“It's a bit more watery than what I originally used...” 3 Steps to Early Detection of Counterfeit Products Starting from Customer Reviews

"Something feels off" — The unexpected voice that detects fakes first

I'd like to start today's post with a story that brand managers can probably all relate to. When is the first moment you realize that a counterfeit of your brand has entered the market? In actual counterfeit response practice, the first concrete signal often starts from the voice of a customer holding and using the real product in their hands.

The case of the famous beauty brand 'Derma Factory' was no different. It was only after accumulating over ten inquiries such as "The container shape is different from usual" and "The formulation and scent don't seem like the original product" that they realized counterfeits were circulating.

Rather than the brand conducting a full market search and finding it first, they entered into full-scale confirmation thanks to a repeat purchase customer who used the product often letting them know "something is off."

If you think about it, it might be a matter of course. The person comparing genuine and fake side-by-side every day is not the brand manager, but the customers who purchase that product several times and use it in their daily life because they like it. In the end, it can only be the customer who first notices the subtle feeling when pressing the pump, the familiar scent, and the difference in texture touching the hands. In fact, it is also a part that is not easily visible from a seat inside the brand.

So, the next step is what really matters. Depending on how the brand acts after receiving this signal, the scale of the risk changes completely. Whether to dismiss one inquiry as a simple complaint, or read it as an abnormal signal indicating the start of a risk. This single judgment determines the speed of subsequent response. The difference between some brands confirming fakes weeks later and other brands moving quickly in just a few days after receiving the same inquiry comes from right here.

Then, how should we specifically capture and respond to these valuable first signals in practice? The start is, after all, the CS channel where customers' voices accumulate as text.

Step 1 — Catching Counterfeit Signals with CS Text

Actually, grandiose clues are not caught from the beginning. Looking at just one inquiry, it's easy to pass off thinking that a specific customer is just sensitive. However, if the same feedback is continuously repeated, from then on, it becomes a meaningful signal that needs to be counted as data. The text signal captured first at the customer service channel when counterfeits are circulating is this word "different." Although the expressions vary, such as "The scent is different from before," "The container is a bit weird," and "The formulation seems to have become watery," in the end, it means that the familiar product feels unfamiliar.

The problem is that these inquiries come in scattered one by one in the customer service channel. If approached only as individual cases, they are usually judged as simple complaints, handled, and then closed. As mentioned before, Derma Factory also grasped the existence of counterfeits only after accumulating over ten such inquiries. In other words, it means that if they had gathered these inquiries into one bundle and counted them, rather than letting them flow as individual cases, they could have detected the abnormality at a much earlier point.

Therefore, in practice, simply looking at the "content" of the inquiry is not enough. You also need to look at "where" this inquiry data is concentrated. The key is analyzing which specific product and during which period the sense of incompatibility felt in customer reviews is concentrated. If inquiries saying "different" have been uniquely focused on a specific product within the recent month, it is highly likely to be a clear counterfeit early warning that is hard to pass off as individual differences. In addition to this, it is also good to look at the 'temperature' of the inquiries together. If the intensity of dissatisfaction is growing beyond the level of simply "it's different" to "please refund me" or "isn't this a fake? I will report it," it could mean that the distribution of counterfeits has already progressed quite a bit even within the same SKU. It is good to look at the change in customer tone as much as the increase in the number of cases.

Looking into the reviews of unauthorized seller stores, this pattern is revealed more clearly. In the review section of external open market stores, which are not channels directly managed by the brand, doubts such as "Is this genuine?", "The scent seems a bit different," and "The packaging is a bit flimsy" are often repeatedly accumulated across multiple reviews. It means buyers are leaving traces of suspicion first, but these reviews are just outside of brand management and do not stand out easily to the manager's eyes. If the star rating is low but the reason for dissatisfaction is focused on "it doesn't seem genuine" rather than the quality or effect of the product, it is good to mark that store separately. This is because it becomes a target to be reviewed with priority in the Step 3 'Image Comparison Work' to be conducted later.

💡 If you suspect counterfeits? Try this right now

Search for keywords like "different", "weird", and "genuine" in CS records from the past 3 months and reviews from your own mall and open markets. Then check if the data is concentrated on a specific SKU or a specific period. If a noticeable concentration is found, clear evidence has been prepared to move on to the next step.

Counterfeit Response Process: Step 1: Reading signals with text → Step 2: ? → Step 3: ?

Once the suspicious product group and timing have been narrowed down through text data in this way, now it is time to check the substance of that sense of incompatibility directly with your own eyes.

Step 2 — Concrete "How it is different" from the feeling of "It is different"

From here, it's the real crossroads of practice that determines the speed of counterfeit response. Inquiries saying "different" accumulated in the CS channel only tell you the direction of the risk, but in themselves, they do not become written grounds to raise an issue with the seller or report it to the platform. This is because sensory expressions like different scents and weird touch have different standards perceived by each person and are in too subjective a domain. Since it is easy to get rejected even if you submit report documents as they are, in the end, this vague sense of incompatibility must be changed into objective differences that can be checked by eye.

Fortunately, the difference between counterfeits and genuine products actually distributed in the market is not as vague as you might think. Looking at the identification points disclosed by Derma Factory, most of the points where differences are clearly captured are concentrated in areas that can be compared by eye, such as containers, labels, and packaging. In the case of 'Niacinamide 20% Serum', the fake showed characteristics of having a longer pump neck length, a larger inner diameter, and thinner wall thickness than the genuine. The PET recycle mark printed at the bottom of the container was also a point of divergence from the genuine. The 'PDRN 4% Ampoule' showed the difference more intuitively. The unit box height was lower than the genuine, and the upper color was printed in orange, unlike the black of the genuine.

Differences may also occur in finer details. According to reports, there were cases where the text originally supposed to be 'Sodium DNA' among the fake ingredient indications was printed incorrectly as 'So.dium DNA', the packing at the mouth of the container was completely omitted, or the gloss of the dropper material was distinctly different from the genuine. Looking at them one by one, they seem minor, but if several of these non-matching items overlap, it is difficult to see it as a coincidence.

What a practitioner should pay attention to here is the fact that these differences are divided into two tracks in nature.

1. Image-type differences: Areas that are revealed immediately even with a single photo in the online sales page, such as color mismatches or print typos.

2. Physical-type differences: Areas that can be confirmed only when the product is actually purchased and received, such as neck length, wall thickness, and presence of packing.

It is good to classify in advance whether the clue we caught is a difference identifiable with online images, or a difference known only by securing the physical product. That way, the next step of gathering evidence becomes much easier.

But there is an interesting point here. While the cause of the customer first raising the issue was 'scent' and 'formulation', the clear difference the brand actually found came from the product's 'container' and 'markings.' Scent or touch is difficult to logically prove or show to a third party, but differences in container structure or print typos become definite weapons that can be shown to anyone by just taking a photo side-by-side with the genuine. So to speak, it is a work of changing subjective "feelings" into objective "evidence."

Therefore, if you have confirmed that CS signals are concentrated on a specific product, do not delay and prepare the genuine physical product of that product to compare it with the information of the suspicious product by item. Pointing out container shape, label markings, unit box color and height, and print status one by one. Rather than holding onto caught clues like scent, finding visible forms and errors in markings makes the response speed much faster. To add one more tip, I recommend leaving this comparison result as a checklist by item. If you clearly organize the caught items like "different neck length / unit box orange misprint / ingredient marking typo," they directly become the evidence list in the seller's objection or platform report documents later.

💡 If you suspect counterfeits? Try this right now

Prepare the genuine physical product of the product selected in the previous Step 1, and compare the container shape, label markings, unit box color/height, and print status with the suspicious product by item. The key is pointing out visual differences confirmed by eye instead of scent or touch.

Counterfeit Response Process: Step 1: Reading signals with text → Step 2: Comparing with actual genuine product → Step 3: ?

Step 3 — Confirming Counterfeits with AI Image Analysis

If you have come this far, you are in a state of having secured quite a lot of circumstances of counterfeit distribution. This is because you checked on which product and from when the signals were concentrated (Step 1), and checked where the containers and markings are different (Step 2). However, combining these two steps does not immediately turn into the "counterfeit confirmed" stage. To be exact, it is closer to having secured a strong circumstantial suspicion that 'the probability of it being a counterfeit is very high.'

Because visual comparison in the end only depends on human eyes and memory. Even while comparing with the genuine side-by-side, subjectivity gets mixed in the judgment, wondering "Does the neck seem a bit long?" or "Is the color subtly different?" However, to raise an objection to malicious sellers or report to the platform to take action, you must be able to objectively show "this image and that image are different by this much," not just a simple "feeling." You have to objectively confirm the risk section narrowed down by intuition again.

The final step for this is image comparison and analysis. It is collecting 'images' of the actual sales page of the product where inquiries occurred and comparing them side-by-side with the genuine official images. Recently, AI image similarity analysis, which calculates how much the design elements of two images match with values, not human eyes, is also utilized. It is a method of checking whether they are using completely identical genuine images without authorization, or subtly editing them, through a 'similarity score'. This method shows its power especially when counterfeit sellers are widely spread across multiple open markets. If you take one problematic sales image as a standard, you can find other malicious stores using the same image as it is all at once. Since fakes usually do not end with just one seller, estimating the distribution range and scale of spread based on images is also work that can be done at this stage.

Adding image comparison to the section narrowed down by text in this way, the circumstantial suspicion that "it seems different" is finally changed into a clear confirmation that "this product is different from the genuine in this part." Only then is the ground for response properly equipped. And it is also good to remember that evidence organized numerically like this is not an end in itself, but becomes data that can be directly put on the seller's objection or platform report documents.

💡 If you suspect counterfeits? Try this right now

Compare the sales image of the product where inquiries occurred and the genuine official image side-by-side. This is the step of checking metrics through AI image similarity analysis for products whose visual judgment is ambiguous, to change circumstantial suspicion into solid 'evidence.'

Counterfeit Response Process: Step 1: Reading signals with text → Step 2: Comparing with actual genuine product → Step 3: Utilizing AI image similarity analysis

Making CS the Fastest異常 Signal Detection Channel

To summarize the practical flow from catching counterfeit signs to confirming evidence, it is ultimately organized into three steps. Counting repeated "different" text signals in the customer service center and review channels based on a specific product and period (Step 1), checking specific differences in containers, labels, and markings behind those inquiries by comparing them with the genuine physical product (Step 2), and finally, placing sales images and genuine images side-by-side and confirming with similarity (Step 3) in order. It is a process of narrowing down from sensory incompatibility to objective evidence.

Going back to the beginning, the core shown by the case of Derma Factory is that it was not the brand's surveillance system but the customer who actually used the product who first recognized the counterfeit. This mechanism will not change much in the future either. If so, the CS channel, which the customer's voice reaches first, can be a channel that detects abnormal signs fastest, rather than simply responding to complaints.

Of course, one inquiry does not immediately mean counterfeit distribution. Because it could be a simple renewal or a fine difference by product manufacturing lot. However, if the perspective of reading those inquiries as risk signals without letting them flow as individual cases is established, CS plays the role of the first button of risk management. It is a matter of how quickly the brand hears the signal that the customer is already sending. Perhaps at this very moment, important signals we fail to read are accumulating in the CS channel.