The Blog

view all

More precision on Castillo v Caturra

Earlier this month, I had the honor of presenting to the SCAA Symposium the preliminary results of the work we did with friends in the research and specialty coffee communities on the Colombia Sensory Trial.  The following day, the Colombian Coffee Hub published this summary of my presentation, which concluded with my observation that the data showed Castillo can be good.  Very good.

Indeed, it does appear that the Trial delivered important validation of the quality potential of a variety that has been consistently maligned in a certain segment of the specialty market.

I think it is important, however, to articulate three caveats to the Colombian Coffee Hub summary.


Colombian Coffee Hub





N = 21

First, the small sample size for the trial—just 21 farms—did not allow us to generate the statistical robustness we had hoped for. The preliminary results I presented using scores from a small number of 2014 harvest samples are not statistically significant.  They don’t tell us anything about the relative quality potential for Castillo or Caturra outside of Colombia. Or even outside of Nariño. Or even what these coffees are capable of for the 2015 harvest.


Second, the findings reported on in the Hub’s summary reflect only the quantitative results of the two cupping panels we held at Intelligentsia Roasting Works in Chicago in October 2014 and January 2015.  They do not incorporate the insights of the sensory analysts at Kansas State University who applied the new SCAA/WCR sensory lexicon to this work.  Their contributions “quantify the qualitative” by and are vitally important complements to the average scores I reported.


And third, we should introduce a bit more precision to the conclusion you publish in the final paragraph in boldface: “There are no evidence of differences between Caturra and Castillo when we’re talking about taste or flavor.”  That is not exactly what the data showed.

What the data from the cupping panels showed was that there was no statistically significant difference in the average scores that the two varieties earned during the cupping panels. And the cupping panel data showed that there was no statistically significant difference between the average scores for any of the specific sensory categories included in the analysis. The fact that the two coffees earned roughly the same average overall and category-level scores does not mean there is no difference between the two varieties in taste or flavor. In fact, this is precisely what the very preliminary data from the Kansas State work suggest—that while Castillo and Caturra can both be good, a good Castillo has a different taste or flavor than a good Caturra.

The sensory analysts at KSU were able to succeed where the cuppers failed to generate statistically significant evidence of separation bewteen the two varieties. They did not show that one variety was better than another but that there is a substantial and quantifiable difference in the sensory characteristics of the two varieties. A good Castillo was fruity but not citric, with notes of dark chocolate and roasted nuts, while a good Caturra was floral with cocoa and caramel notes. This research is still ongoing—I look forward to the opportunity to share more information with readers of Coffeelands and the Colombian Coffee Hub when I have it.


  • Miguel says:

    I really appreciate this research and its intent to help farmers and policy makers to make better decisions.

    It would be great to see how cost of production and yields are linked to those samples. In addition to the differences (or lack of significant difference) on flavor/cupping results in the sample in Narino, how much “more expensive ” is for a farmer in this sample to grow Castillo? and how greater those yields are (over time)? It might not be a sample that allows for statistical extrapolation to Colombia or to the region, but I think it stills provides great information to the industry, development practitioners and farmers to make (or support) decisions about what to plant next.

    Thanks for the research and the info Michael/CRS

    • Michael Sheridan says:


      Thanks for the comment.

      Stay tuned for a discussion of production costs in Nariño based on data from our Borderlands project.

      We do not yet have COP data disaggregated by variety, and it may be very difficult to get. From what we have observed, farmers are not managing their Castillo in a way that is substantially different than the way they are managing the other varieties they have planted. While this means that the two varieties in our Colombia Sensory Trial were managed identically, it probably also means that they were not both managed optimally. This is one of a slew of other caveats about the CST findings–the samples faithfully represent the way the two varieties are being brought to market today in Nariño, but they don’t necessarily represent the highest expression of each variety.

      In summary, we have set for ourselves the objective of understanding whether there are real differences in the field in how much growers are actually spending to produce these two varieties. If/when we get there, we will publish our findings here.


Leave a Reply

Your email address will not be published. Required fields are marked *

Note: You can use basic XHTML in your comments. Your email address will never be published.

Subscribe to this comment feed via RSS