this mobile on the fly tech first will make a grip.

1 Like

At +/-2% it’s intriguing and might be useful to track daily levels if enough representative samples are taken.

Wonder if ±2% is 1%±0 .2 or 1%±2%


I came to here to look for info on this. Boss is very interested but not willing to be the guinea pig.

Would love to hear some first hand experiences. Does anyone know of the company that makes it and their reputation before now?

Saw them at a trade show, they demonstrated it. Works like a charm. Their white paper, if accurate demonstrates their accuracy with around 35,000 samples between 0 and 25% THC

Only works with flower currently.

1 Like

Makes a world of difference doesn’t it

Above 20% total THC, it’s +/-10%? I wonder how accurate it is above 30% total THC? I would guess greater than +/-10%. I mention that because few phenos put out around 30%+.

The Total THC predictions of the Purpl PRO validation measurements are shown in Figure 1. This plot shows the average predicted Total THC via the Purpl PRO devices compared to the HPLC measurements from the respective laboratories. The plots also show the target accuracy bounds of ± 2% up to 20% THC, and ± 10% of the measured value above 20% THC (dashed diagonal lines).

Side note: HPTLC is interesting, but maybe should have its own thread?


The method validated in this study offers advantages over previously developed TLC methods for cannabis analysis because it utilises a simple decarboxylation procedure to convert THCA into ∆9-THC and requires only normal-phase HPTLC plates with an automated spotter and scanner. The method was shown to be comparable within a small degree of error (±0.5%) to results obtainable from a validated HPLC method. This chromatographic method can be utilised for qualitative fingerprinting of cannabis since good separation is achieved between the principle neutral cannabinoids (CBD, ∆9-THC, THCV, CBG and CBC), forensic analysis of cannabis, quality control of hemp and quality control of medical cannabis.

I remember a unit like this coming out a few years ago that also used a database driven system 4 coming up with its readings I don’t remember it being very accurate

1 Like

Me too. And me either.

I’m wondering what the “predictive” part of the measurement is.
Does that mean it guesses the potency or actually measures it.
Hugh Goldsmith

1 Like

If I understand correctly this is like (or is exactly) near ir, and uses a machine learning algorithm to predict concentration from a very messy spectrum

As I understood it, the spectrum identifies a theoretical strain of cannabis that has been previously identified and measured using GC or HPLC and then the algorithm says " this strain historically has measured x% potency ) so therefore this sample has x% potency also.
Its not actually measuring the potency, its inferring it.

Happy to see interest in our Purpl PRO, and wanted to clarify some of the questions/comments above. The Purpl PRO utilizes near-infrared spectroscopy (NIRS), a technique originally spurred by the USDA ~70 years ago and currently utilized in nearly all manufacturing industries for quality assurance/control, including pharmaceutical production, food & beverage, chemical processing, etc.

NIRS is a quantitative technique wherein a scan of a sample generates a near-infrared spectrum, which is a series of peaks and valleys whose heights correspond to the optical absorption of particular molecules (in this case, THC and CBD). The NIRS model is trained by measuring a wide range of samples so the heights of these peaks and valleys can be mathematically calibrated against the chemical concentration provided by HPLC. This approach is akin to those of techniques like GC and HPLC where the scanned sample peak heights are calibrated against concentrations of a purified standard. As @TheGratefulPhil noted from our whitepaper, that meant over 35,000 NIRS scans to build a reliable prediction model. There are competitive offerings in this space that use other approaches like the “pattern matching” mentioned by @srihugh1, but such techniques are quite error-prone and non-quantitative, and as such do not have the pedigree of NIRS.

The Purpl PRO currently measures THC up to 30%, and CBD up to 15%, as there were not enough samples available above these ranges at the time of our last calibration. It is useful to note that many retail samples were labeled as higher than 30% THC, but their true potency via HPLC were not nearly as high. But as new genetics and growing techniques allow higher concentrations, we will be able to extend this range. We will also be releasing software updates in the coming months that will allow the Purpl PRO to measure parameters other than flower potency, these include moisture, water activity, and extract potency, among others. These additions will not require a new device, but simply an update of the phone app.

We began selling the Purpl PRO this year, after years of R&D to build a reliable tool for this market. Our aim is to continue to provide reliable solutions based on science rather than marketing, and we welcome you to read more on our website, on our Instagram feed, and via other user forums. More content will be added as we develop new solutions, explain our solutions, and share the experiences of other users. Again, thanks for your interest!


I met one of your guys at Seattle Cannacon back in 2017, glad to see you guys finally have some units in the market!

If you are still undergoing R&D/testing on the concentrates portion, I run a legal i502 facility south of Olympia in WA and would be happy to do trial concentrate runs on a unit, we deal with concentrates that have a wide variety of consistencies, colors, clarity and potency. Could definitely add a decent volume of scanning data for ya if interested.

1 Like

Welcome to the community!

1 Like