Logo testing before launch: methods and metrics
Why an approved design is not a guarantee of success
A designer sees the logo with a practiced eye and evaluates the composition, balance, and style. A consumer sees the logo on the run — in a feed, on packaging in a store, on an app icon among a dozen other icons. These are two different modes of perception, and the gap between them is the main reason why logos that look perfect in presentation fail in real use.
Testing closes this gap. But there is a common mistake here: the logo is tested at the final stage, when the options have already been narrowed down to one or two, and the budget and time for changes are almost exhausted. The correct entry point for testing is the concept stage, when there are still 3-5 directions on the table and the decision on selection can really be adjusted based on data, not on the art director's intuition.
Monadic or comparative testing — what is the difference
This is the first fork question in the methodology, and the answer determines what data you will get as output.
In monadic testing, each respondent is shown only one logo—without competitors and without alternative options nearby. The respondent evaluates it on its own, which removes the effect of comparison and provides a more honest reaction to the brand perception 'in a vacuum,' that is, as a real buyer would see it, without keeping alternatives in mind. One participant evaluates no more than one or two logos per session—showing them ten variants at once would cause the evaluations to shift relative to one another, and you would get ranking rather than an absolute reaction.
Comparative testing, on the contrary, is built on the direct confrontation of options: the respondent sees several logos at the same time and chooses the best one according to the specified criteria.
| Parameter | Monadic | Comparative |
|---|---|---|
| What is shown | One logo per respondent | Several options side by side |
| What it measures | Absolute perception, the reaction of 'as is' | Relative superiority of one option over another |
| When to apply | Final check before launching one approved option | Choice between several concepts at an early stage |
| Risk | Higher cost — requires a larger sample for each option | Distortion due to the comparison effect, the winner may simply be 'less bad' |
Practice: at the concept selection stage, a comparative method is needed, it will faster reveal the favorite. At the final check stage of the selected logo — monadic, it will show how the brand will be perceived in reality, without competitive context.
Which metrics are actually needed and which are just clutter
The temptation to measure everything at once is a big mistake: a overloaded questionnaire reduces the quality of answers and blurs the focus of the study. There are four metrics that provide real diagnostic value:
- Recall — the ability to recall the logo after a short distraction of attention; shows the strength of memorization in memory;
- Recognition — recognition without hints, when the logo is shown among other visual stimuli;
- Category fit — logo correspondence to the industry: is it perceived as a logo of a bank, cafe or IT company, does it meet expectations of the category;
- Sentiment analysis — emotional response, the ratio of positive and negative associations.
If the research budget is limited, choose metrics according to the task, not according to the module. Launching a brand in a new category — category fit is critical. Entering a saturated market with strong competitors — recognition comes to the fore. Metrics such as 'awareness' and 'emotional response,' which are passed from article to article as a mandatory set, without being tied to a specific brand task, become meaningless numbers.
Example from practice: a regional coffee chain conducted monadic testing of a new logo on 120 respondents. Recall stood at 72% — a good figure in itself, higher than the industry benchmark of 60-65% for the food service segment. However, sentiment analysis showed 15% — a noticeable tilt toward the negative. Analysis of responses revealed the cause: the strong typographic element was memorable, but the icon was perceived as 'cold' and 'corporate' for a coffee shop. The solution was not to redesign the concept, but to refine the visual hierarchy — soften the icon's form and shift the emphasis to the font, which actually provided the recall. After refinement, sentiment moved into the positive zone while maintaining the same level of memorability.
Paul Rand's test as a quick quality filter
It is worth keeping Paul Rand's test on seven criteria separately in the arsenal — a classical methodology for design evaluation, not the result of quantitative audience research. It does not replace surveys of real consumers, but quickly filters out weak variants before budget is spent on research. A logo is evaluated on uniqueness, visibility, flexibility, memorability, universality, timelessness, and simplicity. Six criteria are scored from 1 to 10 points, simplicity is scored separately, from 1 to 15, for a maximum of 75 points total. A result below 60 points is a signal that the logo is not ready for launch, even if it visually pleases the team.
How many people to survey and who exactly
There is no universal number of '100 respondents is enough' — the sample size depends on how subtle differences you want to catch. For the qualitative stage (focus groups, in-depth interviews) 15-20 people per segment is sufficient, this is enough to reveal obvious problems in perception. For the quantitative stage, where a statistically significant difference between variants is needed, the minimum threshold is 100-150 respondents per tested logo. This is not an arbitrary figure: with a 100-person sample and 95% confidence level, the confidence interval is approximately ±7-8 percentage points, and with 150 — it narrows to ±6-7. This means that a difference of 5-7 points between variants in a 100-person sample might fall entirely within the margin of error and not indicate a real advantage of one logo over another.
The audience for the test should mirror real brand buyers, not 'people in general'. If a brand is aimed at teenagers, testing on a 35+ audience will yield pretty but useless data. It is optimal to divide the sample into 2-3 segments (for example, current customers and potential new audience) and compare results between them separately, rather than averaging them immediately.
Where and on what to test the logo
A logo that reads well on mockup A4 often loses its shape on a smartphone screen or becomes a blob on a small app icon. Therefore, testing should be conducted not on one 'ideal' mockup, but across several contexts of use:
- large format — outdoor advertising, signage, banner;
- medium format — packaging, business card, printed materials;
- small format — application icon, favicon, avatar in social networks;
- monochrome rendering — check for case of printing in one color or fax/stamp;
- animation or static logo on screen — if the brand is digital.
The difference in ratings between these formats is not an anomaly, but a normal part of the data: it shows where exactly the logo is weak, and often suggests a specific technical solution (simplify the detail, increase the contrast), rather than the need to change the concept entirely.
Platforms and testing formats: how to choose a channel
A metric without an appropriate data collection channel is worth nothing — an in-person focus group and an A/B test in social media give fundamentally different types of information, and mixing them up is the second most frequent error after testing at the final stage.
| Channel | Pros | Cons | What it shows |
|---|---|---|---|
| In-person focus groups | Depth of responses, you can ask 'why', you can see live reactions | Expensive, small sample size, moderator and group influence on opinion | Reasons for perception, emotional triggers, but not statistics |
| Online surveys (quantitative) | Large sample size, cheaper per respondent, easy to calculate metrics | No depth, respondent may answer carelessly |
Why an approved design is not a guarantee of success
A designer sees the logo with a practiced eye and evaluates the composition, balance, and style. A consumer sees the logo on the run — in a feed, on packaging in a store, on an app icon among a dozen other icons. These are two different modes of perception, and the gap between them is the main reason why logos that look perfect in presentation fail in real use.
Testing closes this gap. But there is a common mistake here: the logo is tested at the final stage, when the options have already been narrowed down to one or two, and the budget and time for changes are almost exhausted. The correct entry point for testing is the concept stage, when there are still 3-5 directions on the table and the decision on selection can really be adjusted based on data, not on the art director's intuition.
Monadic or comparative testing — what is the difference
This is the first fork question in the methodology, and the answer determines what data you will get as output.
In monadic testing, each respondent is shown only one logo—without competitors and without alternative options nearby. The respondent evaluates it on its own, which removes the effect of comparison and provides a more honest reaction to the brand perception 'in a vacuum,' that is, as a real buyer would see it, without keeping alternatives in mind. One participant evaluates no more than one or two logos per session—showing them ten variants at once would cause the evaluations to shift relative to one another, and you would get ranking rather than an absolute reaction.
Comparative testing, on the contrary, is built on the direct confrontation of options: the respondent sees several logos at the same time and chooses the best one according to the specified criteria.
| Parameter | Monadic | Comparative |
|---|---|---|
| What is shown | One logo per respondent | Several options side by side |
| What it measures | Absolute perception, the reaction of 'as is' | Relative superiority of one option over another |
| When to apply | Final check before launching one approved option | Choice between several concepts at an early stage |
| Risk | Higher cost — requires a larger sample for each option | Distortion due to the comparison effect, the winner may simply be 'less bad' |
Practice: at the concept selection stage, a comparative method is needed, it will faster reveal the favorite. At the final check stage of the selected logo — monadic, it will show how the brand will be perceived in reality, without competitive context.
Which metrics are actually needed and which are just clutter
The temptation to measure everything at once is a big mistake: a overloaded questionnaire reduces the quality of answers and blurs the focus of the study. There are four metrics that provide real diagnostic value:
- Recall — the ability to recall the logo after a short distraction of attention; shows the strength of memorization in memory;
- Recognition — recognition without hints, when the logo is shown among other visual stimuli;
- Category fit — logo correspondence to the industry: is it perceived as a logo of a bank, cafe or IT company, does it meet expectations of the category;
- Sentiment analysis — emotional response, the ratio of positive and negative associations.
If the research budget is limited, choose metrics according to the task, not according to the module. Launching a brand in a new category — category fit is critical. Entering a saturated market with strong competitors — recognition comes to the fore. Metrics such as 'awareness' and 'emotional response,' which are passed from article to article as a mandatory set, without being tied to a specific brand task, become meaningless numbers.
Example from practice: a regional coffee chain conducted monadic testing of a new logo on 120 respondents. Recall stood at 72% — a good figure in itself, higher than the industry benchmark of 60-65% for the food service segment. However, sentiment analysis showed 15% — a noticeable tilt toward the negative. Analysis of responses revealed the cause: the strong typographic element was memorable, but the icon was perceived as 'cold' and 'corporate' for a coffee shop. The solution was not to redesign the concept, but to refine the visual hierarchy — soften the icon's form and shift the emphasis to the font, which actually provided the recall. After refinement, sentiment moved into the positive zone while maintaining the same level of memorability.
Paul Rand's test as a quick quality filter
It is worth keeping Paul Rand's test on seven criteria separately in the arsenal — a classical methodology for design evaluation, not the result of quantitative audience research. It does not replace surveys of real consumers, but quickly filters out weak variants before budget is spent on research. A logo is evaluated on uniqueness, visibility, flexibility, memorability, universality, timelessness, and simplicity. Six criteria are scored from 1 to 10 points, simplicity is scored separately, from 1 to 15, for a maximum of 75 points total. A result below 60 points is a signal that the logo is not ready for launch, even if it visually pleases the team.
How many people to survey and who exactly
There is no universal number of '100 respondents is enough' — the sample size depends on how subtle differences you want to catch. For the qualitative stage (focus groups, in-depth interviews) 15-20 people per segment is sufficient, this is enough to reveal obvious problems in perception. For the quantitative stage, where a statistically significant difference between variants is needed, the minimum threshold is 100-150 respondents per tested logo. This is not an arbitrary figure: with a 100-person sample and 95% confidence level, the confidence interval is approximately ±7-8 percentage points, and with 150 — it narrows to ±6-7. This means that a difference of 5-7 points between variants in a 100-person sample might fall entirely within the margin of error and not indicate a real advantage of one logo over another.
The audience for the test should mirror real brand buyers, not 'people in general'. If a brand is aimed at teenagers, testing on a 35+ audience will yield pretty but useless data. It is optimal to divide the sample into 2-3 segments (for example, current customers and potential new audience) and compare results between them separately, rather than averaging them immediately.
Where and on what to test the logo
A logo that reads well on mockup A4 often loses its shape on a smartphone screen or becomes a blob on a small app icon. Therefore, testing should be conducted not on one 'ideal' mockup, but across several contexts of use:
- large format — outdoor advertising, signage, banner;
- medium format — packaging, business card, printed materials;
- small format — application icon, favicon, avatar in social networks;
- monochrome rendering — check for case of printing in one color or fax/stamp;
- animation or static logo on screen — if the brand is digital.
The difference in ratings between these formats is not an anomaly, but a normal part of the data: it shows where exactly the logo is weak, and often suggests a specific technical solution (simplify the detail, increase the contrast), rather than the need to change the concept entirely.
Platforms and testing formats: how to choose a channel
A metric without an appropriate data collection channel is worth nothing — an in-person focus group and an A/B test in social media give fundamentally different types of information, and mixing them up is the second most frequent error after testing at the final stage.
| Channel | Pros | Cons | What it shows |
|---|---|---|---|
| In-person focus groups | Depth of responses, you can ask 'why', you can see live reactions | Expensive, small sample size, moderator and group influence on opinion | Reasons for perception, emotional triggers, but not statistics |
| Online surveys (quantitative) | Large sample size, cheaper per respondent, easy to calculate metrics | No depth, respondent may answer carelessly |
Recent articles
MegaFon, BMW, VTB and more 240+ companies. See how it looks in the showcase reel:
Our branding agency is approached to order:
Branding agency services
