Dave Killeen: [00:00:00] Hi there, Dave Killeen, Field CPO of Pendo, and a very warm welcome back to episode seven of The Vibe PM where every two weeks or so I'll be showing you different ways of dancing with AI so you can have more joy and fun and impact with your work. I'm recording from a different location today because it's absolutely chucking it down outside torrential rain.

Anyway, so I wanna show you today something beyond Lovable. Lovable is fantastic, but I wanna show you how to take it up. Another level when it comes to vibe coding apps. I was recently at a conference and I was going down to Auckland to present to about 300 product managers, and on the plane with two hours and Elon Musk, Starlink, I vibe coded in that, that I'm gonna show you on this episode, which is here in the arms.

It's absolutely nuts. Basically it's a KPI driver tree that helps teams figure out what the north star dollar metric is and how the work that they do [00:01:00] connects to the very top. Atlassian recently came out with a stated of product management report, and they said that only 12% of product managers feel that the work they do connects to dollar, board level outcomes. And so here today I wanna show you the vibe coded app to spark your imagination. If you're interested, let me know. I can get you access to it once I productionize it and throw it up online. But I wanna show you how easy it is just in two hours to build something as compelling as it is.

And so what's wrong with something like Lovable? Actually, I think something like Lovable or AI Studio with Google. Has a place to get the foundations of a product off the ground. You have that chat, all the scaffolding comes together, and it's fantastic. But then once you have that foundation in place, you need to iterate quickly and just having a sequential chat, you throw something in, you wait for it to be built. You throw something else in. Once you have good scaffolding in place, once you've got the foundations in place for your app, you have so many ideas that are coming at you. What you need is you need parallel [00:02:00] agents. You need to be able to go into something like Cursor or any other IDE that lets you do this, and just fire up lots of different agents simultaneously in parallel working on your many, many ideas.

And so you go in and you'll just go into one agent. I want that feature into another agent. I want that feature into another. And you have 20 agents in the space of 10 minutes, all doing work at the same time. So the pace of progress as you're building out your app is fundamentally different than just having a chat.

Wait for your update, having a chat. It's just too slow. So I'm gonna show you today how to move into that new world. And so Lovable obviously needs some introduction. You go in, you have a back and forth chat, it makes the edits incrementally, and you just get something really good. And AI Studio, I really feel is going to mop the floor next year.

They've got a huge release coming out in January, and in it you can now build AI apps. And with AI Studio you have the full [00:03:00] breadth of the Google portfolio as such. So casting your eye down here, you have all of these different ways of bringing Gemini intelligence into your app, having prompt based video generation, analyzing images, understanding video transcribing, audio generating speech. I mean, there's so much in here bringing voice into it, Nano Banana, the whole shebang, the full Monte. And it's great so you can get the good guts of an app off the ground, up and running with your ideas. And then whether you're in Lovable, or whether you're in Google, you then just go over here to GitHub and then push it up to the private GitHub account, or similarly do that from studio over here in the top right hand corner and then once you push that up, you then pull that down into Cursor. Once it's in Cursor or any other IDE of choice, then you are off to the races.

Let me show you next.

All righty. So we're now in Cursor IDE. And like I say, you can use this with any IDE that you're gonna use, whether it's Windsor for [00:04:00] gravity, and in here clone the repo. You're gonna clone the repo from GitHub and you're gonna put the URL of your repo that you have in your GitHub account.

Once you do that, it will then suck everything in. From there, you can start playing with it. And by the way, if you just wanna cut to the chase and start playing with something that's already been built, go to GitHub and have a quick look at what's out there and just fork something. I'll quickly show you how to do that now.

So here we're on GitHub trending and I'm looking at the monthly view. I'll actually find this a really interesting, just to pop in and see where that puck is going. What are other developers in the community getting really excited about? And you can switch this from today, this week, and this month. And so what I've done is I actually found something here that if you wanted to copy this, this essentially is a modern version of Salesforce.

Look at it, gorgeous. And so what you can do if you want to and not have to kind of get your foundations done and the likes of Lovable or AI Studio for the purpose of what we're showing you here in this podcast is you just go up here over the code, take this, copy that, [00:05:00] and then put that into what I just showed you a second ago into your IDE.

Bring this whole code down locally onto your machine, and then we're gonna do the next thing. Okay? And just like a cook, we program, here's what I made earlier. So I'm now in Cursor. Imagine I had them connected that repo that we just saw, that CRM. I haven't done it in this case. What I've actually got now is a code, which is the KPI driver tree, which I'm about to show you in a second after this short introduction here.

This is Cursor and this is just the agent code. You don't need to see any of the files. You can see the files if you want to, but quite frankly, I do not want to, I do not code. I haven't got a clue how to code, so we just get them all outta the way. Now, this is what I wanna show you over here on the left hand side.

Basically what I've done here, this is my history, and literally every single agent I fire up, you go in here and you say, "Hey, can you build me X, Y, Z feature? Please thank you. Love you." Always say, love you. And then, right, and then boom. And you get that off the round. All that goes. Next one. "Hey, can you move X, Y, Z across the page, please?

Thank you. Love you." And they're [00:06:00] all firing off for you. They're all just going boom, boom, boom, boom, boom, and you can see them all running. I mean, obviously clue what I'm asking you to do. But the point is, when you're doing this, when you're working in this way, you could just fire off all of these ideas and they don't trip over each other.

It's really effective. Very, very powerful. And like I said, compare that sequential moving forward in a chat back and forth. Whether it's AI Studio, whether it's notable right now, you can do. Parallel chats in there. You will be, I'm sure, soon. But for now, this is by far the fastest way to move forward at lightning pace when it comes to vibe coding.

And so now let's hit the scene, the pièce de résistance. I am going to this conference. I want to tell people the importance of KPI laboratories and how Pendo can support driving your metrics in your product that's confidently connect to the north star metrics and therefore you are creating value for your business.

Remember that shocking stat from Atlassian's State of Product Management report. If you haven't read it, do read it. Somewhat depressing, but it says that 12% of people in product feel that they're connected. The teams are connected to north star dollar [00:07:00] outcomes, conversations, concerns that their board would be interested in, and therefore, the majority of us, 88%, basically, we were just shipping things, no idea if it's having an impact on the dollar. So anyway, I created this five code app in two hours with Elon Musk starlink on the plane, which is amazing. Completely transforms a 16 hour journey, and then I showed it to the audience and the reaction has been phenomenal and from other leaders I've been talking to since, to see how much this actually helps them understand how they see their business and how their teams are connecting to that north star outcome. I'm now gonna show you the recording from that presentation, and there's a hilarious bit in there, which actually has the AI look at the AI and tell the AI off and say to the AI, your north star metric is not very good. Here's another instrument really, really effective. You're having the AI red team itself, which in principle is incredibly powerful for what average you're doing generally with AI. I'm gonna press pay now. Here we go. Okay, so we're gonna go with Trade Me and see if we can generate a decent KPI strategy for them.

Here we go. So now it's [00:08:00] doing. It's off to the races, is now pulling all the information it can find on the internet about this particular company. Coming up with an inferred north star metric, building up the driver tree, and then pulling together this detailed knowledge that I'm about to show you. I'm gonna quickly fast forward in the interest of time so we can all have a drink a lot sooner.

And when you look at this, so gross merchandise value, typically what you'd expect to see for an e-com marketplace like this. That's all very good and expected. And then we mentioned those three vectors earlier, increasing revenue, reducing cost, and reducing risk. What is now done is pulled together what initiatives it can find underway within Trademe on each of those vectors, along with quotes in the C-suite and the CEO down here as well to give us a bit more context and color.

Now, here's where it gets interesting. Now we see under GMV, we have then down here the KPIs that then next ladder down from there. So increasing the number of successful transactions seems common sense. Increasing customer lifetime value seems common sense. And so you've got those two metrics here. Now, if you wanted to add more in, we could add more in if you wanted to.

You see then metrics that sit beneath. Then increasing number of successful transactions, and then in product metrics itself in terms of what you can control as a product [00:09:00] org to drive up those buyer conversion rates above. Right. So that's all good. All well and good. This is fantastic. So now what you have down here, the red team analysis, let's the AI act as devil's advocate to stress, test the strategy, identify potential blind spots, and find opportunities for improvement.

This bit's bonkers, folks. Wait till you see it. Basically you're having the AI tell the AI off. It's looking at a starter for 10 on the KPI driver tree and basically is saying what we've designed here is basically not very good, honestly, even though we probably have 60% here. Having the AI now to do a second pass and basically critique itself is so powerful.

Let me click on that. I'll fast forward and show you the results. Okay, and here we have it. So let me read this out to you. Alright, let's cut to the chase. I've reviewed your KPI structure and it's a decent first draft, but it has become that it has some structural flaws that will cause problems down the line.

My job is to find the cracks before they become chasms. Here's my analysis. So gross merchandise value is logical, but one dimensional. GMV is a product [00:10:00] of two key variables, the number of transactions and the average transaction value. But the entire KPI tree is built around the first variable, with a complete and total absence for the second transaction value.

You are set up to celebrate $1,000,001 sales just as much as $10,100 sales, even though the latter is likely far more profitable and indicative of a healthy marketplace for high value goes like cars and property. Your stated high margin verticals. Basically. Then it comes down here with a recommendation.

Add a second primary driver under GMV called increase average transaction value. Under this, build out sub drivers and user actions focused on encouraging high quality listings, promoting bundles and optimizing user experience in your high margin classifieds like motors and property. Honest to God folks, this is just bonkers.

Now in the next bullet. You're incentivizing a landfill. This is so fun, not a curated marketplace. Of course I am. I'm such an idiot. So the driver, increased listing volume and quality is a [00:11:00] classic example of mismatched goals and metrics. Dave, while you've named quality every single underlying user action.

It measures either quality of listings or monetization of sellers. There is nothing here that actually measures the quality of a listing before it sells, blah, blah, blah. Here's a recommendation. Pull your socks up, Dave. You're stupid. It. I could go on. But anyway, so I'm assuming you want to take all of this.

I apply the feedback and then you get the idea. The KPI driver tree updates itself based on the AI telling the AI that it's shit and honestly nuts. But very effective. And so then I've also got this thing in here to help teach you about the importance of certain metrics if we do a deep dive on this metric.

It will then have the AI come up and explain why this metric is important, what benchmarks look like within the industry and ultimately what ideas it has for you to then drive that metric further forward. So again, it becomes this kind of educational tool. The purpose of this is to really show you how you can quickly design a sparring partner for you that fits around the work and that you do as a geek, right?

And [00:12:00] this is not difficult. This literally took a few hours on the plane. I mean, clearly I went to town on this because I've now created a back and forth podcast with two people. I've got no idea who they are. Having a banter about this whole strategy is fantastic. So easy. An AI assistant with voice.

Honestly, if I get lonely, have a back and forth. Add more context in here. And then down here, what you can do as well as you say up here is strategic alerts, it looks for alerts on the competition. Um, and you can then, yeah, do whatever you wanna do here as well. So. Bonkers, bonkers, bonkers. And then this will give you the strategic play work for you to basically articulate exactly what's happening back to your org and how you are going to use this tool to drive the KPI driver tree and ultimately dollars for your business and your shareholders.

Absolute crazy nuts. So I'm gonna wrap this up right quickly because all the last episodes I've been giving you so far, I've always promised to you I'll be under 10 minutes per episode. This one's gone on a little longer. But an awful lot in your head that I hope you've really enjoyed and you'll find useful.

Many thanks for joining me on episode seven of Vibe PM. Please spread the vibes and I'll [00:13:00] see you next time. Take care. Bye.

Registrations for Pendomonium 2026 are now open and you really don't wanna miss this one, folks. We're bringing together the most inspiring minds in product and leadership who will challenge your thinking on everything from product-led growth to the future of tech to actually gaining value from your AI agent investments.

This product festival is designed to spark curiosity, create conversation, and build community. While spotlighting the newest tech for software experience leaders, I would like to personally offer you exclusive 30% discount when you use the code. The Vibe PM 30. That's the Vibe. PM or lowercase one word.

And the numbers three zero. Get your discounted ticket at pendo.io/pendomonium. See you there.