> Here’s a glimpse of what you would learn….

In this episode of the Ecomm Breakthrough Podcast, host Josh Hadley speaks with John Li, co-founder of PickFu, about using consumer feedback to make smarter eCommerce decisions. John explains how PickFu helps sellers validate AI-generated content through real human feedback, particularly for main product images, which drive 75% of click-through rates on Amazon. They discuss PickFu’s newest features, including multi-question surveys, Amazon SERP mockups, and AI-powered insights. John shares compelling case studies demonstrating strong ROI from simple image tests, and outlines an iterative testing playbook to help sellers continuously optimize listings and scale their businesses efficiently.

Here are the 3 action items that Josh identified from this episode:

Timestamps:

00:00:00 Introduction & Importance of Human Validation in AI Era
Discusses the need for real human feedback to validate AI-generated creative assets, especially for e-commerce images.00:00:29 Podcast Introduction & Guest Background
Host introduces the podcast, John Li, and PickFu’s mission to help brands make data-driven decisions.00:02:13 Is PickFu Still Relevant in the Age of AI?
Explores whether AI can replace human feedback and the continued importance of human validation for image testing.

00:04:23 AI vs. Human-Generated Images: Performance Insights
Compares the effectiveness of AI-generated images versus human-created ones, emphasizing context and quality.

00:06:04 PickFu Platform Updates & New Features
Overview of new PickFu features: multi-question surveys, Amazon SERP mockups, image stack, A+ content testing, and screen recording.

00:08:31 Integration with AI Agents & MCP Server
Explains PickFu’s MCP server, enabling users to run and analyze polls via AI chat agents like Claude and ChatGPT.

00:10:02 Main Image Testing & Impact on Click-Through Rate
Discusses the critical role of main image testing, its effect on CTR, and shares a case study of a failed launch turned successful.

00:13:42 Iterative Testing Playbook for CTR Optimization
Outlines a step-by-step playbook for optimizing main images through iterative testing and competitive benchmarking.

00:18:50 How to Structure and Automate Testing with AI
Describes how PickFu’s AI features and MCP server help automate the testing process, including prompt engineering and image generation.

00:20:33 Compounding Gains from Continuous Testing
Highlights the value of frequent, automated testing for incremental improvements that compound over time.

00:22:34 Future of E-commerce Creative Testing & Automation
Discusses the vision for AI-driven, automated creative testing and the evolving role of humans in the process.

00:23:54 Case Study: Simple Test, Big ROI
Shares a case where a single $85 PickFu test led to a 12.5% CTR increase and $3,800 in extra revenue.

00:25:15 Who Should Own the Testing Process?
Explores who in an organization should manage testing—owners, creative directors, or listing managers.

00:27:28 Action Steps for Sellers New to Testing
Advice for sellers: benchmark against competitors, focus on top and mid-tier products, and optimize main images.

00:29:31 Panel Quality & Validity of PickFu Responses
Explains how PickFu ensures high-quality, valid consumer feedback and differentiates from competitors.

00:32:13 Sleeper Use Cases: Product Selection & Packaging
Highlights underutilized PickFu use cases like product selection, pricing validation, and packaging design.

00:34:40 Actionable Takeaways & Recap
Host summarizes three key action items: frequent main image testing, leveraging video tests, and preparing for AI-driven automation.

00:37:10 Rapid-Fire Questions: Book, AI Tool, E-com Influencer
Guest shares his favorite book, AI tool, and a recommended e-commerce influencer to follow.

00:38:35 Closing & Contact Information
Provides information on where to learn more about PickFu and how to connect with John Li.

Tools and Websites
“PickFu“: “00:01:57”
“Claude AI“: “00:09:09”
“ChatGPT“: “00:09:09”
“Nano Banana“: “00:20:33″Books
“Guns, Germs, and Steel“: “00:37:15″Case Studies and Examples
“Jeff’s Case Study”: “00:10:54”
“Yes Bar Case Study”: “00:24:05”

Action Items
“Main Image Optimization”: “00:35:37”
“Run Video Tests”: “00:35:37”
“Implement AI for Testing”: “00:36:33”

Notable Mentions
“Steve Chou“: “00:38:06”

This episode is brought to you by eComm Breakthrough Consulting where I help seven-figure e-commerce owners grow to eight figures.

I started my business in 2015 and grew it to an eight-figure brand in seven years.

I made mistakes along the way that made the path to eight figures longer. At times I doubted whether our business could even survive and become a real brand. I wish I would have had a guide to help me grow faster and avoid the stumbling blocks.

If you’ve hit a plateau and want to know the next steps to take your business to the next level, then email me at josh@ecommbreakthrough.com and in your subject line say “strategy audit” for the chance to win a $10,000 comprehensive business strategy audit at no cost!

Transcript:

John Li 00:00:00  But in a world where there’s ever more and more AI generation, you actually need real human validation even more to validate for which like decide which one to go with. Right. Like, you can, you can easily spin up 50 different image main image variations, but which one are you going to choose? You’re not going to be test 100 or 50 different image types. And if you’re not, someone’s going to have to choose. And if that’s someone is you, then you’re just going with your gut.

MC 00:00:29  Welcome to the E-com Breakthrough Podcast. Are you ready to unlock the full potential and growth in your business? You’ve already crossed seven figures in sales, but the challenge is knowing how to take your business to the next level.

Josh Hadley 00:00:43  You’re not losing market share because of bad ideas. You’re losing market share because you’re testing too slow and competitors are innovating faster than you. But what if you could validate everything instantly with real feedback and move faster? That’s exactly what today’s guest has made possible. Welcome to the Ecom Breakthrough Podcast.

Josh Hadley 00:01:00  I’m your host, Josh Hadley. I scaled my own brand from 0 to 8 figures in sales, and now my mission is to take it to over nine figures on my journey to nine figures. I bring you the unfiltered conversations with the smartest minds in eCommerce. Past guests include Ezra Firestone, Kevin King, and Michael E Gerber, author of the E myth. Joining me today is John Li, the co-founder of PickFu, a powerful consumer feedback platform that helps ecommerce sellers, authors, app developers, and marketers make smarter, data driven decisions. Before launching Pick Food, John spent years at Microsoft as a software software engineer and program manager, but his entrepreneurial curiosity led him to create a simple tool to test ideas with real people. That tool evolved into Pig Foo, now trusted by tens of thousands of Amazon sellers and brands to validate product images, packaging and listings before they go live, saving them from costly mistakes and giving them the confidence to scale. John is a true advocate for taking the guesswork out of ecommerce growth.

Josh Hadley 00:01:57  And today he’s here to share the exact feedback, playbooks and the tests that you can run to drive more browsers into buyers and more ideas into eight figure success stories. With that introduction, welcome to the show, John.

John Li 00:02:10  Thanks so much, Josh. Super, super glad to be here.

Josh Hadley 00:02:13  John, I love what you guys have built with PickFu, and it’s definitely evolved over the last few years. I was telling you earlier, I was like, I it’s rare for me to run into anybody that does not know PickFu, but yet you’re telling me, hey, we’re still scaling, we’re still getting new subscribers and new brands every single day. So it’s impressive what you guys have been able to build. But I think, like the first question I want to ask you, and I think maybe some of our listeners would be curious to hear which is, is PickFu still relevant in the world of AI? Because I know there’s also been some conversations where it’s like, just upload your images to AI, tell it to be a a feedback mechanism, and then have AI give your feedback.

Josh Hadley 00:02:49  And I’m like, I don’t know that the robots know that exactly the same way humans would interpret the images. And so there’s a lot of context that goes behind that. But talk to me more about that, John. What’s AI doing to this image testing arena right now?

John Li 00:03:03  Yeah. No great. Great question. Josh, I think, you know, broadly speaking, from until the time where we can say that AI perfectly replicates every individual’s, you know, behavior and everything else. I think that there is still a space and an important space for human validation and real consumer feedback. In fact, I think from what we’ve seen, you know, we see Amazon sellers and brand builders use AI more and more for that image generation. And that’s been fantastic for everyone because, you know, it’s so much easier to create a really compelling image or some kind of compelling creative. But in a world where there’s ever more and more AI generation, you actually need real human validation even more to validate for which like decide which one to go with.

John Li 00:03:46  Right. Like, you can you can easily spin up 50 different image main image variations, but which one are you going to choose? You’re not going to be test 100 or 50 different image types. And if you’re not. Someone’s going to have to choose. And if that someone is you, then you’re just going with your gut. And I would definitely not trust an AI to randomly choose one out of 52 to use as your main image or as your A+ content. So what we’ve seen on our platform is we see a growing number of AI generated content in terms of creatives, and it’s fantastic to see that because now, because we know that these sellers are actually using real human feedback and real consumer feedback to to validate their design choices.

Josh Hadley 00:04:23  So if more people are using AI images like, what are you seeing, John, from the testing side, are you seeing AI images outperform what once was, you know, human generated images or real lifestyle photos or whatever. Tell me what you’re seeing that way.

John Li 00:04:38  Yeah, the concrete answer is that it actually depends, right? Like you, we’ve seen so much AI slop out there and it’s it really depends on the quality of the image and also the context of it. Because, you know, as I’m sure you know, it’s not just about the quality of the image, but it’s also about what kind of message you’re trying to send with the image. For example, if your product is a multipack. Are you prominently displaying that it is a multipack on the packaging of the image, or are you showing product in use? Obviously depends on the category. You have to think about the buyer and if the buyer is going, if your target buyer is going to be buying this as a gift or buying it for themselves, that affects how you’re going to position the image too. So I don’t think it’s just. Is it AI or is it not AI? But it really depends on the quality and the context of what you’re putting into the image.

Josh Hadley 00:05:21  Yeah, no, I think it makes a massive difference.

Josh Hadley 00:05:24  Understanding like what sits behind it and how much time you spent engineering and creating those images. So that makes a lot of sense now, John, for maybe some of the users who maybe used pick fu in the past, maybe haven’t used it for a while, there are other competitors on the market. I know for me I’m not the one testing main images on my account anymore. So we have team members doing that. So it’s been a couple of years since I have personally been into the platform. What are some of the new developments and things that you guys are incorporating, especially with AI. I’m sure you guys have AI integrated here as well, but tell me more about like, how should people be fully utilizing the platform to ensure that, like at the end of the day, they have the best listing to beat the competition?

John Li 00:06:04  Absolutely. We’ve been kind of on a tear, building up a whole bunch of new features. So for example, it’s not just about single question polls anymore. You can do up to 16 questions, surveys, any kind of any kind of poll type that we support.

John Li 00:06:16  We’ve added a whole bunch of support for things like building a mock up of your listing and testing your mock ups against your competitors, being able to change those images on those mockups, titles, prices, everything. So have a fully simulated sandbox to to hypothetically like stress test these different creative changes we have we have the Amazon Serp mockups where you can actually put that in a simulated Amazon search results page so people can click on it and tell you which product they would buy. We’ve recently added support for image stack for image sets and A+ content, So now you can test the full image stack against like different versions of image stacks or you versus the competitors. You can test full A-plus content laid out, laid out beautifully vertically. Test that against your own versions or the competitors. And we’ve also added support for screen recording. So if you want to send, if you want to send your target audience to landing page, whether that is your Amazon listing, your Amazon storefront or your DTC, you know your DTC home page and have them walk through it.

John Li 00:07:15  It’ll record their screen. You can see a transcript of what they’re talking about, and we’ll even use AI analysis to summarize all that for you. You can do that on pig food now. So those are some of the core components. We’ve also added support for projects so you can group all your polls into projects. We’ve added AI support to summarize insights across all the polls within a given project. So now you can have all your different workflows separated, separated by projects, and get insights across a whole suite of polls on every single poll. We have AI insights and suggested next steps, so we’ll auto suggest the next poll for you based on the based on the results of that poll. And we’ve also launched what we call pick Fu AI, which is a nice little sidebar where you can chat with your results. So that’ll take into account the full context of that poll on the page. And you can dive in deeper into into any analysis or any insights that that poll has brought up.

Josh Hadley 00:08:03  Love it. Sounds like there’s been a lot of updates there, and I love that you guys like, have specific use cases that help, like take the conversation that much further.

Josh Hadley 00:08:11  I especially love the video, one of getting somebody to think out loud because like oftentimes like what they’re thinking out loud when they first see something or an image or a landing page. Like can tell you a lot even better. Like sometimes that video and audio is worth a thousand words compared to the short little little clip that you get from them with quick comments. Right?

John Li 00:08:31  Yeah, absolutely. And I oh, I neglected to mention one thing, but I do think that when this comes out, it should be live by then. Everything that we’ve added, we’re supporting in our new MCP server, which we’re launching soon. So everything I just described, everything that you can do on the Fu website you can do. Through your favorite agent. Through the pic fu mcp server. So now you, you know, like for myself personally, I haven’t launched a poll from the website in a while, but I do launch a lot of polls. I just do that talking to Claude connected to the MCP server.

Josh Hadley 00:08:58  Fantastic. So explain to somebody listening to this, that’s like MCP. What? So what does the MCP thing mean to somebody using Claude or even like a ChatGPT or some of the other LMS?

John Li 00:09:09  Yeah, absolutely. So MCC are basically a way to power up your your chat agent with additional tools and capabilities. So Pic Fu basically took all that functionality that you see on the platform that we’ve been building out for a long time and put it into an MCP server, which you can connect to Claude, you can connect it to ChatGPT, whichever, wherever you’re using it. And then in your normal conversation with your chat agent, you can say, hey, can you go grab this pick fu, this pick fu poll that I ran and help me do analysis and it’ll go through our MCP server, grab exactly your pool, bring everything back into your chat window, and you can have a conversation about it in your Claude ChatGPT wherever. The MCP server also allows you to strategize create polls as well. So now you can now, you know, pick your favorite agent, connect the MCP, and now you’ve got basically an AI researcher that can read polls, analyze polls, strategize, launch polls, all of that stuff.

Josh Hadley 00:10:02  It’s really exciting, really quick. You could have one team member be in charge of, like, all the creative for your brand and just, like, executing, like, just a million times faster than you ever could on your own. And that that’s always, I think, been the constraint, especially if you have like more than a handful of SKUs that you’re trying to manage and improve over and over again. So, John, tell me like, how important is the main image testing and how important is that like click through rate. Do you have any case studies to maybe share with us to say, hey, like the juice really is worth the squeeze? Because oftentimes, like everybody will go and talk like we should go, you know, test your main images. And most people might do that before they launch a product. But what I’m hearing from you is like the best practices should be. You’re testing like every single day, and you’re testing new images or every single week. But it’s like it’s aggressive testing and you’re always just like trying to beat what your last, best performer was.

John Li 00:10:54  Exactly. You’re you’re always trying to beat the the competition. Right. Because Amazon is a perfect almost like a perfectly competitive environment on those search results pages. So obviously better, you know, as I say, better picks get more clicks. I think the stat is that the main image contributes about 75% to the to the click through rate of a given listing. What we’ve seen is we’ve seen great results when sellers test before they launch, like before they start a product launch, and then we see fantastic results, obviously for existing products as well. So I can share two stories. The first one is around turning around a failed launch. We had a guy, his name was Jeff. He was helping a friend launch a brand, and that friend launched a product twice and two failed launches with ad buys and everything. Each time the after the launch, the product was not ranking anywhere near the front page for his target keywords. So the third time Jeff came to pick Fu and basically reverse engineered the competitor’s images, and by reverse engineering, I don’t mean anything like nefarious or bad.

John Li 00:11:54  What he did was really he took the existing organic top three best sellers in that target category, put it on pick Fu, asked people to do a click test like click on the area you like the best and leave an explanation. And by mining all the responses, he was able to figure out what were the aspects of a really good main image in this target category. So then what he did was he went and created created main image variation, tested that against the top three, failed miserably, read the results you know, went back to the drawing board. Iterated again. He did. I think he ran nine different polls, so he took nine iterations of we’re going to make a better main image, tested against the existing competition failed, but iterate and go back to the drawing board until he got to one that was winning consistently against the existing the existing best sellers. and remind you this was all before he relaunched again. Right. So, so sort of sort of off in his little secret lab here.

John Li 00:12:45  He’s doing the testing and he’s iterating. And finally, after, I think 11 or 12 PickFu polls, they’re all small polls, and that’s fine. He ended up with a main image for his friends listing that he was confident could do better. So then what happened? He launched, so he relaunched third time now, I think within one week he was getting first page rankings for all the keywords that that they were looking for. And this was only by changing the main image. Everything else was the same bad buying strategy, all of that other stuff. And so that’s one study where it was it was a clear like rescue launch that worked really well.

Josh Hadley 00:13:17  Yeah, that’s those are incredible results. And John, maybe my follow up question to that is like, so what had to change. And I know this is going to vary depending on like the product category and things like that. But like are you seeing any consistent patterns, you know, across different brands or across different segments of the market where it’s like, hey guys, these are like the tactics.

Josh Hadley 00:13:37  If I’m creating a main image today. Make sure you utilize X, Y, or Z in your main images.

John Li 00:13:42  Yeah, no there’s really good question. So I think what changed is actually the it’s really the testing philosophy, I think I think one mistake that a lot of people have around using pick foo and this kind of testing is that they think it’s one really big test that is going to give you a very clear picture of like red or green or whatnot, when in fact the best results come from a series of tests, like almost like a playbook of testing, iterative testing and improvement. So kind of like that story that I just shared, right? Where you’re running small tests, quick tests, and then taking that feedback and those insights and iterating on it and testing again. And it’s usually in those sequences that we see work really well when it comes to Amazon specifically. I mean, we see we see tests across all these different product categories. It’s really hard. You know, as you know, every single category is different.

John Li 00:14:27  There are certain categories that are the Wild West. Other categories are really locked up. So what wins in one category versus another is going to be different. But what we’ve seen consistently that wins is taking this iterative approach of testing and validating and iterating over and over again.

Josh Hadley 00:14:41  So, John, what does that look like? Like what? What is that playbook like? What are the first things that you should be testing? How do you then iterate to where you get to nine iterations? Because I think most people look at this and be like, is my main image better yes or no? Right. And then it’s it’s and again, obviously with AI, I think you can help maybe like digest some of the feedback now from commenters why they’re picking this or that, but what are the what’s your iteration playbook for sellers.

John Li 00:15:08  Yeah. So we actually created a CTR optimization playbook, like based off everything that we were seeing from sellers who were having a lot of success. And generally the steps are that you gather information first, compare against the competition, iterate and then validate.

John Li 00:15:23  And so those four steps, what I mean by those is in the in the beginning you you gather information about what the buyers are looking for. So you’re trying to understand the buyer intent. What this means in practice is that we recommend running an open ended poll to your target audience. Let’s say you’re selling a dog chew toy. We would recommend that you first run an open ended poll. You’re not even testing any creative, but on pick few, you can run open ended polls just to gather feedback. It’s like a digital focus group, right? And you ask your target, your target audience, dog owners. When you’re buying a chew toy for your dog, what are the most important features that you’re looking for? Really broad. Let them answer, take that information, rank them. Sort of. And now you have a clear list, independent of any kind of creative of what your target audience values in that product that they’re looking for. So that’s step one. You use that for the rest of the series, right? Then step two would be that that comparative like competitive comparison.

John Li 00:16:18  Maybe you have a product or a listing already. You can put that up against the top 3 or 4 listings in that category. Run the poll. It really doesn’t matter. If you lose, you’re probably going to lose because they’re the top three best sellers right now. What you want to know is understand why, right? And that’s where the that’s where the high quality like consumer comments come in. Whether or not your listings in that comparison or not, you take those you you take that those responses and now you understand. Well, maybe this listing has better product in use. This listing has stronger claims on the packaging. This listing has more vibrant colors, but you take all that you synthesize that you combine that with what you learned in that first poll, and now you have a good base of potential hypotheses to improve your image. So now you’re on to that that iteration phase where whether you have an image or not, you can start doing that testing that iterative testing of like, let’s try this, let’s try this, let’s try this.

John Li 00:17:07  Let’s keep testing against ourselves and test different variations until we’re getting, you know, stronger and stronger and stronger. And then when you’re confident at a certain point, you can bring that, you can bring that strongest image that you have, go back again and validate against a competition to see how you stack up against it. There’s going to be situations where obviously the goal is not to win, right? Because you might be coming up against situations where you’re going up against a strong brand. If you’re, you know, if you’re selling cleaner and you’re going up against Clorox. What people are going to click on Clorox because they know it’s Clorox. But what you’re trying to do is just show movement. If in the if in that initial poll, you’re, you know, you’re listing only got maybe 5% of the votes, but you’re able to keep iterating on that main image until you’re getting 10% of the votes. That is I mean, that’s double the clicks. That generally translates to better sales on Amazon. So you are going to see meaningful difference on.

Josh Hadley 00:17:52  That fantastic playbook in the way I would even enhance this is like at the very beginning I would download like the reviews from your competitors. Totally understand, like, hey, this is why these are like this. The feedback that the customer segment is giving about, you know, this market or this product, why they’re buying it, etc. and then you stack that in addition to the poll run, those together have AI kind of synthesize that for you. So my question for you, John, would be this, you know, AI is only as good as like the prompts that you give it, the process that you have behind it. So does pick. You have like really good prompts to where it’s like, here is the reviews. Here’s the output from the open ended pull and does pick. You have like the right question to ask where it does like it pulls everything properly out of AI to then go give that to like your creative director or whoever on your team to go and implement that and begin at least iterating with some new main image options.

John Li 00:18:50  Yeah. Great question. And yes, we’re trying to build that into all our AI features and we’re trying to improve that all the time. So what we have right now, like I mentioned before, on every single poll we have, we have AI insights where we’ll extract the top, I think 3 to 6 insights from the responses that we’re given along with the suggested next poll. And then you can use AI on our AI sidebar to analyze the results as well. I mentioned the pick for MCP. So what we’re actually doing is we’ve embedded into that MCP the playbook that I just talked about. So when you connect it, you should be able to figure out you can ask it for the CTR playbook, and it’ll will give it to your AI agent to download that playbook to know exactly how to apply. You know, first you should be doing this. Then you should be doing this thing. You should be doing this, and this is how you should be analyzing the results. And so I was actually running this last night to take, you know, one shoe dog product.

John Li 00:19:40  And in I use cloud desktop. And I had the MCP connected to it. And in a single session took maybe an hour or so of just prompting, like I started with, I started with one, sort of grabbed a random ugly listing, did this whole process, and then I had this really nice listing listing image. By the end of it that was that was scoring better than the competition by the end of it. And that took about an hour and just going back and forth and probably ran about five holes total. One other thing that we added to the MCP server is actually image generation. So I was able to both synthesize, use the agent to synthesize the responses from all the polls all, and then pass it back to the MCP server, which uses like nano banana and some of the best image generation tools out there to iterate on the images using AI and then test again. So it was all sort of a self-contained loop. All took about an hour or two. And by the end of it I got like a nice new main image.

Josh Hadley 00:20:33  That’s incredible. I think that, like, what you just touched on, just like I had massive unlocks in my mind where the people that are going to be winning in the e-com space over the next 3 to 5 years, and even maybe in the next 12 months, are going to be the people that are like, they’re iterating just like you did. But 24 over seven. Like literally they have one team member that oversees everything, and then they have autonomous agents that light up a pole, change the image.

John Li 00:20:58  Absolutely.

Josh Hadley 00:20:58  And it may be expensive to like, you think like, oh, it’s 100 bucks for this test, $100 for this test. And it’s like, okay, that may be true, however, but in your first example of like, you went from a failed launch to a very successful like you’re now on the main page on the top of search results, like it’s going to pay for itself if done properly. And I love the AI like can actually integrate with itself to like create the main images and riff off of it.

Josh Hadley 00:21:24  It’s like take the feedback and again in the creative world. It’s not even relying upon one person because like, just because one person’s style they like this doesn’t mean that like, oh, well, I have to have my creative director, you know, follow our brand guidelines. It’s like, I’m okay with nano banana just being like, I don’t know. Here’s the feedback we got from the market. So here’s the main image. It’s like might not have been my style, but let’s go test it and we test it. Sure enough it’s better. And then you launch it to the Wild on Amazon. It’s like oh it is better. And so I think that it just removes the human subjectivity out of it, which gets me really excited. And so like this is an area that I think like is a sleeper for most people. Their main image is it’s not it’s not going to be revolutionary. It’s not going to attend my business. But I actually think it can and it can connect your business by like small compounding returns where it’s like, yeah, my main image got my click through rate, went from 1.25 to 1.3, and then it went from 1.3 to 1.35, and it gets better every single week.

Josh Hadley 00:22:27  Yes, that compounds into being able to like ten x your business over time. So yeah. John, anything else you would add to that?

John Li 00:22:34  Yeah. I’m really excited that you were talking about that. I didn’t want to I didn’t want to overwhelm you with like all this AI talk. But honestly, I that is the direction that we’re building because this is the vision that we’re seeing as well in terms of the future of work and how to really unlock, you know, really unlock individuals and workflows. And with automation getting, you know, humans need to be in the loop at the right time, but humans don’t need to be the ones that are always doing that work of running the polls and validating and everything else. So exactly what you said about you can set ups, you can set up schedule prompts and automations and so on to test across your entire catalog. And probably in those cases, you don’t want to run $100, so you can run a $15 poll. You know, we will.

John Li 00:23:12  You know, we’re always looking to make sure that we’re trying to effectively support all these future use cases. So you should be able to you know, that’s kind of where we see work and where we see the growth of e-commerce and the way that creatives and selling works in the future. So we’re definitely building towards that.

Josh Hadley 00:23:29  Yeah, we’re definitely in the infancy of AI, but at the same token, it’s moving fast. And so being able to jump on that train, you’re going to be ahead of the competition. And if you don’t like, you will be quickly behind the competition.

John Li 00:23:41  Absolutely.

Josh Hadley 00:23:42  John, you had another case study that you had mentioned. So I’d love to hear more. Kind of like how people have been using pick foo in very creative ways that maybe some of us listening aren’t really thinking about utilizing pick foo in this way.

John Li 00:23:54  Yeah. So there’s another case study we had where it’s, you know, the first one I was talking about was the iterative was the iterative process. And that was probably, I think 12 polls, maybe about $600 worth of testing.

John Li 00:24:05  Obviously we see some of that. We also see much simpler situations where you have an in-house design team and you’re just trying to figure out the right image. And so we have this one brand called Yes Bar. There’s a couple of years back, I think there are Danielle’s clients, but essentially they worked on a main image, validated it with Pig Foo. I think it was a single test just at the end of their creative process. 85 bucks. So it was a single $85 test. It strongly validated because from what we’ve seen, if you win, if your images win 7030 on pick Foo, you’re going to see higher click through rates on Amazon. So that’s kind of a broad stat that we see consistently. And that’s been validated. So they ran their test 85 bucks. It did win 7030 change it on Amazon. And within two weeks they had made I think click through rates went up 12.5%. And they made $3,800 extra in the next two weeks. So that was a $3,800 fine. 3715 ROI on an $85 test.

John Li 00:25:00  That’s on a single test, right? And I’m not trying to cherry pick these because there’s we see this a lot. We there’s just a lot of sellers who are a little more private about like, they’re a little more private about their sales growth. But we hear from them all the time that it’s very effective.

Josh Hadley 00:25:15  Yeah. It’s just the simple optimizations that make a big impact. Absolutely, John, what do you see like for most sellers is this like the owner is typically the one managing this and running these polls? Or are you often finding that it’s like a creative director or brand managers? What are you seeing that way, and who do you think is the right person to head that up? Especially as we move into the world of AI, where again, I don’t think it removes humans, but it does allow us to do techniques more with the current humans that we have.

John Li 00:25:43  My my tongue in cheek joke initially was going to be that as owners, we would never want to put any responsibility on ourselves to do any work, right.

John Li 00:25:51  But honestly, I think that I think in the world of AI and automation, I think I think it’s whoever is responsible for making sure that listings are performing well. So that depends on the org structure. And if you’re a single seller, then it might be you, right? You need to have the right combination of both context and vision for what you want to see on your brand, but also the time and the availability to run these polls and to do this kind of testing. So typically what we see that works best is whoever’s in the organization, typically it’s a creative director or the listing manager who or the Amazon manager, depending on the size of the org, who who is are running these kinds of playbooks on their list, on their catalog, on a regular basis. That’s where because a lot of times we find that the individual seller who’s owning the business, they they may be bought in, but they’re just so busy with everything else, right? They’re busy, you know, they’re busy with inventory.

John Li 00:26:38  They’re busy creating your product, all of this other stuff. So whoever it is in the in the organization whose job it is to to ensure that the listings are there and that they get good, that they’re performing well and that they’re managed properly, I would put I would put testing on their plate.

Josh Hadley 00:26:51  Awesome. Now, John, if I’m listening to this episode and I’m like, man, this, this sounds awesome, I definitely it’s reinvigorated me to, you know, get into my main image testing and do more of it. If I drank that Kool-Aid and I come to you and I say, all right, John, I’m sold, what should I do now? Tell me. I’m going to commit resources, even my own time to dive into this. I want, I want 100 x my business through main images and testing and validation and things like that. What would you encourage sellers and those listening to do? First and foremost, and maybe again, some other use cases to just like really double down on what’s performing right now.

John Li 00:27:28  Yeah. So I would do I would suggest doing a lot of competitive testing to benchmark your catalog against what’s out there. I would take the top performers in your catalog and just validate that what you think are the strengths are actually the strengths against your top competition. So the top one, two, three sellers in your those target categories I would take your top, I would take your top best sellers and run benchmarking tests against those just to validate that, you know, you’re not leaving potentially even more money on the table. Right. Because there might be other aspects of the image that you could improve compared to the organic best sellers that could result in an even bigger lift for your best sellers. Then I would take the middle of your catalog, the ones that are like doing okay but not, you know, not selling nothing, but not not crushing it. And those I think are the real opportunities. Those are the ones where I think from, you know, on like the page 2 or 3 of, of the of the listings of the keywords that you’re looking for, because those are the kind of the ones within striking distance.

John Li 00:28:24  And those are where I would focus a lot of the attention on, because what you want to get is you don’t want just a catalog where you have like a tiny handful of bestsellers, but you want more bestsellers, like for differentiation and in your catalog. So I would focus on those striking distance listings in the middle and spend some time doing benchmarking and also looking for and really focus on improving the main images of those to see how, you know, can you get can you get something from a page three to a page two, from page two to a page one? And that’s where you’ll see a lot of opportunity for growth.

Josh Hadley 00:28:51  Fantastic. I like that playbook. So John, another selfish question that I have for myself. And you don’t have to call out any specific competitor names, but I am curious with like the lay of the like the testing market there, there’s been some rumors where it’s like, hey, it’s just a bunch of Filipinos taking tests, or it’s not even like legit consumers. So tell me, like how you guys have it set up, like who is actually taking these tests? And is the data actually, like, valid? And then how does that compare with like, is it true like maybe some of the other tools that are out there on the market, maybe like you’re not getting what you thought you would pay for? Because I think like if we’re talking about this level of testing, it can add up really quick.

Josh Hadley 00:29:31  And so that’s why it’s so important to actually get valid responses back to ensure that, like you’re moving in the right direction. Sure.

John Li 00:29:39  Oh good question. So I don’t know a lot of I don’t spend a lot of time looking at the rest of the testing landscape. I know what we do. What I can share is that we take quality. We take quality and validity. It’s like super important at pick Foo for us. We are building. We build our panels on top of existing enterprise consumer panels that are out there. These are the same panels that the giant CPGs like Procter and Gamble and Mondelez use for their own research, except when you’re in a large enterprise like Procter and Gamble, you have a centralized insights team working with third party consultants who tap into these panels, who where the studies are, you know, five figures and five months. And like, they take forever to go back and forth. And a lot of times the consultants are just figuring out who to target, how to target them.

John Li 00:30:30  And then on the other side, spending a lot of their time filtering out for bad data, bad responses. You know, quality control for the for the responses coming back. So we see Pig Fu as a way to shortcut that entire enterprise insights process using software and our own tools. So we spend a lot of we spend a lot of time on the quality and targeting side right now. Right now, our panel is about 20 million reachable consumers across 15 different countries. And we see a lot of great use cases where you have sellers in one country wanting to target, let’s say, potential buyers in Germany or in Japan, you know, or in Korea or in Brazil. Like there’s a lot of really interesting sort of cross-border internationalization use cases On the quality side of things. We have a ton of different layers around making sure that that respondents are paying attention. They are not writing gibberish, that they are who they say that they are. And I will just volunteer that we reject over 50% of the people who try to answer our surveys.

John Li 00:31:26  So quality is paramount for us. If anyone finds that a response is inaccurate or not, paying attention or whatever will happily replace that. You know, like we we would love, you know, we believe that we’ve built a reputation for high quality, high quality consumer feedback that you can rely on to make these decisions for your business. And we’ll stand by that.

Josh Hadley 00:31:46  I love what you guys have built there, especially on top of like, the existing like large consumer panels that the top brands, the big enterprises are using. So I think that’s very unique to be able to tap into that. Now, John, I’m curious, as we wrap things up here, are there any other use cases that we haven’t talked about that you feel like, hey, the listeners, like we talked a lot about main images and click through rates, but like what are some of like the sleeper tests that like, people may not be doing that actually are very impactful.

John Li 00:32:13  So many. We could talk for hours about this.

John Li 00:32:15  I think I think people people don’t use pick foo early enough in the process around product selection and product design. You should think about pick foo as a way to tap into your target market and understand what they want. The fact that Amazon sellers use it for main images, that’s great, but you’ve already decided that you’re selling this product and you’re just trying to optimize that. But when it comes to product selection, like pick Foo is amazing. To help you validate whether or not you should even be selling this product. So I’ve heard great case studies where someone came and said, okay, well, I ran a pricing. I wanted to sell this really fancy toilet paper holder, and I set up my colleagues and I knew exactly how much margin I needed to make off each one. And I ran a pricing test on pick foo. I put a couple of different prices and asked, you know, and ran a bunch of polls around that, and it validated that I could not sell this thing for the price that I needed to sell it for in order to make a profit.

John Li 00:33:06  So in that case, the pig foo test actually validated that he should not sell this product. And he came back and he’s like, well, you just saved me $10,000. You know, you just saved me an arm and a leg of like, you know, failing on a product and like, pounding my head against the wall, trying to sell this thing where I couldn’t make a profit for it. Right. So the earlier you test, we like to say test before you invest, right? Like the earlier you test, the the more impactful its going to be. So product selection is one of them. And also product packaging I think is a huge thing particularly if you’re selling into retail. You know everyone wants to diversify like Amazon’s great but so are many other marketplaces. Being in retail is fantastic. So we have a lot of brands who are testing, who are testing their packaging, making sure that the claims are most appropriately prominent on packaging, even even setting up mock ups where they’re they have an opportunity to be on an end cap in a retail store, and they’re testing out the end cap, display design and everything else.

John Li 00:33:55  And so those are really neat to see as well.

Josh Hadley 00:33:57  Yeah, I’ve got my head spinning of just different ideas and things that we could test and how we could do so much better behind that. But I really like the insight of like, hey, before you even consider going into that product category, like vet the product and you could probably innovate the product even faster. People talk about wanting to have like a utility pan but then say, oh, but I’m not very creative. I don’t have many good ideas. Guess what? Leverage everybody else. That’s going to give you an idea and feedback about your product. And you’re like, actually, that’s a really good idea.

John Li 00:34:25  We actually see a lot of those poles of lists of features listed, like if you were going to buy blah, what kinds of, you know, which of these features would you rank as, like the most useful and most valuable? Like we see a lot of those. So those are great.

Josh Hadley 00:34:37  Fantastic. John, this has been a fantastic episode.

Josh Hadley 00:34:40  I’ve got three action items that I’ve listed here that I think the sellers and those listening can go implement in their business, but you let me know if I’m missing some here. All right. Number one, main image optimization needs to go from maybe a once a year or once every six month testing cycle to a every other week I am testing something new and it is being pushed live to Amazon. And then the nice part is, the best feedback that you get is Amazon’s PPC data. It’s going to give you really quick. Like why does that click through rate. Yeah today. So that’s number one is like test your main image. Utilize polling software testing it against your competitors. And with the advent of AI, you should be able to execute this on a weekly basis without having to have the human doing every single step along the way. So that’s action item number one. Action item number two is we didn’t dive too much into this, but I’m going to say run video tests because video tests are going to be like worth.

Josh Hadley 00:35:37  Those videos are worth a thousand words compared to just maybe a single test that you do on the image itself, because you’re going to audibly hear what people have to say. And if you’re in the DTC space and you’re saying, hey, why am I struggling on Shopify? Why is this landing page not converting off of my meta traffic, hitting people and adding that on there with like, they might even be completely confused. Like I was expecting to see x, y, and z. I have no idea where to find it. I don’t even know how to add this to cart or whatever, and it might be something very obvious to you, but to the external user you have no idea. But hearing that audio goes a long way. And then my third and final action item here is begin prepping for implementing AI. With this, I think the power of Pickfu and the brands really crushing it in the future is not just like, hey, we use it to test and validate products. Like that’s the playbook of, you know, the early 2020s, so to speak.

Josh Hadley 00:36:33  But like if you want to move into the next decade of e-commerce, it’s how do I have agents that are executing this playbook for me? And now we do ten x, the number of tests that we previously did. And guess what? Naturally, your business is going to be that much better, stronger, faster and bigger because of all of these tests, because these are micro improvements that compound just like compound interest. These are compound improvements to your main image, to your click through rates and to your conversion rate. John, is there anything you feel like I missed here or something else we need to expound upon?

John Li 00:37:07  No, I think those are fantastic takeaways, John.

Josh Hadley 00:37:10  Love it. The final three questions I have for you here. Number one, what’s been the most influential book that you’ve read and why?

John Li 00:37:15  I wish I could say I read more, but I would say that honestly, it was. I think it was guns, germs and steel. Actually.

Josh Hadley 00:37:21  Interesting. I have not heard that one before.

John Li 00:37:23  It’s about the way that human societies evolve based on geography, context, even weather patterns, latitude, longitude, all of that stuff. I find it super. I found it super fascinating to think about how history evolves and how things change completely based off the context of, you know, the situation that they’re in. And so that has changed the way I see things.

Josh Hadley 00:37:44  Love it. Second question what’s your favorite AI tool that you’ve been using and how have you been using it?

John Li 00:37:49  I got to give a shout out to Claude Co work on the desktop paired with the Pig Fu MCP server. It just unlocks a whole bunch of really interesting use cases, not just for testing, but also for day to day work.

Josh Hadley 00:37:59  Two final question for you here. What who is somebody that you admire or respect the most in the e-commerce space that other people should be following and why?

John Li 00:38:06  Really good question. I’d give a shout out to Steve Chu. He’s been around for a while, but he’s really smart. Guy gives very open opinions about the state of e-commerce and the different tools out there.

Josh Hadley 00:38:17  And he’s got he’s got a lot of context and he’s got a podcast as well. It’s one that I actually started listening to when I first.

John Li 00:38:23  It’s amazing. Yeah, he’s great, great guy.

Josh Hadley 00:38:26  So good one to follow. Well, John, this has been a fantastic episode. If people want to learn more about you, learn more about Pick Fu. where’s the best place for people to go?

John Li 00:38:35  Yeah. So you can just go to the website, pick food, pick fool.com. I’m on LinkedIn. You can reach out to me there.

Josh Hadley 00:38:40  Awesome. John, thanks again for your time today.

John Li 00:38:42  Cool. Thanks, Josh. It’s been great.

MC 00:38:44  Thank you for listening. Visit Ecom breakthrough com for more information. If you’ve enjoyed today’s episode, the best way you can show your appreciation is by clicking the subscribe button and quickly leaving a review. See you again next time!