Chocolate making, professionally

Bean-to-Bar Process: Every Step a Chocolate Maker Controls

By Oliver BennettPublished October 15, 2026Updated October 15, 202615 min read

The bean-to-bar process runs the same chain in a two-person workshop as in a factory: check and clean the beans, roast, crack and winnow, grind and refine, conch, temper, form and pack. For the person running it, every step comes down to three things: what you decide, what you measure and what tends to go wrong. The plain story of how cocoa beans become chocolate is in how chocolate is made.

Many published cocoa figures come from trade contracts or tasting-sample protocols rather than production recipes, so check which kind of number you’re looking at.

The process at a glance

In a small bean-to-bar workshop, the same one or two people usually make every call, from accepting a sack of beans to the date on the wrapper. A few early decisions shape the rest. Roasting whole beans or nibs changes how you winnow and how you show that the roast is safe. Refining and conching in one melanger drum, or on separate machines, sets how far you can adjust each one on its own.

Sourcing beans

The beans set a ceiling on the chocolate, and some faults can’t be fixed later in the process. A moldy, musty off-flavor is the standard example: End’s presentation to the International Cocoa Organization of the industry’s quality guide lists it as one the manufacturer cannot remove.

That guide, Cocoa Beans: Chocolate and Cocoa Industry Quality Requirements, comes up often below. European cocoa and chocolate trade associations and the Federation of Cocoa Commerce (FCC) wrote it, so it’s an industry text, but it’s the most accessible one that sets the ISO standard and the trade’s contract terms side by side.

Choosing origin and supplier

Origin shapes much of the flavor, but for a maker the supplier’s consistency matters as much. The guide defines consistency as more than the share of defective beans: it covers bean size and degree of fermentation too, and it advises against blending poor beans with good ones just to meet a minimum grade.

Ask about cadmium before you buy. The guide ties the problem to beans from certain regions of some producing countries, particularly in Latin America and the Caribbean, and EU limits on the finished chocolate rise with cocoa content, so for a high-percentage single-origin bar a supplier analysis is worth requesting. The limits are tabled in chocolate food safety for small makers.

What to check on arrival

Three checks cover most of what a small maker can test without a lab: moisture, a cut test and a bean count. The ISO 2451 bean standard requires beans dried to no more than 7.5% moisture, according to the industry guide’s summary of the ISO cocoa standards. Its cut test grades a sample of 300 beans by five defects: moldy, slaty, insect-damaged, germinated and flat.

Those are trade limits, and the guide warns that beans close to the ISO Grade I limits are unsuitable for good quality chocolate. How to run each check, what the defects look like and how the ISO, FCC and US contract limits compare are in cocoa bean quality checks.

Sorting and cleaning

Sorting takes out bean clusters, flat and broken beans, and anything that isn’t cocoa. Broken beans are worth removing early, because the industry guide notes they can significantly raise the free fatty acid content of the fat.

Cleaning is also where hygiene zoning starts. Raw beans carry what the guide calls an intrinsic risk of microbial contamination, so give raw beans and roasted nibs separate containers, tools and routes through the room. The trade’s foreign-matter thresholds and the guide’s hygiene advice are in cocoa bean quality checks.

Roasting

Roasting has more settings than any other step and no standard profile. The industry guide’s roasting protocol advises mapping temperature, time and loading separately for each type of bean and for the specific roaster you use.

Roast profiles and what they change

A roast profile is more than a temperature and a time. The same protocol, written for tasting samples, treats the load as part of it (one layer, topped up with filler beans), starts the clock 2 °C below the set point and counts the oven itself as a variable, down to how evenly air moves and how fast it recovers after the door opens. Very dry beans (below 6.5% moisture) or very wet ones (8.5% or more) may need different settings. For tasting samples in a convection tray roaster it suggests 112, 120 and 130 °C, each for 25 minutes, as low, medium and high roasts: sample-evaluation settings, not a production recipe.

Peer-reviewed evidence on what the profile changes is thin. In a Brazilian study by Rocha and colleagues (2017), consumers judged chocolate from beans roasted at 80, 120 and 160 °C for 20, 40 and 60 minutes. Temperature mattered and time, within those levels, didn’t; the 160 °C chocolate tasted burnt, and 90 to 110 °C scored best. With one bean source, one roaster and one panel, that range is something to test, nothing more.

Published craft settings vary widely. A 2021 paper by Yan and colleagues quotes fine-chocolate protocols from 110 to 160 °C (usually 120 to 140 °C) for 5 to 120 minutes (usually 10 to 35), taken from an earlier review. Its authors’ background is noted in chocolate food safety for small makers, which uses the same paper’s kill data. A small-melanger study by Clark and colleagues (2020) roasted for 17 minutes at 163 °C, adjusting large-volume guidelines with trial roasts.

What to log for every batch, then: set temperature and, if you can measure it, bean temperature; time and when you started counting; batch weight; and the moisture and bean count of the lot. Taste is the result you’re steering, but those are the inputs that let you repeat it.

Whole beans or nibs

Most of the industry roasts whole beans, for the sake of winnowing. A 2022 review by Peña-Correa and colleagues notes that industry uses continuous drum roasters for 20 minutes to 2 hours, and that water escaping from the shell and the edges of the nib loosens the shell and makes it easier to winnow.

There is a cost. The same review describes cocoa butter migrating toward the shell during whole-bean roasting, with losses of as much as 0.5% of the cocoa butter at deshelling. Roasting nibs reduces that loss, but then you shell raw beans, and the industry guide’s winnowing protocol notes that shells on under-fermented or lightly roasted beans stick tightly to the nib.

The choice also changes the food safety math. Nascimento and colleagues (2012) found Salmonella more heat-resistant in nibs than in whole beans at 110 to 130 °C, so whole-bean kill data can’t simply be carried over to a nib roast. The figures are in the food safety guide.

Equipment classes for small batches

What separates small roasters in the published sources is how well they hold a repeatable load. A tray convection oven can be topped up with filler beans; the roasting protocol notes a rotary drum can’t, so with a drum, keeping batch weight constant is up to you. Fluidized bed roasting of nibs, in a stream of hot air, ran 16 times faster than oven roasting in a 2022 study by Peña-Correa and colleagues and favored the pyrazines behind roasted chocolate aroma without the typical burnt volatiles. That is experimental work.

Cracking and winnowing

Winnowing sets how much of each bean you keep and how much shell ends up in the chocolate, and small machines tend to trade one against the other. It is also a food safety step: the industry guide calls the shell often the most contaminated part of the bean, for microbes including molds, for polycyclic aromatic hydrocarbons (PAH), for ochratoxin A and for some pesticide residues.

The Codex standard for cocoa mass (CXS 141-1983) allows shell and germ of no more than 5% m/m calculated on the fat-free dry matter, and names the lab methods for measuring it: AOAC 968.10 and 970.23. US rules in 21 CFR 163.110 cap shell in cocoa nibs at 1.75% by weight on an alkali-free basis.

Yield is the other side. Industrial winnowers size the broken pieces and adjust the airflow to each size class. Laboratory-scale breakers and winnowers skip the sizing, and the guide’s winnowing protocol says their nib yields are frequently as low as 62 to 78% of the starting raw beans, a tasting-sample figure that still shows where small-scale winnowing loses material.

The same protocol breaks and winnows beans as soon as they cool after roasting (usually 20 to 60 minutes), so the nibs don’t pick up room off-flavors, and turns nibs into cocoa mass within 48 hours of roasting or stores them at most seven days at 10 to 24 °C in sealed bags.

On every batch, weigh the nibs out against the beans in, and now and then send nibs to a lab to check the shell left in them.

Grinding and refining

Refining sets the particle size of the chocolate and with it the mouthfeel. For a small maker the real choice is the machine class, because it also shapes how conching works. The full comparison is in chocolate refining explained.

Melangers, roller refiners, ball mills: when each makes sense

A melanger is the usual starting point. Clark and colleagues (2020), citing trade press, describe it as the most common refining tool for small-batch makers because it scales easily and costs little up front, and it usually refines and conches in the same drum. The price is time and a dependence on recipe: it takes hours where a roll refiner takes minutes, and a low-fat recipe refines poorly in it.

A roller refiner is the industrial route. It runs continuously, turns out a dry flake and commits you to a separate conche. A ball mill covers the widest range of scales and recirculates the mass until it is fine enough, though two small studies found ball-milled products needed more force to start flowing than roll-refined ones, so test flow on your own recipe.

Whatever the machine, measure particle size and write down how. A micrometer, a grindometer and a laser D90 read different things, so a “20 µm” from each is not the same number.

Conching

Conching finishes the texture and changes the aroma. It breaks up the particle clumps left after refining and frees the fat trapped inside them, as chocolate refining explained describes, so the chocolate flows better. That flow, measured as viscosity, governs how the chocolate behaves in the mold or over an enrober.

What conching does to acids

Conching is often described as driving off acids, but the data are mixed. In a 7-hour, three-phase laboratory conche run by Albak and Tekin (2016), the measured aroma compounds totaled 37,606 ppb before conching and 30,197 ppb at the end, about a fifth lower, yet acetic acid, the vinegary acid left from fermentation, rose from 2,471 to 9,797 ppb in the hot pasty phase and ended at 5,353, still above where it started. In Clark’s melanger, one form of acetic acid was gone by 8 hours and another stayed abundant to the end. Conching lowers the total load of volatiles, but single acids vary from study to study, so judge your batches by tasting rather than by the clock.

Melanger or conche: temperature

A melanger is not a heated conche. Small-batch makers typically run melangers at ambient temperature, while conches are generally heated up to 80 °C, according to Clark and colleagues. Friction warms the mass anyway: to about 40 °C in a 2019 study by Hinneh and colleagues, and to 78 °C in a low-fat batch (27% fat) after more than 3 hours. Neither reaches Maillard temperatures, per the work Clark cites, and in Clark’s melanger time affected flavor more than temperature did.

Heat does speed up aroma loss: the industry guide’s sample protocol keeps ground cocoa mass below 55 °C because above that volatiles escape at a substantially higher rate.

How long to conch

No peer-reviewed study gives a target conching time for a melanger. Clark’s data run to 24 hours, and Clark’s 2019 thesis cautions that larger melangers would need longer times and that more testing is needed to know whether results scale linearly with batch volume. The 7-hour lab-conche program above shouldn’t be copied onto a melanger. Combined machines like the refiner-conche are compared in the refining article.

Aging

Aging is optional: some makers rest finished chocolate in blocks before tempering and molding, and say the flavor settles. The closest published source is the industry guide’s note that chocolate, particularly semi-sweet chocolate, changes its flavor profile and mellows in long storage, and that holding it 2 to 4 months before assessment may not be practical. That isn’t quite the same as resting blocks. If you try it, taste an aged block blind against fresh chocolate from the same batch.

Tempering at scale

At small scale, tempering can be done on a marble slab, in a double boiler, in a small tabletop tempering machine or with pre-crystallized cocoa butter, according to the industry guide’s tempering protocol. Factories run continuous machines, and research follows them: Bolliger and colleagues (1998) found a high-shear crystallizer matched a conventional tempering machine on crystallization quality with a much shorter residence time and a lower outlet temperature. Volume and the consistency you need pick the route. The goal is the same at any scale, cocoa butter set in its stable crystal form (one of the cocoa butter polymorphs); what tempering is and why it matters explains why, and how to temper chocolate at home covers the hand methods.

Don’t expect one universal curve. The industry guide melts at 45 °C, cools to about 27 °C and rewarms to about 31 °C, while a University of Wisconsin thesis by Jin (2019) gives around 50, 26 and 32 °C. A chocolatier working with couverture takes temperatures from the supplier’s data sheet. A bean-to-bar maker is the supplier, so the working temperatures for each recipe come from your own tests. Seeding during cooling speeds crystallization and improves it in the stable form, Jin notes.

Then check the result: a temper test on a strip or knife tip, or a tempermeter, whose cooling curve shows crystallization heat as a plateau on the temper curve. The Wisconsin thesis calls a slope between -1 and 1 usually well tempered, while noting the reading says only a limited amount about crystallization.

Molding, enrobing and other forming

Tempered chocolate should go into molds right away, and the industry guide’s protocol has cooling follow immediately; fridges and air-conditioned rooms used to cool filled molds should be checked for off-odors first. Cooling too fast leaves unstable crystals, hair cracks and pores, and condensation is one route to sugar bloom; why chocolate turns white covers how the two kinds of bloom, fat bloom and sugar bloom, form and what holds them back.

Molding a shell for a filled piece, depositing (including one-shot depositing) and enrobing all need chocolate that flows the right way, which is where the viscosity you set in conching pays off. Compound coatings skip tempering, which is why they’re common in enrobing; real vs compound chocolate has the trade-offs. For drops and chips, see how chocolate drops and chips are made. Larger lines cool molds in a cooling tunnel. Solid, shell and one-shot molding are compared in chocolate molding.

Packaging and shelf life

In the EU, the maker sets the shelf life. Regulation (EU) No 1169/2011 defines the date of minimum durability as the date until which the food keeps its specific properties when properly stored and gives no fixed period for chocolate (Art. 2); “use by” is reserved for highly perishable foods (Art. 24). The date carries storage conditions where needed, and its format follows the length: month and year for more than 3 months up to 18, the year alone beyond that (Annex X). The rest of the label, market by market, is in chocolate labeling rules.

For plain chocolate, what ends shelf life is usually quality rather than safety. A study by Iacumin and colleagues (2022) notes that molds don’t develop in chocolate because its water activity is low (below 0.6), though it recorded mold in bars with a milk cream filling, so measure water activity for filled products. Published shelf-life figures by chocolate type, such as Iowa State University Extension’s, describe commercial products and can’t replace testing your own recipe and packaging.

Food safety in a small production

A bean-to-bar workshop works under the same framework as a factory, HACCP, applied in proportion to its size. In the EU a HACCP-based procedure is a legal duty under Regulation (EC) No 852/2004; the US equivalent is the FDA’s preventive controls rule.

The main hazard in cocoa is Salmonella, and the roast is usually where it’s brought under control. Conching is a weak kill step, so a melanger at around 40 °C can’t stand in for a validated roast. The seven HACCP principles, the roast kill data and the EU limits for cadmium and ochratoxin A are in chocolate food safety for small makers.

The control table: what to measure at each step

The last column says what kind of value each reference point is; where no source gives a number, the table leaves it out.

Going from workshop to factory

No published source puts a volume on the point where a step outgrows a small machine. The sections above show where small equipment costs the most instead: nib lost in winnowing, hours in the melanger, and at larger scale the move to continuous roasting and tempering. Planning that move is covered in how to start a chocolate factory, the machines step by step in chocolate production line, and sizing in chocolate production capacity.

Cost pulls both ways. Hand work raises craft production costs, as Cadby (2021) notes and how chocolate is made explains, yet Hinneh and colleagues note that the cost of conventional equipment seems to be an impediment for producers in developing countries. Food safety guidance assumes a large plant too: the industry’s roast validation approach was written for large processors.

Sources and open questions

Questions the published record leaves open:

• No peer-reviewed study compares the flavor of whole-bean and nib roasting under the same conditions.
• Fluidized bed nib roasting isn’t confirmed as equipment a small maker can buy off the shelf.
• The Codex cocoa mass standard was read in its 2014 amended version; a later revision may have changed the shell limit.
• Nib storage at production scale and the shell left by small winnowers lack independent data.
• Aging blocks before tempering, and batch against continuous tempering for small makers, lack independent studies.
• Small-scale molding figures (vibration, fill weights, cooling times) come only from equipment makers or line operators.
• Regulation (EU) No 1169/2011 was read as first published; US date-labeling rules were not verified.