Your Product Launch Is Probably Missing This Critical Data

If you just take that at face value, it's like, okay, hey, boom, that's a one star review. But the thing is, is there's a lot of rich information that has been shared in that one review. So we know the reason for the one star review is because of packaging and delivery. But what do we also know about the consumer based on this? That they're a parent, that they have kids, that they're a brand loyalist, that they're complementing the quality of the product. So, so really what it is about enabling is taking every individual data point and tagging as much information as possible on it based on what is being kind of deduced. And then what that helps with is when you now do that across tens of thousands of data points, what it gives you the ability to do is use all of these individual tags to answer various questions. You're listening to Food Tech Talk Supply Chain Insights from Farm to Fork, a podcast discussing the latest trends and technologies in the food and supplements industries. Brought to you by TrustWell, featuring conversations with regulatory experts, quality and safety champions and thought leaders across the industry. I'm your host, Katie Jones, CEO at TrustWell. If you're loving the conversation, don't forget to subscribe and stay tuned to the end of the episode where you'll hear our guests tell us their key food industry predictions for the next 12 months. All right, let's jump right in on today's episode of Food Tech Tips Talk Supply Chain Insights from Farm to Fork. We're joined by Gautam Kahnamaru, co founder and CEO of Yogi. Gautam's background spans natural language processing at Microsoft, advanced AI development, and now helps consumer brands transform massive amounts of unstructured software feedback into actionable business intelligence. Through Yogi, he's helping companies move beyond siloed data to better understand reformulation risk, consumer sentiment and innovation strategy at the SKU level. Today we'll explore how AI is reshaping product development, what Amazon's achieving review structure signals for the future of consumer data, and why brands need to rethink where they invest their innovation resources. Welcome. Thanks for having me. Excited to be here. So we'll start first with what sparked your interest in natural language processing and AI. Everybody's on the bandwagon now. You've have an extensive amount of time and spent a lot of time in your career on natural language processing and AI. What sparked that first interest for you and then how has evolved as you're working solve problems for consumer brands? Yeah, you can say I've been Definitely on the bandwagon for a while. Worked at Microsoft outside of college and I mean, they've been working on AI technology since they started the company. And so I think it's taken a lot of iterations to kind of get here. But the reason for the interest in natural language processing, I think what's fascinating to me has always been so much of work and knowledge ends up being transferred and just shared in text. And so natural language processing was always the kind of thing where if you could have a 5% or even 10% improvement on what it takes to process text in a better way, the amount of downstream impact that has just across any industry you can think about is orders of magnitude. And so I think it was one of those like single choke points, if you will, where any improvement there has a lot of downstream impact at the end of the day. And the reason that we ended up falling kind of on consumer goods is I think what's really interesting is these consumer goods companies are some of the most well known and prestigious brands in the world. Right? Like Coke and Pepsi and Colgate and Apple and Microsoft. Everybody knows these companies across the world and they're very, very consumer facing and have to be super, super consumer oriented. But it is very difficult to really understand what going on and what people are talking about for these companies. And so I think that's where we kind of ended up falling into this industry to focus on. Now let's talk about a topic that is very top of mind for our customers, something that we are dealing with from an ingredient perspective, but reformulation. So reformulations can be very risky for very established brands. But when you think about unified consumer data, how does that data help companies protect trust while still being able to innovate? Right. While still being able to drive major product changes, whether that's a reformulation in taste or color, really any kind of element that consumers have a lot of affinity for and a lot of brand loyalty too. Yeah. Whenever you're looking at, specifically at a reformulation or renovation, you're touching something that's, for lack of a better term, coming off the top of my head, something sacred. Right. Like something that hundreds of thousands of people are used to using, maybe used to using on a daily or weekly basis. And so whenever you're going to touch something there, it is very important for you to. Whenever these companies are making these decisions, again, depending on what their move is, what the strategic move is, they're maybe trying to make it so that nobody will notice, or maybe there will be some notices. But we expect it to be a little bit more sweeter. You have these hypotheses essentially when you're building your reformulation or renovation and you're going out with it. And knowing whether you hit on those hypotheses or not is extremely important. And it's very important to know it as soon as possible. Because if you put yourself in the mind or the situation of a consumer, right, if I go, let's just take the new Coke example. If I go and I buy Coke on my weekly or bi weekly grocery trip and then one random time it tastes different, chances are I may not give Coke a second chance or I'm going to give it one more chance and say, hey, maybe that was a bad batch. And so there's a very small window for you to either satisfy those people or bring them back. And so I think this is where being able to understand all of the sources of consumer feedback in one centralized place makes a huge difference. Because you don't know where people are going to talk about what their experience is. It's not guaranteed that they're going to call into the call center or write a review or write a post or put it somewhere else. And so you really need to be listening everywhere to understand where this is. So that you can not only see, hey, are we hitting on our hypotheses that we had, but you're able to react fast. And we've had customers who, they go in with one hypothesis, it turns out not to hit. And luckily they've been in a position where they can revert back to the old formula relatively quickly to kind of reassess their plan, which is maybe the nuclear option in these cases, but something you do need to consider because once market share walks out the door, it's very hard to get it back. Looking to improve your supply chain at TrustWell, we've been connecting the dots between food industry and data for over 40 years to give you more control and visibility. Our comprehensive food logic platform sets a new standard for compliance, transparency and quality in the food industry. Request a free demo. The link is in. The show notes Such a good point around the nuclear option because there's so much time and effort and expense that goes into that process when it's important to have that single source of truth right in terms of the feedback and one centralized area that you have all of that information for the product teams, what does that look like? Does that data and how does that input help improve the decision makings across all of the different teams? When you are working on a Free formulation. It's a great question. I always tend to think of these in a visual manner, so I'll try to explain kind of the way it breaks out. But it's essentially one of those charts where you have multiple inputs going into a single, let's just say choke point and then multiple things coming out of it. I'm sure everybody's kind of seen this almost like spider web or wormhole diagram. And usually what it ends up being at the end of the day is kind of like we were talking about the, with the reformulation example. You need to be listening across a multitude of different sources, right? Because every customer or every consumer is different. They're going to give you feedback in different areas. And so making sure that all of that is being put together into one centralized place ends up looking like kind of different. But it usually ends up becoming one whatever database or software system where all of this information is being put in. And there's different levels of quality to that, right? Like most big companies now have a data warehouse where they're putting all of their data in. But then there is something different about being able to cross correlate, oh, hey, here is the product listing on Target and here's that same SKU number in our call center and here's that same mention in a social post. So there is like quality and nuance that goes into kind of putting those together. And then the output of all of this is all the different teams accessing the same system to answer the question. And this is where it starts to become super important. Because a lot of consumer goods companies out there, what tends to happen is like marketing will see one thing because they pay attention to social data and then they'll bring that up to the product team maybe, but then the product team will say, hey, but when I look at call center data, nobody's complaining about this, so therefore it's not an issue. Right? And then an E Commerce person will be kind of on the side being like, oh, hey, our star rating went from a 4.5 to a 3.7. And then they don't know who to share that with, right? So if everybody's talking from a different angle, source of truth, at the end of the day, it's almost like somebody speaking German while the other person's speaking Spanish while the other person speaking English, they could all be talking about the same problem. But things are being lost in translation at the end of the day. And so really having it so that the entire organization is coming to the same source of truth means that Everybody is speaking the same language that when marketing is coming in and saying, Hey, 20% of our complaints are talking about taste. That somebody on the R and D team can go back and say, oh, I see how they got to that answer. I believe it. And here's an extra nuance that I see. And I think that is what makes the biggest difference. Because I think the important thing to note or that I think I've come to appreciate is like, everybody in the organization is incentivized to build the best product and the best experience. It's just that everybody has a different path and viewpoint on how to get there. And really making it so that everybody's path, at least from the data analysis perspective, overlines, really helps the org as a whole become kind of more aligned and calibrated at the end of the day. I couldn't agree more. We do a lot of work in the cross section of R and D and food safety. Right. And how can those functions, who I firmly believe across the entire organization, want to provide the best, safest product, but in the, you know, speed to market, that's where things can break down. Right. And that access to information about the safety and verifiability of suppliers. Right. Is where things disconnect. So, yes, could not agree more. I always think that the best intentions are there, but when you don't have that connectivity of data, it makes making decisions very hard. So you talked about reviews, and I think it's interesting because there could be varying levels of consumer action based on an experience, so. Right. So in your example, they may not be calling the call center because it's not, you know, it hasn't gone from a 4, 2 to a 1. That might have elicited more of a response from the consumers, but it went to a 3, 5 or whatever it was, and that's not good enough. When we think about reviews, can you educate our listeners a bit on how Amazon has shifted their review process and what's going on there? So the way and people who have shopped on Amazon, I'm sure, are used to this experience. Well, you'll go onto a product page and there'll be two to five to 10 different variations of that product that you can kind of click through. Right. A classic example is let's actually take two separate examples. You have a shoe that comes in a black, white, green, yellow, purple color. Right. So you click in and you can pick the color you want, add it to your cart, and then on the other end you'll have a beverage, let's say it's like Capri sun or something like that where you have like the six or seven different flavors that you can click through and hit submit. So the way that reviews and ratings, or at least like the top line star rating used to work in Amazon is you can kind of think of it as essentially the average of all of those variations together. And so what this would mean is like if you saw a 4.6 star rating for Capri sun, it would kind of be the average of all of those six or seven variations. But oftentimes what wasn't unusual is for let's say their most common flavor, their cherry flavor, to have 95% of the reviews and that is the one out of four, six. And then this long tail of other variations might only have 10 or 20 reviews each and maybe they have like a 5 star rating or a 3.0 star rating. So pretty much what some people would do, maybe some more nefarious players, is they would purposely group some of their lower performing products on the same page with their best performers or put an innovation product in the same page so that automatically they're already at a 4, 6 or 4.7 star rating. The change that Amazon made I think this earlier this year, I think it was, it was kind of a rolling update, is that now when people select the variations, the star rating will change accordingly. So even if the cherry flavor has a 4, 6, if you go switch to the new innovation one, you'll see that one only has 10 reviews and is maybe at a 4.2 or something. Now if you think about this generally as a consumer, this is actually for their benefit, right? It is giving a more accurate view of what people are saying and stuff like that. It is just a big behavior change for manufacturers and E commerce teams because this is how it used to work. But I think the reasoning behind this is probably kind of like twofold on the Amazon end. One is maybe try to protect the consumers more, be more consumer forward, consumer centric. But then the second is as they are getting more into AI and agentic shopping and agentic commerce and everything like that, being able to share cleaner and cleaner data with the models downstream that are giving the suggestions is more beneficial and more useful, right? Like everybody's heard the garbage in, garbage out scenario that is still prevalent no matter how good an AI gets. And so it kind of falls along those lines as well where it's like cleaner breakdown can be shared with these LLMs at the end of the day so that they can give their suggestions. Oh, that's Interesting. I think that to your point, best for the consumer promotes transparency. A bit of a shift in terms of how brands need to think about the aggregate review compared to the individual, but also more data points to understand exactly where they fall in terms of the product reviews. Consumer reviews are messy and emotional by nature. Right. It's something, I mean, especially in the field that we work in, predominantly in food. So how does it work AI separate meaningful product signals from noise. So it's very important to kind of listen to all the signals, but it's also important to be able to dissect the signals and understand what's actionable and what's perhaps not. The best way to kind of think about this is like almost doing as much tagging as possible on individual data points. So just to give a sense of like, if you saw a review that said something like, let's say it was a one star review where somebody's like, hey, I've been purchasing, let's just say it's a, like a bread brand or something. So I've been purchasing this bread for 20 years. My mom used to purchase it for me as well. It's always the first thing I pick up when I go to the grocery store. I use it to make the sandwiches that I give to my kids when they head off to school. Always found that like the quality is good. It's always tasted the same. Like, huge, huge fan of this brand. I'm just disappointed that it came squished when it was delivered one star. So the thing is, is like if you just take that at face value, it's like, okay, hey, boom, that's a one star review. But the thing is, is there's a lot of rich information that has been shared in that one review. So we know the reason for the one star review is because of packaging and delivery. For example, maybe it's just purely delivery. So we can, we can tag that. What do we also know about the consumer based on this, that they're a parent, that they have kids, that they're a brand loyalist, for example, that they're complimenting the quality of the product. And so there's all of this other information that's been shared as well. So really what it is about enabling is like taking every individual data point and tagging as much information as possible on it based on what is being kind of deduced. And then what that helps with is when you now do that across tens of thousands of data points, what it gives you the ability to do is use all of these individual tags to Answer various questions. So now what it can be is like, hey, what are the top five issues that have been popping up over the past month that were not popping up the previous month? And now you can kind of use the AI system to be like, okay, let's isolate all the negative complaints or the negative mentions within this larger content. Let's understand what those issues were. Let's compare them to May, since it's June currently, and do that analysis. Or, hey, what percentage of people that write feedback do we know for a fact? Or parents, for example? It's like, okay, hey, now we know to isolate kind of a different set of mentions. So it really is about kind of the thing that comes back to like the fascination about natural language processing is like, there are so many variables when it comes to text in terms of what people are saying, how they feel, sentiment, emotions, this, that. So it really is about like boiling that infinite set of possibilities into a set of like, buckets, if you will. And I think as AI gets higher and higher from a quality perspective, there's more and more buckets that you can use to break down information, which just means you can answer more questions downstream at the end of the day. Yeah, and do it faster. The text review. Right. Nothing landmark about that, but the ability to then process massive amounts of that text data, to then extract the contextual themes of it in a very short amount of time. I'm just incredibly powerful for brands to have access to that information. So we are both CEOs of tech companies and I'm sure that you address this question, you know, with your customers or with your teams probably as much as we do. There seems to be brands debating whether to build internal data systems or to buy specialized tools. And I think that AI has helped increase that proliferation. Right. Like, it's the massive amount of, you know, vibe coding that could happen within an organization. When you counsel your customers and leaders in the space, what framework should they use to help make that decision? There's two important dimensions that I think are maybe the two most underappreciated when these decisions are being made. I think the first one is understanding, like, I think it's called like the Pareto principle, essentially, where like 20% of the effort gets you 80% of the way there. But really when it comes to high quality software output, at the end of the day, it really is that last 20%, that last mile, where true insight and quality gets made at the end of the day. And so, I mean, vibe coding and cloud code and Codex and all that stuff is Amazing, amazing tools. I'm a coder at heart, so the amount that you can output is, is great. But I think the thing that a lot of maybe executives don't appreciate when you haven't been in the software world for very long is like getting the MVP or prototype is just step one of a 30 step journey. There is the maintenance of the tool, right? What happens when it goes down? What happens when this data source changes its format? What happens when, I don't know, a feature request comes in from internally within your team? Like how do you deal with all of that? And I think starting to appreciate that that maintenance is going to take headcount and effort at the end of the day that you have to process when you're doing your ROI calculation, it is no longer like, oh hey, I was able to get this answer from this software system. Now I can vibe code my own version of that software system. Okay, cool. Hey, I can get that same answer. But does it still hold the test of time two months from now, four months from now, six months from now? Because whenever you put something out there, especially from a coding perspective, the half life, if you don't maintain it is very aggressive at the end of the day. So I think that is one important consideration point. And then the second is like also just really looking internally and gaining an appreciation for what can we do best. Because there are some things that these consumer goods companies have knowledge of internally or context of or historical perspective on that nobody else has. And if they can utilize these tools to really focus on building that intel and that know how into some systems, that is I think where the opportunity is like there's one or two customers that we have that I think are doing this like build versus buy decision making the best that have essentially prioritized, not necessarily building entire systems from scratch versus utilizing this code and these tools to be the glue in between systems with their context and their know how to really uplevel what their entire team and organization is able to do. So I think those are kind of like maybe the two sort of consideration points that people don't always maybe appreciate until they get too far into it. That's a great point. Perhaps just you know, having the, you know, the IT framework or the base on to your point, what they do really well. So if that's you know, distribution or you know, other areas, but then augmenting that with additional integration points or data sources, that's a really good point. So when you think about, it's hard to predict, you know, it's like oh, you know, a couple of years from now, it feels like we should say a couple of months from now. Looking ahead, how do you see AI changing the relationship between consumers and food brands, let's say, over the next next Couple of years? 5 years seems way too far out with the pace of change that we're seeing. So what's coming next, do you think? I think every step that we see with these technology sort of improvements is going to really just empower consumers more. So the way that I like to think of this is even if you just think of the shift from more in store shopping to E commerce shopping, what it really empowered consumers with is put aside the convenience of like, I can sit on my sofa, kitchen, hit a button and it'll be delivered the next day. But really what it's also enabled is a lot of inherent knowledge for consumers as well as a lot of choice. Right. It used to be if I'm picking a laundry detergent, I would go to target. Whatever 20 or 30 options are there is what I'm going to select from. Now if I go on a walmart.com, target.com, amazon.com, I have 200 that I can select from. Right. So the amount of choice that I have has increased greatly. But also I know a lot more about that product before I hit the buy button than I used to. Right. Because I used to have to go in there. Oh, hey, Arm and Hammer is a brand that I've known for years. It's a brand I trust. Okay, that's what I'm going with now. It's like, okay, this is what this person said. Oh, this person's like me because they, I don't know, play or have kids who play a lot of sports. And so stain removal is really, really important. Oh, it turns out that this other brand is being talked about really positively with stain removal. Let me go ahead and pick that up. And so just what I know about the product has increased orders of magnitude before I hit the purchase button. And I think that is just going to continue as this kind of AI technology becomes more and more just available. Right. Like, you can fast forward in a lot of ways, but LLMs will probably know so much about me that it's already going to take that into account before it gives a suggestion on what product to do. So it's already going to have done six steps worth of decision making and processing before it shares information. And there's probably a multitude of other option or possibilities out there as well. It's incredibly exciting times. And you know, I say, I don't know, I feel like I say this every year. It's like it's, you know, never more exciting to be at the cross section of tech and food. And it's like, yeah, every year is just more and more exciting. And because we're seeing more and more adoption, pushing the envelope further and just seeing, you know, what technology can do, providing consumers more information, more transparency, higher amounts of safety and all while helping food companies do what they do best. And I think that that seems like obviously Yogi very much centered on that. So Gautam, I really appreciate the time and the insights. Really fascinating conversation and very excited to see watch what's happening with Yogi. I appreciate it. Thanks again for having me. Thank you for joining us for another bite sized nugget of food tech talk supply chain insights from farm to fork. Make sure to subscribe and if you love the podcast, leave us a review on Apple Podcasts, YouTube, Music, Spotify or wherever you get your podcasts. To learn more about TrustWell and request your free demo of our technology platform that connects food formulation products, nutrition analysis and compliant labeling with traceability, recall readiness and supply chain transparency. Please Visit us at www.trustwell.com and click Get Started or click the link in the Show Notes to learn more. Thanks for listening.