How Can You Tell Whether Consumers Really Perceive a Difference in Beer?

Triangle and Duo-Trio Tests: Making Evidence-Based Technical Decisions

Not Every Recipe Change Is Perceptible. Find Out Before Spending More Money.

The Question Behind the Change

Imagine the following situation.

You decide to increase the hop dosage in your IPA because you believe the aroma could become even more intense. Or you consider reducing the amount of a specialty malt to lower production cost without compromising the beer’s identity.

The following week, someone on the team comments:

“Now it is much better.”

Another person responds:

“Honestly, I did not notice a difference.”

Then the question that actually matters appears:

In practice, this question is harder than it seems. A person may expect the modified beer to be different because they already know what changed. Another evaluator may be influenced by serving order, temperature, fatigue, previous samples, conversation with the team or the simple expectation that a new version should be better.

That is why a technical sensory decision should not depend only on comments made around the brewhouse. Sample coding, controlled temperature, balanced serving order, silence during evaluation, adequate intervals between samples and a clear question are part of the test design.

Discrimination tests, such as the triangle test and the duo-trio test, help answer one specific question: whether a sensory difference can be perceived under controlled conditions.

Why Opinions Are Not Enough

Informal sensory analysis is part of the daily routine in breweries. Teams taste beer during fermentation, compare lots, follow stability, evaluate packaging changes and discuss whether a recipe adjustment made sense.

This routine is valuable. The problem begins when managerial decisions are based only on comments such as:

“I thought it was better.”

“To me, it was more aromatic.”

“It seems to have lost body.”

“I did not notice a difference.”

These comments may reflect real perceptions, but they may also reflect expectation, fatigue, order effect, previous knowledge, conversation among evaluators or personal preference. Without a controlled test, it is difficult to separate perception from interpretation.

In a brewery, this distinction matters because recipe changes, raw material substitutions and process adjustments can affect cost, consistency, production capacity and brand identity.

What Discrimination Tests Are Used For

Discrimination tests are used when the brewery wants to know whether two samples can be distinguished sensorially. They are useful in situations such as:

• recipe changes;
• supplier changes;
• raw material replacement;
• reduction or increase of ingredient dosage;
• process changes;
• changes in time or temperature;
• packaging changes;
• comparison between production lots;
• effect of storage conditions;
• stability evaluation over time.

These tests are designed to answer a narrow technical question: whether there is enough evidence that the samples are perceptibly different.

They do not answer whether one sample is better, preferred, more accepted, more intense in a specific attribute or commercially more attractive. When those are the questions, other methods are needed, such as preference tests, acceptance tests, descriptive analysis or consumer studies.

Triangle Test

In a triangle test, each evaluator receives three coded samples. Two samples are the same and one is different. The evaluator must identify the odd sample.

If the evaluator is only guessing, the probability of choosing the correct sample is 1 in 3, or 33.33%.

The triangle test is useful when the brewery wants to compare a recipe change, supplier change, raw material replacement, process adjustment, batch difference or an effect whose sensory direction is still unknown.

It is statistically efficient because the chance probability is lower than in two-choice methods. However, it demands more attention from assessors because three samples must be compared. This can be a limitation when the products have strong persistence, intense bitterness, high alcohol, pronounced acidity, marked carbonation or aromas that cause adaptation and sensory fatigue.

Duo-Trio Test

In a duo-trio test, the evaluator receives a reference sample and two coded samples. One coded sample is equal to the reference and the other is different. The task is to indicate which coded sample matches the reference.

When the evaluator is only guessing, the probability of a correct answer is 1 in 2, or 50%.

In the constant-reference version, the same reference is used for all assessors. This reference may be the current recipe, a standard batch, the regular product or a control sample familiar to the team.

In the balanced-reference version, samples A and B alternate as reference across assessors. This is useful when neither product is clearly the standard or when the brewery wants to avoid favoring one sample as the known reference.

The duo-trio test is usually easier for less experienced assessors because the comparison is anchored by an explicit reference. The trade-off is lower statistical efficiency: because the chance probability is 50%, more responses or a larger panel may be needed to reach the same level of evidence.

When to Use Each Method

• The triangle test has higher statistical efficiency.
• The duo-trio test is simpler for assessors because it uses an explicit reference.
• The triangle test may be limited when samples cause persistence, adaptation or sensory fatigue.
• The duo-trio test is useful when a known standard exists and the brewery wants to compare a new version with that standard.

Method Chosen for the Example

In the example below, the brewery evaluates a change in hop dosage. The beers are expected to be similar, and the main question is whether the change creates a perceptible sensory difference.

The triangle test is appropriate because the objective is to compare the original recipe with the modified recipe without telling assessors which one is the standard. If the brewery had a well-established control sample and wanted assessors to compare the new version directly with it, a duo-trio test with constant reference could also be considered.

Sample Preparation and Test Execution

The quality of the result depends heavily on how the test is prepared. A sensory test can lose validity if the samples are not controlled with care.

Important factors include:

• serving volume;
• temperature;
• glass type;
• coding;
• serving time;
• carbonation;
• turbidity;
• foam;
• lighting;
• environment;
• interval between evaluations.

Samples should be identified with random three-digit codes. Codes such as 111, 222, A, B, C, control or new version should be avoided because they may suggest order, hierarchy or identity.

Evaluators should not discuss the samples during the test. Conversation can create bias and influence later responses.

Visual cues also need attention. If color, turbidity or foam reveal the identity of a sample, the test may no longer be measuring the intended sensory difference. In some cases, controlled lighting or opaque cups may be necessary.

When aroma and flavor are the main focus, the test must be planned to reduce adaptation, carryover and fatigue. The order of presentation should be balanced. In the triangle test, the six possible sequences should be distributed across assessors whenever possible. In the duo-trio test, the position of the matching sample and the use of the reference must also be balanced.

How the Statistical Interpretation Works

Counting correct responses is not enough. Some correct answers will occur by chance.

In the triangle test, a person who guesses has a 1 in 3 probability of being correct. In the duo-trio test, a person who guesses has a 1 in 2 probability of being correct.

The statistical analysis compares the number of correct responses with the number expected by chance. The significance level defines how much risk the brewery accepts when concluding that a difference was detected.

A common criterion is alpha = 0.05. This does not mean absolute certainty. It means that, under the assumptions of the test, the probability of obtaining that result by chance is low enough to support the decision criterion previously defined.

How to Interpret the Results

Why Nine Correct Responses in 15 Assessors?

For a triangle test with:

• 15 valid evaluations;
• 9 correct responses;
• chance probability of 1/3;
• one-sided binomial test;

the approximate p-value is:

p = 0.0308

Since:

0.0308 < 0.05

the result is statistically significant at the 5% level.

A result with eight or fewer correct responses does not prove the beers are equal. It only means that, with that design, panel size and statistical criterion, there was not enough evidence to demonstrate a difference.

Example: Changing Hop Dosage in a Beer

A brewery regularly produced an American Light Lager using 2.20 lb of Saaz hops at the end of the boil in a 17.04 BBL batch.

The team believed that increasing the dosage could create a more intense aroma and improve the beer’s sensory perception.

Before permanently changing the recipe, the brewery decided to produce two versions:

• original recipe: 2.20 lb of Saaz;
• modified recipe: 4.41 lb of Saaz.

A triangle test was conducted with 15 valid evaluations.

The result indicates that the change was perceived under the test conditions. However, it does not answer whether consumers would prefer the new version, whether the aroma became more attractive, whether the price could be increased or whether the additional cost is justified.

The brewery then considered a second hypothesis: reducing the hop dosage from 2.20 lb to 1.65 lb per batch. In this case, another triangle test with 15 valid evaluations produced seven correct responses.

This result must be interpreted carefully. It does not prove that the beers are equal. It only means that, under those conditions and with that number of evaluations, the test did not provide enough evidence of a difference.

If the brewery wants to demonstrate similarity, a formal similarity test, a larger panel or a different statistical approach should be considered.

Decision Flow

Financial Impact of the Sensory Decision

The technical result becomes more useful when connected to the economic decision.

In this example, the brewery produces five batches per month, totaling approximately 85.22 BBL per month. The hop cost is US$13.61/lb.

Scenario 1: Increasing Hop Dosage

The modified recipe increases the dosage from 2.20 lb to 4.41 lb per batch.

Additional hops per batch:
Approximately 2.20 lb

Additional hops per month:
Approximately 2.20 lb per batch × 5 batches = approximately 11.02 lb per month.

Monthly additional cost:
Approximately US$150.00

Annual additional cost:
Approximately US$1,800.00

The triangle test showed that the difference was perceived. From a management perspective, the next questions are:

• Was the perceived difference positive?
• Do consumers prefer the modified beer?
• Does acceptance improve?
• Is the market willing to pay for that difference?
• Does the additional investment make sense for the brand positioning?

Scenario 2: Reducing Hop Dosage

The second hypothesis reduces the dosage from 2.20 lb to 1.65 lb per batch, saving approximately 0.55 lb per batch.

Monthly saving:
Approximately US$37.50

Annual saving:
Approximately US$450.00

The test did not show enough evidence of a perceptible difference. This may support further evaluation, but it should not be treated as definitive proof that the beers are identical.

Before implementing the reduction permanently, the brewery could repeat the test, increase the number of assessors, use a formal similarity test, evaluate stability over time or run the change at a controlled production scale.

The final decision should consider the sensory evidence, the economic impact, the brand risk and the production context.

Limitations of Triangle and Duo-Trio Tests

Discrimination tests are powerful, but they must not be overinterpreted.

They do not indicate:

• which beer is better;
• which beer is preferred;
• whether consumers will accept the change;
• which attribute changed;
• how intense the difference is;
• what caused the difference;
• whether the change should be implemented commercially.

The result can be affected by the selection of assessors, number of evaluations, sample control, sample homogeneity, order balancing, temperature, coding, fatigue, adaptation, carryover, environment, statistical method and the objective defined before the test.

When the objective is preference, affective methods with representative consumers are more appropriate. When the objective is to demonstrate that two products are similar, a similarity test should be planned. The absence of statistical significance in a difference test is not the same as proof of equality.

Conclusion

Triangle and duo-trio tests help transform sensory discussion into technical evidence. They are especially useful when a brewery needs to evaluate recipe changes, raw material substitutions, supplier changes, process adjustments or batch differences.

In the example, the increase in hop dosage produced a perceptible difference, but that result alone did not prove preference, acceptance, commercial advantage or financial return. The annual additional cost of approximately US$1,800.00 still needed to be evaluated against positioning and market response.

The reduction scenario generated a possible annual saving of approximately US$450.00, and the test did not provide enough evidence of a perceptible difference. Even so, that result should be treated with caution because absence of significance does not prove equality.

The value of sensory analysis is not to replace management judgment. Its value is to improve the quality of that judgment.

Instead of asking only “I thought it was better?”, the brewery begins to ask: “What evidence do we have?”

That shift is what connects sensory analysis to technical management, cost control, product consistency and safer decisions in the brewery.