Building a Generative AI Venture from Portugal: João Aroso, CEO of Leadzai (Part 6)

Sramana Mitra: Now, do you also do the A/B testing on landing page optimization and things like that? Because that would be part of your lead generation process, right?

João Aroso: That’s a very clever question and a part of our roadmap because that obviously affects the end result. However, currently we rely only on the customer’s landing page.

Sramana Mitra: Where is the lead collected? Is it collected on the customer’s website?

João Aroso: Yes.

Sramana Mitra: Do you get a notification when the lead is happening? Because you’re charging for a lead.

João Aroso: Yes. We track that the lead happened on the customer website and we charge every time a lead is generated.

Sramana Mitra: Can you elaborate on your business model? You’re buying traffic and then you’re charging for leads. So obviously, there’s a traffic arbitrage going on, which means you are financing the upfront traffic acquisition. How does that work in terms of profitability and financing? Do you need to put in your own money to finance all this? How do you manage the cash flow of all this?

João Aroso: We do need to put in some money to finance this. It’s not a guaranteed arbitrage. We just don’t know the cost in advance because I don’t know how much traffic I need to buy in order to generate a lead.

So, if I was selling clicks only, it would be less risky for selling traffic. We’re doing traffic arbitrage but towards leads, right? So, we do have risk. Currently we lose money in 9.7% percent of our campaigns. That’s part of the risk of our business. We do have campaigns that work exceptionally well and we raise way more money when we make way more money than we anticipated. But we do have campaigns that also go spectacularly wrong and we lose way more money than we anticipated in that campaign.

Sramana Mitra: With the initial 1.5M you started and then brought your company all the way to GPT-3 launch and switched it to GPT-3. Did you raise a round at that point?

João Aroso: Yes. We raised shortly after. So far, we’ve raised 7.5M.

Sramana Mitra: And you’re about to go in to raise more money. How much?

João Aroso: About 2M.

Sramana Mitra: How many clients have you serviced in this mode? How much data has your model handled so far? You obviously have a very clear understanding since you gave me a very clear number on where you lose money and so forth. You have good metrics understanding. Tell me more about what that is based on. How many clients? How many campaigns?

João Aroso: Currently, it’s been over 16,200 campaigns.

Sramana Mitra: That’s across how many customers?

João Aroso: That was the number at the end of March 2024. So, the average campaign per customer is about 1.02. So, it’s pretty much almost as many customers. I think it’s very rare that we have a customer that runs more than one campaign. That speaks a lot about the type of customers that we have.

If you were working shopping, for instance, they would have product-specific campaigns and promotions. However, we’re running everyday businesses like the small accounting firm that does a local service to their community. They do taxes and accounting for a company. They always have pretty much the same campaign. They’re always acquiring the same type of customers. It’s not very dynamic. So, most of our customers run one campaign.

Sramana Mitra: And what about pricing? Is there a kind of ballpark in the price of the leads that you sell?

João Aroso: Actually, I haven’t looked at average pricing because the markets are so different.

Sramana Mitra: Actually, that’s my question. I guess the price of leads depends on the average deal size of what they are selling. So do you normalize to that?

João Aroso: You know, we not only cross industries, but across geographic markets. I remember we stopped looking at the average price when we started working in Latin America.

All of a sudden, everyone was concerned. You know, how is our average price going down so much? What’s happening? It’s because we started working in Latin America.

When you convert the Argentine Pesos to US dollars, a million pesos comes to ten dollars or something. So, we stopped looking at an average aggregate of it all because the numbers in the US are very different than the numbers in Europe. The numbers in Southern Europe are radically different than the numbers in Northern Europe. The numbers in Latin are very different.

So, I don’t know the average price of our leads. But what do I know? That our pricing is built today on a campaign level. So, we do not have a price for plumbers in New York.

When you create your campaign, we use our own technology to estimate the price for that specific campaign with a mix of anticipated cost plus some perception of value of the lead. We price the lead based on how much how much we anticipate the traffic is going to cost, how much we know from our experience that conversion rates would be for that type of customer and for those type of markets. We even take into account macroeconomic variables like purchasing power of that specific target. With the formulas that we have in place for that, we come to a price per lead.

So, each campaign is different. And even the same campaign, if you create it in different moments in time, you might get a different price. Because, I don’t know, cost per traffic went up, went down, or, you know, our combo of data realized something is slightly different.

One of our biggest challenges today is working out our pricing structure. Pricing is so much an integral part of our business model and of our product in itself, right? One could argue that my main USP and my product is pricing, right? So, it’s a difficult but fascinating problem to solve.

I believe that eventually our pricing model will evolve to what others did in the past – to a kind of a bidding world. You would tell me how much you are willing to pay per lead. If I’m able to serve you leads at that cost, I will, if not, I won’t. I think we need a lot of scale to be there. Because ultimately, you know, you can’t have a bidding pricing if there’s not more than one person bidding for something, right? We need a different scale to be able to make that work. But I think that in the end, that’s where our product is going to go.

This segment is part 6 in the series : Building a Generative AI Venture from Portugal: João Aroso, CEO of Leadzai
1 2 3 4 5 6 7