How Commercial Roasters Should Set QC Tolerances for Cambodian Coffee

Commercial coffee quality control works best when the roastery defines what variation is acceptable before a problem occurs. A tolerance is not permission for inconsistency. It is a practical boundary that separates normal production variation from a batch or lot that needs review.

For Cambodian coffee and Fine Robusta, tolerances should be built around the actual product. There is no responsible universal national specification for roast color, moisture, defects, sensory score, or espresso performance that applies to every Cambodian lot.

The roaster should define its own acceptance system from buyer specifications, reference samples, production data, sensory targets, and intended use.

Start with product identity

Quality control begins by confirming that the correct coffee is being evaluated.

Record:

• Sku;
• green lot code;
• supplier;
• origin;
• region where verified;
• harvest;
• process;
• roast profile version;
• roast date;
• production batch.

If identity is wrong, later measurements are meaningless.

A strong system makes it possible to trace a customer complaint back to the relevant green and roast records.

Separate green-coffee acceptance from roast acceptance

Green QC asks whether incoming coffee meets the purchasing specification.

Roast QC asks whether production transformed the accepted green coffee into the intended roasted product.

Do not combine them into one vague “quality” score.

A roast can be executed correctly on a green lot that has drifted. A green lot can be excellent and then be roasted poorly.

The system should identify which stage changed.

Define lot-level green specifications

Useful green-coffee fields may include:

• moisture;
• water activity where used;
• defect method and result;
• screen distribution where relevant;
• foreign matter;
• odor;
• visible condition;
• sample identity;
• process;
• storage or packaging condition.

The buyer should name the measurement method and acceptable range in the contract or internal specification where appropriate.

Avoid pretending that one number defines Fine Robusta quality.

Use tolerances that match measurement precision

A QC limit should not imply more precision than the method can deliver.

If a measurement device or sampling method has meaningful variability, an extremely narrow tolerance may create false failures.

Use calibrated or maintained equipment where appropriate, consistent sampling, and the same measurement procedure.

The purpose is to detect meaningful change.

Maintain an approved green reference

When practical, keep a representative reference sample from the approved lot.

Compare incoming or later inventory with the reference for:

• appearance;
• odor;
• moisture trend;
• physical preparation;
• sensory result.

A reference gives the team something concrete to compare rather than relying on memory.

Reference storage should also be controlled so the sample itself does not become misleading.

Define roast color tolerance from repeated successful batches

Roast color can help monitor consistency, but the acceptable range should come from the roastery's own data.

Begin by identifying several batches that meet the sensory target.

Measure them consistently using the same preparation method.

Then observe the natural production range.

The tolerance should be narrow enough to detect meaningful drift and wide enough to reflect normal controlled variation.

Do not copy a color number from another roaster or another coffee.

Whole-bean and ground color are not interchangeable

Color readings can differ depending on whether the sample is whole bean or ground, measurement device, grind, sample depth, and preparation.

Choose one standard method for release QC.

If both are recorded, treat them as separate measurements.

Consistency in method matters more than the apparent precision of the number.

Track roast weight loss

Weight loss can reveal production changes that color alone may miss.

Create a reference range from approved batches.

If weight loss moves materially while color remains similar, investigate heat application, moisture condition, batch loading, or profile behavior.

Do not use one universal percentage target for Fine Robusta.

The acceptable range should belong to the specific product and machine.

Time and temperature are process references, not final quality

Production software may track:

• charge condition;
• turning point;
• time to color change;
• first crack where detectable;
• development time;
• end temperature;
• total roast time.

These data help diagnose why a roast changed.

They should not automatically release a batch without sensory confirmation when sensory QC is part of the program.

A batch can follow the curve and still taste wrong because green coffee or machine conditions changed.

Define sensory acceptance separately

A useful sensory release standard describes what must be present and what must be absent.

For a Cambodian Fine Robusta espresso, the roastery might define requirements around:

• cleanliness;
• sweetness;
• body;
• bitterness quality;
• aftertaste;
• key flavor direction;
• absence of roast taint;
• absence of unacceptable defects.

Do not invent a national flavor profile.

The sensory specification belongs to the product.

Use a reference cup or retained roast sample

Where practical, compare the current batch with an approved reference from recent production.

The comparison can help detect gradual drift that becomes normalized when each batch is evaluated only against memory.

Because roasted coffee changes with age, compare samples at similar rest where possible.

Record the comparison conditions.

Define critical defects that trigger automatic review

Some observations should bypass ordinary tolerance discussion and trigger investigation.

Examples can include:

• foreign odor;
• severe roast scorching or burning;
• contamination;
• wrong coffee in the batch;
• packaging failure;
• major sensory defect;
• visible foreign material.

The roastery should decide which issues require hold-and-review rather than normal release.

Build a hold-and-release process

When a batch falls outside tolerance, do not automatically destroy it or ship it.

Use a defined workflow:

1. place batch on hold;
2. verify measurement;
3. repeat sensory evaluation;
4. check green lot and roast records;
5. compare reference;
6. decide release, rework where appropriate, alternate use, or rejection;
7. document decision.

This prevents both unnecessary waste and accidental release of a true failure.

Tolerances can differ by product application

A retail single-origin product may have tighter sensory identity requirements than a blend component.

A hotel breakfast coffee may prioritize broad consistency and batch-brew performance.

An espresso product may require strong consistency in body and bitterness balance.

Define tolerances according to the product promise.

One QC sheet does not need to govern every SKU identically.

Blend components need incoming and finished-product QC

A Fine Robusta component can fall within its own specification while the finished espresso blend still misses the target because another component changed.

Check:

• individual green components;
• roast consistency by component;
• blend ratio;
• finished sensory result.

The release decision belongs to the finished product as well as the ingredient.

Espresso QC should include application performance

If the coffee is sold primarily for espresso, test it through a standardized espresso setup.

Record a reference dose, beverage ratio, grinder, machine, water, and roast age.

The objective is not to make the roastery's QC recipe the customer's required recipe.

It is to detect meaningful batch-to-batch changes under controlled conditions.

Milk-drink products need a milk check when relevant

A coffee can pass straight espresso QC and perform differently in milk.

If milk performance is central to the commercial product, include a standardized milk beverage periodically or for profile changes.

Evaluate:

• coffee persistence;
• sweetness;
• bitterness;
• balance;
• aftertaste.

This connects QC to the use case customers actually buy.

Batch brew and filter products need different checks

A filter-focused Cambodian coffee may need stronger control over clarity, aroma, acidity, body, and process expression.

Use a repeatable brew method for QC.

Do not force espresso metrics onto a filter product.

The product brief should determine the test.

Packaging belongs inside QC

A correctly roasted batch can fail commercially if packaging is wrong.

Release checks may include:

• correct bag;
• correct label;
• correct net weight;
• seal integrity;
• valve or closure where applicable;
• lot or batch code;
• date coding;
• carton quantity.

Packaging errors can create regulatory, traceability, and customer-service problems.

Net weight should be controlled operationally

The roastery should use appropriate weighing and filling controls for the product and destination requirements.

Do not rely on visual estimation.

Legal requirements vary by market, so destination compliance should be confirmed separately.

The QC principle is that declared pack information must match the actual production process.

Supplier transitions need temporary tighter attention

When changing green lots, suppliers, harvests, or processing, increase QC frequency until the new coffee is understood.

Run additional:

• green measurements;
• sample roasts;
• production pilots;
• sensory comparisons;
• application tests.

Once the product stabilizes, routine frequency can be adjusted.

A new lot should not be treated as identical simply because the product name stays the same.

Create warning limits and action limits

Some roasteries benefit from two levels.

A warning limit indicates the batch is drifting toward the edge of normal behavior and should receive extra attention.

An action limit means the batch is outside the approved range and should be held.

This can reduce overreaction to small variation while still identifying meaningful process change.

The limits should come from real production data.

Avoid tolerance creep

Teams can slowly widen acceptance because repeated borderline batches become familiar.

Keep the approved reference and review why limits were established.

If the business wants to change the tolerance, do so deliberately with sensory and commercial evidence.

Do not let the standard move accidentally because rejecting or adjusting batches is inconvenient.

Use trend charts, not only pass/fail

A batch can pass today while a metric has been drifting for ten roasts.

Track trends in:

• roast color;
• weight loss;
• total time;
• sensory scores or descriptors;
• green moisture;
• complaint rate.

Trend data can identify problems before the action limit is reached.

This is more useful than treating every batch as isolated.

Complaint data should feed QC

Customer complaints can reveal issues that internal sampling missed.

Record:

• Sku;
• batch;
• complaint type;
• customer preparation where known;
• retained-sample result;
• root-cause conclusion;
• corrective action.

Do not assume every complaint proves production failure. Use the information to identify patterns.

Repeated complaints from the same batch deserve stronger investigation.

Fine Robusta requires careful language in QC records

Do not use “Fine Robusta” as a pass/fail field without defining what the roastery means.

The lot should be evaluated through measurable physical, sensory, and commercial criteria.

Historical Fine Robusta protocols can provide context, but current internal release should identify the actual methods being used.

A category label is not a substitute for the record.

Cambodia origin should be controlled as identity, not flavor guarantee

QC should verify whether the product is the correct Cambodia-origin lot.

It should not assume every Cambodian coffee must taste the same.

Region, process, harvest, and green characteristics can change sensory expression.

The job of QC is to protect the specific SKU promise while keeping origin claims accurate.

A commercial QC scorecard

A practical record may include:

Green

Lot code, moisture, water activity if used, defects, screen, odor.

Roast

Machine, batch size, profile version, color, weight loss, process data.

Sensory

Cleanliness, sweetness, body, bitterness quality, aroma, aftertaste, key notes.

Application

Espresso, milk, filter, or batch-brew result.

Packaging

Correct pack, weight, seal, label, code.

Release

Pass, hold, corrective action, approver.

Review tolerances periodically

A tolerance system should evolve when enough evidence exists.

Review when:

• machine changes;
• green supplier changes;
• batch size changes;
• packaging changes;
• product brief changes;
• repeated complaints occur;
• production data show the current limits are not useful.

The review should improve the system, not simply make it easier to pass.

Red flags

QC is weak when tolerances are copied from unrelated coffees, measurements use inconsistent methods, sensory release is replaced entirely by curve matching, batches outside range ship without documented review, or the team cannot connect complaints to production records.

Another warning sign is when every crop is expected to fit the old profile without revalidation.

Bottom line

Commercial roasters should set QC tolerances for Cambodian coffee from the specific green lot, machine, roast, product brief, and application. Use measurable physical references, sensory acceptance, application testing, packaging checks, and hold-and-release rules.

The strongest tolerance system does not pretend that production never varies. It identifies which variation is normal, which requires attention, and which threatens the customer promise.

Explore OCC Roasting Program or Contact OCC for commercial roasting and quality-control development.