A smoky Ghanaian lot, a six-week launch and 1.2 tonnes of reference beans
A bean-to-bar maker in Copenhagen receives 2 tonnes of main-crop Ghanaian beans at 7.1% moisture for a 68% bar that must launch in six weeks. In a blind panel of six trained assessors, every sample shows medium-to-high smoke/phenolic and a distinct ham/bacon note; two assessors additionally flag musty/earthy. Bitterness and astringency both sit above the retained reference. A triangle test against that reference, using the same recipe and roast, returns 6/6 correct responses. The supplier explains that the lot was dried on a wooden platform with a fire underneath during two days of rain, and offers a 4% discount. The maker holds 1.2 tonnes of reference beans, enough for four weeks of normal production, and has already sold the launch into 90 retail doors.
1. Accept, reject, or blend? Give your decision with the reasoning and the numbers behind it.
Read the worked answer
Reject as the launch chocolate; do not blend into the flagship.
Statistical case. In a triangle test the chance of a correct guess is 1/3. Six correct out of six gives p = (1/3)^6 ≈ 0.0014 under the null of no difference, so the difference is real, not panel noise. A trained panel detecting it in 6/6 at medium-to-high intensity means consumers will detect it too.
Sensory case. Smoke/phenolic from direct fire drying sits in the same roasted, burnt-phenol space as the bar's own roast notes and cannot be roasted out or conched away - it is in the bean, not on it. At a medium-high intensity, blending at 10–20% would still be found by a trained panel and probably by a loud minority of customers. The ham/bacon note is a classic fire-drying/contamination marker; the musty/earthy flag from two assessors is the more serious signal because it points to mould and possible mycotoxin risk, which no amount of blending makes acceptable.
Commercial case. A 4% discount on 2 tonnes does not cover a failed launch across 90 doors. Keep the 1.2 t of reference beans for the launch, extend by buying spot reference-grade stock or by delaying four to six weeks, and use the rejected lot (if mycotoxin-clear) in a deliberately smoke-positioned product - never as the flagship single origin.
2. What confirmation and root-cause work would you commission before making the rejection final?
Read the worked answer
Sensory rejection is a judgement; back it with objective checks so the supplier conversation is evidence-based rather than taste-based.
- Mycotoxin screen: request or commission ochratoxin A and aflatoxin results on a representative sample. Any musty/earthy descriptor plus fire drying over two rainy days is a mould-risk combination; a positive result converts a quality decision into a food-safety one.
- Moisture and water activity: 7.1% is within normal range, but check the distribution across the lot and re-test after arrival, since wet patches drive mould.
- Cut test: look for mould hyphae, insect damage and under-fermentation, which would confound the bitterness/astringency reading.
- Reference-anchored panel: re-run the panel with an attribute list and intensity scale (0–5) on coded, randomised samples, including the retained reference, so you have intensities rather than adjectives.
- Roast trial: roast a 2 kg sample at your standard profile and at a lighter profile and re-taste. If smoke does not move, it confirms the defect is in the bean.
- Ask the supplier: drying method and fuel, duration, whether the fire was under or beside the bed, whether beans touched smoke, and how many other lots were dried the same way.
3. How do you report the outcome to the supplier and to your own buying team and retailers?
Read the worked answer
To the supplier. Send a one-page report: lot code, arrival date, sampling method, panel size and method, triangle test result with the probability, attribute intensities against the reference, mycotoxin status, and photographs of the drying method if available. State the decision (reject or accept at a defined discount for a non-flagship use) and the specific corrective action required: indirect drying, a cover during rain, no fire under the bed, moisture monitoring at the end of drying, and a pre-shipment sample for sensory approval. Keep the tone factual; the aim is a supplier who changes practice, not one who stops answering emails.
To the buying team. Present three options with costs: delay the launch four to six weeks, buy spot reference-grade beans at a stated premium, or reformulate the launch bar to a different origin. Put the numbers next to the 4% discount so the trade-off is explicit.
To retailers. If the launch slips, tell them early and give a date. Never describe a smoky lot as "artisanal smokiness" if the panel has called it a defect - one honest delay costs less than 90 doors discovering the problem in a customer review.