Follow a mango from the cuvette to a fit-for-purpose verdict. Every chart is a real run of the Lab engine.
600 whole mango fruit from three seasons; dry matter measured in the laboratory, from 10.9 to 22.9 % (mean 16.1).
The laboratory number is the truth the model learns from. The instrument never sees it directly.
Collect reference values for as many samples as you can before you scan.
Illustration, not data. Source: Spectra Lab engine run, 9 Oct 2026, mango teaching subset (CC BY 4.0).
Every fruit scanned from 684 to 990 nm: 103 points.
The curves look alike. They differ in height, slope and a few small bends.
Look at your own spectra before you model them.
Raw spectra: the curves differ mostly by height and slope.
Source: Spectra Lab engine run, 9 Oct 2026, mango teaching subset (CC BY 4.0). Twelve of the 600 calibration fruit, coloured from low to high dry matter.
Four pipelines on the same model: raw, 1st derivative, 2nd derivative, SNV.
Error on the unseen season: 1.00 with 1st derivative, 1.20 with SNV.
Try several. The best choice depends on your data.
Source: Spectra Lab engine run, 9 Oct 2026, mango teaching subset (CC BY 4.0).
PLS with 1 to 15 components; a smoothing window from 5 to 31 points.
The error flattens at about 10 components. The smoothing window barely moves it.
Do not chase the last 0.01. Stop where the curve goes flat.
10 components with 1st derivative.
Source: Spectra Lab engine run, 9 Oct 2026, mango teaching subset (CC BY 4.0).
Window 11 points: the Lab default.
Source: Spectra Lab engine run, 9 Oct 2026, mango teaching subset (CC BY 4.0). 1st derivative, engine Savitzky-Golay with a 2nd-order polynomial.
100 fruit from a fourth season.
Typical error 1.00 % dry matter, bias -0.19, RPD 2.94.
Always hold out a whole season or lot, not random rows.
1st derivative: typical error 1.00 % dry matter, bias -0.19, error after removing bias 0.99, RPD 2.94, which the Lab bands as rough quantitative.
Source: Spectra Lab engine run, 9 Oct 2026, mango teaching subset (CC BY 4.0). 100 fruit from a season the model never saw. RPD here is the spread of the reference values divided by the error after removing bias.
The Lab's bands: below 1 no predictive value, 1 to 2 screening only, 2 to 3 rough quantitative, 3 or more quantitative.
This model is 2.94, just under the quantitative line.
Compare the error with the limit you need, then decide.
Your limit is 1.00; 1st derivative gives a typical error of 1.00 on a season it never saw, and an RPD of 2.94 (rough quantitative).
Source: Spectra Lab engine run, 9 Oct 2026, mango teaching subset (CC BY 4.0). Your limit is your own requirement; the Lab does not set it.
Next step: calculators · waitlist · the NIR course.