‘JND’, the ‘just-noticeable (color) difference’, is a core concept in color science. Next to that one, there is also ‘JAD’: the ‘just-acceptable (color) difference’. But these are not sufficient, if you ask me. That’s why we should add SAD: the ‘smallest actionable (color) difference’. Which color difference would make shoppers leave a package on the shelves and pick another brand? SAD is distinctly different from JND, and probably also quite different from JAD. With the recent Mondelez vs Aldi lawsuit, SAD is very relevant.
CONTENTS: JND & JAD | Aren’t JAD and SAD the same? Lessons from economics vs behavioral economics | Studies on JAD | How to find the SAD? | But, is SAD real? | Don’t people notice these color differences??? | Why is this important? | Updates
Originally published: 28/06/2025 – Last update: 07/07/2025
If you look at studies on color discrimination (identifying color differences), they all more or less follow the same pattern: under standardized conditions (either a monitor or a viewing cabinet) a dozen or a few dozen people (very often students from a discipline where color is relevant) are explicitly asked if they see color differences, or they have to adjust one of two colors so they both match. Based on that, the ‘just-noticeable difference’ (JND) will be defined.
In some studies, participants were not only asked if they noticed a difference, but also if this difference was acceptable or not in that specific setting: the just-acceptable difference (JAD). But there aren’t that many studies on this. Even with the help of Google Gemini, I could only find a few. And many of them can’t be transferred to printing: e.g., the JAD of tooth color is something very different from brand color reproduction on packaging…
The study I designed a decade ago is of the few relevant JAD studies in printing. And that clearly shows there is a difference between JND and JAD. In that study, the participants (over 100, with about half involved in some way with printing or design) were not only asked if they noticed a color difference between a pair of boxes (both flat and folded), but also whether they considered the difference acceptable or not. Plus, a third question was asked: whether the difference would influence buying behavior…
Now, if you would conclude from this that JAD (the just-acceptable difference) and SAD (the smallest actionable difference) are the same, or at least very similar, there is one issue with this study: people were asked if a color difference would influence buying. They had to make a conscious decision on that. While the influence of color in shopping – and with that color differences – is something that plays on the unconscious level… Nobody who is buying a can of Coca-Cola will consciously look at the red and evaluate it based on color memory (which is flawed, BTW), or adjacent packages.
By the way, this is also a limitation of all color-related studies I’ve seen thus far: they were conducted in a controlled environment, and people were asked questions about the colors. I can’t remember seeing studies where color, and certainly not color differences, were evaluated based on observations, rather than survey questions.
Aren’t JAD and SAD the same? Lessons from economics vs behavioral economics
Here we are touching a very fundamental issue: people are not rational human beings. In an ideal world, with perfectly rational human beings, JAD and SAD could (or should) be the same. But people are not rational. And that’s what has also been wrong in economic studies for a very long time… It was only when prof. Daniel Kahneman and his group began researching ‘behavioral economics,’ and it became clear that people can respond in a way that is very different from a rational one. His work was rewarded with a Nobel Prize in Economics. His book Thinking, Fast and Slow, is highly recommended.
Prof. Dan Ariely, who worked with Kahneman for some time, showed in one of his experiments that what people say they will do (or not) can be quite different from what their actions are when they end up in that situation. In his book ‘Predictably Irrational‘, he shares a study that perfectly shows this, where people said they would never do X, but the moment the situation occurred, they acted exactly the way they claimed they would never do… (I won’t share the whole experiment here, but you can find it by reading the book.)
And this is where JAD and SAD are different: the JAD is the ‘rational’ part of a human being stating what is acceptable and what is not, while SAD is how that same human being behaves during shopping, probably with a time constraint and longing to take a bite of that special cookie. And that will lead to a different result: I’m pretty confident we are more tolerant in real life than in a JAD experiment.
Studies on JAD
When I asked Google Gemini for studies on JAD, it concluded that the studies “collectively illuminate the multifaceted nature of Just-Acceptable Differences (JADs) in color reproduction, moving beyond a simplistic numerical threshold to encompass a complex interplay of perceptual, contextual, and human factors.” Multifaceted and complex.
A little later in the report, Gemini also mentions that the reviewed studies “unequivocally demonstrate that JADs are not static values but are profoundly influenced by a complex interplay of factors”, including viewing conditions, object characteristics and content, cultural background, and observer expertise. In a shopping environment, viewing conditions are not standardized (no D50, no equal illumination, and no neutral surroundings), the placement (turned or tilted) can be an influence, and – obviously – most observers are not trained in color quality control and will have less expertise in color discrimination.
Gemini also had some advice: “Future advancements will likely involve more adaptive, context-aware metrics that integrate spatial, temporal, and semantic information to predict human acceptability better, moving towards a more holistic understanding of color quality.” And SAD is, IMO, part of that more holistic understanding of color quality…
The recent lawsuit from Mondelez against Aldi not only highlights the need for SAD but also reveals that Mondelez claims Aldi’s packaging is similar enough to trigger consumers into buying the Aldi product, thinking they are purchasing the real deal. It also shows that the SAD might be pretty high… Looking at the samples shown in the complaint, even a 10 dE00 difference in color will still confuse consumers and lead them to believe the Aldi copy is the real deal… So the SAD might be a double-digit delta E… Quite far from what some people in the industry are claiming…
To summarize:
- JND is neutral and binary: you either see a difference, or not
- JAD is not neutral: it’s a judgement, an opinion, an emotion, but also nothing more than that
- SAD is consequential: this will impact sales.
How to find the SAD?
Well, that’s a good question… The easy answer: not the way JAD is researched. The Gemini report provides an overview of psychophysical methodologies used in JAD studies, including direct scaling or rating, paired comparison, threshold determination, and rank ordering. All of these methods involve conscious decisions. Not a single study is based on observations of humans during shopping, when we often act on unconscious impulses. In case you haven’t read it yet, here’s an introduction to the System 1 and System 2 thinking, concepts from behavioral economics that explain a lot about our irrational behavior. And as I explain in that article, quality control in a press room involves clear System 2 thinking, whereas shopping is mainly driven by System 1.
It is essential that people are not aware of the fact that they are part of a study on color, a study aimed at finding the SAD. So, any questions about color and its influence on purchasing products are prohibited. These would not give a reliable result. If you ask someone if color is important, they will probably answer positively: why would you otherwise do a study on that? Participants are looking for the expected behavior and want to affirm this…
(message to my 10-year-younger self: nice try, that third question, but the results are not what you intended to find)
To demonstrate the importance of using the proper wording, consider the results of this brief poll from my presentation at the FOGRA colour management café. I asked the participants (all printing and color professionals) when they would consider buying a different brand than their favorite one. And NOT ONE person picked the ‘looking a bit different’ option. Just by using that more generic term (which does include color differences), nobody picked that one. If ‘visible color differences’ had been an option, I bet multiple people would have picked that one. Especially the attendees selling color solutions…
In a lab test, the SAD should be investigated indirectly. E.g., trying to find a setting where packages are shown but without any reference to color or visual properties. Perhaps experiment with the packaging design? Or have a series of refreshments on the table and see which ones are taken, which ones are consistently left behind. But I admit, it’s not an easy one to set up a good test for this.
But the best way would be to observe people in a real-life setting, like a supermarket. And, if relevant, conduct post-purchase interviews. Some things that could be checked: What are the colors of the packages on the shelves? Are there differences? If so, which packages will stay on the shelves? Is this linked to the color? How long is the interaction time with an individual package? If it’s been a long time and a package is put back on the shelves, ask that person why they put it back, using an open-ended question, again without mentioning the color.
With these two different approaches, we may also want to distinguish between SAD-Lab (tested in a lab environment) and SAD-IRL (tested in real life). These could be different, depending on how the SAD-Lab is tested.
BTW: there will not be just one SAD for all print products, it will depend on the application. The SAD for the packaging of a luxuary product will probably be lower than the packaging for breakfast cereals or a soda can. But: this needs to be tested. And we will need a lot of testing, because SAD will be very case specific: for a paint job in a car body repair shop, the JND, JAD and SAD will be very close, if not identical. We should not hold the printing of a disposable package to the same standard as such a paint job…
If you are an academic researcher or a student looking for a thesis subject, I would be happy to assist you in further exploring this! Feel free to contact me! I don’t have the means to execute an extensive study myself, but I would be happy to assist others and provide feedback.
But, is SAD real?
I can imagine that some people involved in color quality (in print) on a daily basis may have difficulties with this new concept, specifically the Smallest Actionable Difference. So, let’s take a few examples from real life, packages I bought over the years.
This first one is fun: I saw these when I went shopping just before going on holiday with my then-girlfriend. I immediately noticed the color difference (since the color part of my brain is usually activated when I enter the cookies aisle), but I wanted to check if she noticed it. And guess what: she only noticed it when I explicitly asked her if she noticed a color difference. Not when I asked her if she noticed something, not when I asked her if she noticed a difference… That’s how I came up with the ‘uncertainty principle of visual color evaluation’.
These two packages of Dash were purchased with a year or so difference. When the first one was almost empty, I purchased a new one and placed them next to each other on top of the washing machine, which is located in a spot that I pass by a few times a day. And it was only after over a month (!) that I noticed the difference in the blues. And for the record: the most vibrant one is the oldest one.
And check out this example I already discussed last year, with Dallmayr coffee.
Or these packages of different LU Prince products…
And what about these Oreo packages, where both old and new logos, with a 10 dE00 color difference, were used.
Some of these are interesting, because several people have handled these products. The people responsible for putting the primary packages in the secondary packaging (when applicable), the people putting the products on the shelves, plus their supervisors or the shop owners. And none of those seems to have made an issue out of this difference. Maybe they even didn’t notice it…
And in the Oreo case, when the marketing and/or product managers decided to go for a different blue in the new logo, they didn’t mind… Plus, as I showed in the Mondelez vs Aldi article, Mondelez defines the Oreo brand color as ‘blue’. No Pantone number, no CIELab or any other more precise color definition. It’s defined as blue in the lawsuit and the registered trademark.
Don’t people notice these color differences???
In a previous article, I also showed why people in a shopping environment don’t seem to notice small color differences. Here’s the short list. For more information and links to studies, please check the full article.
- Color control is a special skill; you need training
- Our color memory is poor, even in an ‘undelayed’ comparison
- A supermarket is very different from a press console:
- Lower light level and probably not D50
- Not flat
- Samples are not placed on top of each other for close evaluation
- Cluttered background
- Very different timeframe (usually <1 second ‘Total Fixation Duration)
- Predictive processing: how our brain processes input
- Framing
Why is this important?
Color differences can be consequential. If a product doesn’t look genuine, consumers might decide not to buy it, leading to missed sales. But at what point doesn’t a product look genuine anymore? That’s a difficult one.
If you can’t see a difference, it’s genuine, that’s easy (unless it’s counterfeited, of course). However, when you notice a difference, it may have an impact on sales. But not with every color difference. A ‘just-acceptable difference’ (JAD) was invented to provide an answer to that. However, the number of studies on JAD is limited, and those I found were all conducted in a laboratory environment where participants had to evaluate the colors consciously. Which is not the case when we are shopping: that’s (mostly) unconscious.
The Mondelez vs. Aldi case underscores the core issue: what is the threshold below which a package is still considered a Mondelez product, and above which it is no longer? That’s the ‘smallest actionable difference’ (SAD) we need to find. And based on the examples in the Mondelez complaint, that could even be double digit (10 dE00)… Quite far from the 2 dE00 that some people are pushing.
Imagine if the SAD were indeed what Mondelez implicitly claims, how much waste, both in time and materials, could be saved if print tolerances reflected SAD instead of JAD. It would be a joyful day for many printing companies…
Disclaimer: Once again, for the record, I’m not promoting bad print quality. Printing companies should adhere to the applicable ISO standards. Best practices, such as PSO/PSD and G7, are great tools for achieving those standards. And printing companies should use the tools that are available to get their print production stable and within a certain margin, to minimize waste (both economic and ecological).
What I am promoting is being more reasonable when it comes to tolerances in print. Tiny tolerances are not necessary; the Mondelez case shows it. Tiny tolerances are hurting printing companies badly.
PS: I also asked Google Gemini to make an assessment of the scientific merits of this new concept, the Smallest Actionable Difference. This is the full report by Gemini. Spoiler alert: it has scientific merit, and it has the potential to advance color science.
UPDATE 07/07/2025: I had contacted John – The Math Guy – Seymour to pick his brain on SAD. And as expected, John did a thorough job, by pointing out an issue with JND… Here is part of our conversation that is relevant to JND:
“There is a bit of an issue with the term JND. It doesn’t affect the point of your blog, but I will correct you just as one eminent color scientist corrected me on this years ago. Danny just scolded me not to use JND and DE interchangeably, so I had to do some reading.
David MacAdams did an experiment a long time ago that determined some tolerance regions in the chromaticy diagram. In this paper, he defined the term/abbreviation JND.
His experiment went like this. His assistant did color matches. A whole lot of them. The assistant had a color to match and had a knob that would move another color around along a straight line in chromaticity, with some point along that line being a perfect match. The assistant was to adjust the knob until he got a match, record the results, and then mess up the knob. He did this maybe 50 times?
Then the line in chromaticity space was rotated. Where the first line went reddish/cyanish, the next line might be orangish/bluish. The assistent did the same test again.
Not knowing how to do the 3D stats on his data, MacAdam took the standard deviation along each line. He called this the “just noticeable difference”. There was some work done trying to play with these numbers to determine tolerance limits for any color, but our current deltaE was not based on this.
So a JND was defined by MacAdam as a measure of how close this assistant could get to the color and think the colors are the same.
All of the modern deltaE equations are based on data sets from a different sort of experiment, one where the subject is given a pair of samples and asked to rate the degree of match on a scale of maybe 0 – 10. I think they might be given a pair of gray values and told that these are one unit apart. This calibrates the subject.
I think that all the modern color difference equations have been calibrated so that 1.0 DExx around L*=50 a*b*=0, is the same as that of DEab.
When a color difference equation is tested against one of the data sets, they use a metric that they made up called STRESS. The first step is to determine the best scaling factor to go from calculated DE to visual assessments.
I am not sure that I have all the details down, and (if I have them right) I am not sure I understand what it all means for the relationship between JND and DE. But… I heard this from one other eminent color scientist. He said that there is a factor of about 1.5 between the two. I think he meant that 1 DE is about 1.5 JND.”
If you don’t know John, here is his blog. He really knows the details about color, including the history of color science. E.g., check out his ‘CIELAB is Icky’ video.
Be the first to comment