,
Podcast
OCT 30, 2024

It’s the Strategy, stupid – Andrew Martinez-Fonts (VP Product, Honeysales)

Share on

This week on the podcast, we speak with Andrew Martinez-Fonts, VP of Product at Honeysales, to discuss product strategy and how that should always come before any technological advances. Andrew stresses the need to truly understand customer needs and align strategies with business goals.

Featured links: Follow Andrew on LinkedIn | Honeysales | 'Five things we learned at the #mtpcon + Pendomonium roadshow - Berlin 2024' feature by Louron Pratt

Key takeaways

Chapters

00:00 Introduction to Product Strategy and AI

02:54 The Importance of Product Strategy

05:46 Navigating AI in Product Development

09:02 Budgeting for AI and Technology

11:51 Data Quality and Processing Challenges

15:09 Cost Management and Economic Considerations

18:02 Outsourcing vs. In-House Development

Episode transcript

Randy Silver: 0:00
Hey everyone. Randy, here I'm flying solo today Well, not solo. As always, we have a great conversation coming up in just a moment and yes, we did talk a bit about AI today and yes, every conversation lately seems to touch on that topic. But what I really enjoyed about chatting to Andrew Martinez-Fonts is that he's really focusing on the strategy first and we get into how he's doing the actual work at Honeysales. Okay, there's no dad joke today. Lily's not here, so let's get right into it.

Lily Smith: 0:37
The Product Experience Podcast is brought to you by Mind, the Product part of the Pendo family. Every week we talk to inspiring product people from around the globe.

Randy Silver: 0:47
Visit mindtheproductcom to catch up on past episodes and discover free resources to help you with your product practice. Learn about Mind, the Product's conferences and their great training opportunities.

Lily Smith: 0:59
Create a free account to get product inspiration delivered weekly to your inbox. Mind, the Product supports over 200 product type meetups from New York to Barcelona. There's probably one near you.

Randy Silver: 1:17
Andrew, thank you so much for joining us live in person in Berlin at the MTP at Pendo Roadshow. Yeah, I'm very glad to be here. Thanks for having me, and you spoke earlier today. It was a really nice talk and everyone here knows who you are and what you do, but people listening and watching probably possibly don't yet. So can you give us a quick introduction? What are you doing these days and how'd you get into product in the first place?

Andrew Martinez-Fonts: 1:40
Sure? Great question. I'm sure most people don't know me. I'm not well known yet. Right now I'm the VP of product at Honeysales, which is a seed stage startup here in Berlin. I joined about nine months ago. I came from a company called Contentful it's also a Berlin-based startup now more of a scale-up, I would say and I got into product first. I came up through design, so I studied industrial design in college, with designing phones and cars and toys and things like that, graduated in 1998, way back in the previous millennium when the internet was starting to happen and I had learned a little bit of web development skills and stuff. So I went into web design for about five years and then made the leap into product management through a very good friend of mine named Chris, who had joined eBay as a product manager.

Andrew Martinez-Fonts: 2:27
And he and I had somewhat similar backgrounds and he said hey, you might like it over here, you might enjoy this job and found an amazing community of product managers not only wonderful people, but people I could really learn from. It was a great experience.

Andrew Martinez-Fonts: 2:40
So since then it's been a combination of, really, I guess, only tech companies, but large and small, of lots of different varieties and business models and industries and so forth, and it's been fun. I would describe myself as kind of like a journeyman PM. I don't have a very particular focus in fintech, but I've done fintech. I don't have a focus in advertising, but I've done advertising. No problem is uninteresting enough for me to get it to.

Randy Silver: 3:07
Okay, I'm not going to ask you about the boring stuff today, Andrew. So with Honey Sales and the talk you gave earlier was talking about how you're using AI there and you don't start with the tool when do you start?

Andrew Martinez-Fonts: 3:21
Yeah, my talk is about AI and product strategy, but really where I think we should all be starting is product strategy. Ai is a super fascinating and interesting tool and I've learned a lot personally and we've made a lot of hay from it at HoneySales. But really I think one of the challenges with AI in particular, and with any new technology, is kind of hype and people sort of get this idea that the technology will solve all of our problems. It's a panacea that we don't have to do any hard work anymore, we don't need to understand what we're trying to do anymore, we don't have to make any decisions, we just AI will sort of solve all our problems. And my point from the presentation was really like you still need a product strategy. You still need to figure out who your customers are, what their problems are, what they're actually going to pay for you to solve and focus on those things, and AI might be one of the good answers to your problems. It's turned out to be one for us of honey sales, but it may not always be.

Randy Silver: 4:15
And you referenced a really nice model from an old guest of ours from Gibson Biddle.

Andrew Martinez-Fonts: 4:19
Tell us a little bit about the approach that you like, yeah, so I started reading Gibson Pettel's work, I want to say maybe five or six years ago, when I was asked to write my first product strategy for my team, and I just like the way he simplified things down a lot, and so the model I talked about today was the DHM model that he describes. It's really not about how to write product strategy, but it's about evaluating hypotheses that could contribute to your product strategy. And DHM stands for Delighting Customers in Hard-to-Copy, margin-enhancing Ways. So delighting customers is making sure your customers really love the product you're giving them to solve the problem. Hard-to-copy is about kind of creating a competitive moat around yourself so that other people can't enter your space easily, and margin-enhancing is about each of your investments generating increased margins for your company, ie profits that you can reinvest into your product development sort of equation, if you will so that you can keep innovating and keep enhancing the product itself, continue to solve customer problems.

Randy Silver: 5:19
So when you were first asked to write this product strategy, what level were you at? Were you a head of products?

Andrew Martinez-Fonts: 5:25
I was in a group product manager role, so I was managing a team of individual contributors and our head of products, or our VP of product, had asked each of the GPMs to write a strategy for our groups, and it was the first time I had been asked deliberately to do so, and I was terrified at the prospect. I mean, the idea was gosh. I've never done this before. I had had an experience in the past where somebody tried to do some business strategy with the product team. It didn't work out that well and so, yeah, I was daunted, I would say, and I think some of what I liked about his writing was just that it was about trying to kind of simplify things down to the most core elements, and I thought it was also something that the audience here would appreciate in terms of describing what makes a good hypothesis.

Andrew Martinez-Fonts: 6:11
So how did it go? The first time you wrote one, I think it went okay. I actually used a different model. I used the Glee model. If there are other Gibson Biddle fans out there, that one is, I think, get big on X, lead on Y and expand, or something like that, into Z, and it's actually, I think, fairly close to the McKinsey three horizons model, which is kind of like what's the current problem we need to address, what's the next one and then what's the one sort of three horizons out? That's like the new, completely new business or something like that. I think it went OK. I don't, you know, it's hard to say like it was well received, but you never know if it's successful until you know the proof of the pudding is in the tasting, as they say. So yeah, it's hard, I don't know.

Randy Silver: 6:57
So in the years since you wrote that first strategy, you've written a number of them since then.

Andrew Martinez-Fonts: 7:04
How has your approach changed? Good question. I think the first time I wrote something it was probably six, seven, eight pages long, probably a little too much, and I think keeping things simple can be really helpful. The motto in my high school was teach us to light and simple things, and I realized as I've gotten older how much wisdom there is in that. And so keeping things short, trying to make them punchy, I think another way of expressing product strategy that I like is identifying your unfair advantage as a company or as a product, and so that sort of also worked its way into my work over time.

Randy Silver: 7:36
But anyway, so when you again, when you first did this, you were a group product manager, which to lots of people who were working underneath you you seemed like a position of authority. But being in that level, you're kind of in the middle of the hierarchy and you saw other people on top of you. Now that you are the leader in a company, how has that changed your approach?

Andrew Martinez-Fonts: 7:57
I don't know that things have changed necessarily. I am at a higher title, if you will, but at a much smaller company, and so the dynamic is not necessarily different. I can imagine it's as if, and as we scale up, I can imagine it would be a lot harder to put together a cohesive strategy, but at the same time, you know, strategy should be straightforward and simple, and so maybe not, I don't know.

Randy Silver: 8:20
And as you're working with companies of more medium size, not the smaller size but as you scale up a bit, are you asking people who report into you to write strategies for different areas?

Andrew Martinez-Fonts: 8:31
now Not yet. No, we're still too small, but I actually think it's a really useful exercise and, at least for me as a mid-level manager, it was really a great challenge and something that expanded my skill set at some level, even if I wasn't super successful. You know, practice makes perfect, and so it was good in that respect.

Randy Silver: 8:50
And as a company that is both selling B2B and as someone who's looking at the way that companies are budgeting these days. What effect is AI having on people's budgeting in the product and dev space?

Andrew Martinez-Fonts: 9:09
This is not my original work. I read a newsletter called Clouded Judgment by a VC named Jamin Ball. He's based in Palo Alto, but I really like his work because he takes a super analytical approach to the world of enterprise software and I come from design, so my analytical skills are not quite at that level. But he recently wrote about this concept that he called bifurcating budgets, that basically, big companies have a tech budget, or in the past they had a single tech budget, and those budgets are bifurcating into traditional technology and AI-fueled technology, with the promise that AI will deliver outsized returns on their investment and essentially that if the tech budget was growing by a certain amount year over year, that in this bifurcated world, the traditional tech was not going to grow at all and the AI bucket was going to grow dramatically. And so I just thought it was a really interesting perspective on, or evidence for, what he was hearing in the market. He also talked about how big companies are buying tech on three to five year cycles and with AI, we don't know what's coming next month, let alone next year or next three to five years, and so there's a lot of uncertainty in the market around like is it worth committing to this AI solution right now, or should we wait six months and see if something better comes along? And I'm sure that's a huge dilemma for, especially for big buyers who are buying big you know, making big purchases of tech. I imagine that the industry will respond by maybe changing commercial terms shorter commitments or something like that and I'm sure that buyers are also buying based on company reputation and stability and funding and things like that.

Andrew Martinez-Fonts: 10:37
I certainly know from my time at Contentful that that was a big consideration. Contentful is in the content management space and it's littered with competitors. There are literally thousands of CMS options out there for companies to choose from. But one of the big reasons that it shows Contentful is because we were well-funded, we were an established company, we had a big customer base. A big part of the Fortune 500 uses the product and so, as a product person, I hate to say this, but sometimes it's not about the product at all. It's actually about the company or the reputation or whatever. So I expect that that may also happen in the AI world Makes sense.

Randy Silver: 11:13
So, at HoneySales, what's the approach that you've taken to developing with AI? What's the lessons that you've learned early on?

Andrew Martinez-Fonts: 11:21
along the way.

Andrew Martinez-Fonts: 11:22
I think initially our assumption was well, we're going to have to build a bunch of machine learning models, we're going to have to get some expertise in machine learning specifically, we do have one engineer who is an expert in machine learning and that's actually been super valuable to us because he understands the.

Andrew Martinez-Fonts: 11:38
He comes from sort of a traditional machine learning neural networks world and understands what these models are capable of doing. But he's also working, not necessarily just building models. It's more about helping guide us through the technology landscape of AI so that we can make the right decisions, so we can take advantage of the latest prompting methods or what have you. And so I would say we take a I don't know, I've never described it in this particular way, but I would say we kind of take an AI-like approach, like we don't invest a ton in building our own models but and as much as possible, we try to use things off the shelf. There was a discussion at a panel earlier where one of the speakers said yeah, you know, we try to build a lot less stuff ourselves these days, right? So much software is available off the shelf as a subscription or as a purchase, and we should just be doing that more, because we're never going to beat the billing company that does subscription billing, for example. Right, that's never going to be an expertise for our company.

Randy Silver: 12:35
Yeah, if you've ever used Wardley mapping, it's moving things from being customization things that are uniquely generative for you or customization things that are uniquely generative for you, or moats for you, to things that are becoming more utilities.

Andrew Martinez-Fonts: 12:49
Exactly, exactly, and I think we all know that the increased cost of maintaining systems that are not proceeding at the pace of the larger market is kind of a waste of time.

Randy Silver: 12:59
So, when you are doing this, though, there are certain factors that really make a massive difference, things you have to really consider, and you mentioned a few of them. I want to ask you about what's the thing that you need to really look out for in terms of data quality? How do you know when you've got it right?

Andrew Martinez-Fonts: 13:14
Ah, yeah, I talked a little bit about data quality in terms of kind of garbage in, garbage out, the typical like make sure that we have the data we need. What we're seeing when we get responses from ChatGPT, which is the model that we tend to use, is that sometimes we don't expect what we wanted. One of the examples I gave today was around formality and pronouns. Like in German there are informal and formal pronouns, just like in Spanish and French, and our customers mostly operate on a formal level because it's Germany and business is a more formal practice. But actually a lot of customers are starting to use informal and turned out we were asking and we give our customers the choice, and a lot of our customers were getting messages back that mixed the formal and the informal and that resulted in a very weird experience for them.

Andrew Martinez-Fonts: 14:00
And so, my most esteemed colleague, how you doing? Yeah, kind of exactly the way the Germans address their letters is most honorable, mr Silver, and they go on from there. And then, yeah, how you doing buddy? And that didn't work. And so we've used different methods to both validate the data that we're getting out of the models and then make sure that we can also put in some automated stuff to check whether this is right in the first place. So a combination of human beings and logic and what someone I talked to last week called sort of the dirty side of AI is like yeah, there's still human beings back there Maybe check them and things. And we've seen this throughout the. You know, a lot of the social networks obviously employ a lot of human beings to make sure that the content is safe and all those things, and I think we're following that pattern a little bit.

Randy Silver: 14:54
And as the company grows, you need to start worrying about speed as well.

Andrew Martinez-Fonts: 14:59
Yeah, that's right, we've, you know, as we start processing thousands and thousands and eventually millions of these signals for our customers, what HoneySales does is we generate, we try to find signals for salespeople about their prospects. So you load your prospects into our product. We try to find signals about what those prospects are doing in the world so that you can have a more kind of contextually relevant outreach to them. But the average salesperson has 1,000 to 5,000 prospects, and those prospects are generating dozens of signals, sometimes a week, and that adds up to a lot of signals.

Andrew Martinez-Fonts: 15:34
And, yeah, it turns out that making requests to and from ChatGPT takes time and slows our processes down, and so we're kind of in this. You know, this has really virtually nothing to do with AI. This is really about data processing. But then you run into these challenges that are like, well, I really want to use AI, but actually I'm now dealing with this data processing issue, and so there are kind of side issues or side effects that you've got to deal with that you might not have thought about otherwise, because you just think, well, this is just going to be fine.

Randy Silver: 16:03
And then comes the big side effect of cost, and so how are you dealing with the cost? This can get exponentially bigger as well.

Andrew Martinez-Fonts: 16:11
Yeah, that's true, we make it into our unit economics. So one of the things I've been careful about doing as we and this is not just about using large language models, but also about buying other software products right, like we're buying all kinds of side features that we need to make our product better and I've got a model that basically takes into account for each new customer who brings a certain number of prospects. Like this is how the economics work out, and there is obviously some optimism that the prices will go down over time and that as our volumes increase, we can get some volume pricing and all those things. But yeah, it is a consideration Absolutely Like there was a moment where I realized that we couldn't provide the product at the price we had been talking about because of these considerations, and so we've got it. We've got it worked out at this point, but like that's something I wish I had started doing earlier.

Randy Silver: 17:03
And, as you said earlier, you're still at the scale where it makes sense to potentially use chat, gpt. Where does it make sense to start actually investing in your own models?

Andrew Martinez-Fonts: 17:10
Well, we haven't done the analysis. I presented something today which was the work by a data scientist named Skanda Vivek. He published an article that I had found and he talked about. He did an analysis about a year ago, so a little bit out of date because things are moving so fast. But his analysis was roughly that if you want to use ChatGPT, you're probably saving money if you're using up to thousands of requests a day, but once you tip over into the millions of requests a day, you should probably just deploy your own hardware, use open source models and deploy them on your own machines, and it's cheaper for you to run those servers than to pay for each token from ChatGPT. So I know those are very fuzzy rough numbers.

Andrew Martinez-Fonts: 17:52
Directionally it makes sense, but directionally it's like look, if you want help writing your term paper, it's probably going to cost you $3 on ChatGPT, but if you want to set up a server that's going to be $100, and obviously the costs are different. But yeah, it's a factor, and speed would also improve. If you've got your own servers you have sort of more probably performance from them or you can at least tune them up and down increase. If you've got your own servers you could probably tune the concurrency rates that you can deal with and things like that.

Randy Silver: 18:22
Okay, andrew, this has been great. I think we've got one last question on this. So you said we're building less these days. We're using more of other stuff. Yeah, we can still do delighting customers that way, but how are we able to really get into things that are hard to copy and margin enhancing if everyone has access to the same set of tools?

Andrew Martinez-Fonts: 18:43
Yeah, it's a great question, and this is really one of the big challenges that I think any company has today trying to compete or trying to keep up with incumbents or what have you. In our case, what we've tried to do is identify those core aspects of our product that are part of our IP, and everything, almost everything else, if we can, is to outsource it. So you know, this is the same pattern, I think, that companies take when they buy a content management system. So, going back to my experience at Consentful, it doesn't make sense to build your own content management system anymore. If you talk to any engineer who's more than 30 years old, they've built one themselves, right? Because everybody thinks this is so easy, it's easy to do and in the end it gets too complicated and too expensive to maintain and all of that, and so that's why Contentful was benefiting from that to a large extent that look, the product is there.

Andrew Martinez-Fonts: 19:33
This company is going to innovate and create better solutions for you much faster into the future, and so one of my sort of strategic perspectives on HoneySales right now is like let's not waste time building the stuff that we don't need to right, that we can just use off the shelf and that doesn't actually provide competitive mode for us. And so there are a few things that are competitive for us Like. One is about picking the right signals, another is about generating the right messages, another is about evaluating the prospects themselves and figuring out how do we approach these people things like that. But pretty much everything else, yeah, let's buy it, because it'll just be easier and it'll cost us less in the long run. Fantastic. Thank you so much. You're very welcome. Happy to be here.

Lily Smith: 20:28
The Product Experience hosts are me, Lily Smith, host by night and chief product officer by day.

Randy Silver: 20:35
And me Randy Silver also host by night, and I spend my days working with product and leadership teams, helping their teams to do amazing work.

Lily Smith: 20:44
Louron Pratt is our producer and Luke Smith is our editor.

Randy Silver: 20:48
And our theme music is from product community legend Arnie Kittler's band POW. Thanks to them for letting us use their track.

Podcast

How CX works with Product at Ikea – Hanneke Hoogewerf (Sr Strategic Experience Designer, IKEA)

How to prioritise your work for impact- Nacho Bassino (Product Leadership Coach)

How Google makes AI work at Enterprise scale – Miku Jha (Director, AI/ML and Generative AI, Google Cloud)

How to achieve better results by changing your mindset – Connor Joyce (Senior User Researcher, Microsoft)

Product-led growth in action – Partho Ghosh (VP of Product, Uberall)

Using AI in healthcare products – Tetiana Telenczak (Head of Product, Merck Group)

How to improve your relationship with data – Heather Williams (Director of Product, Elsevier)

How Citi is accelerating AI in banking – Tariq Maonah (SVP of Product and Engineering, Citi)

Why can’t we get accessibility right? Claire Bauden (SVP Global Product Design, Global Payments Inc.)

Recommended

Product-led growth in action – Partho Ghosh (VP of Product, Uberall)

Product outcomes and company values go hand in hand – Brian Walsh (SVP of Product, Pendo)

Navigating AI bias: Insights and strategies – John Haggerty (Founder, The PM Insider)

How to improve your relationship with data – Heather Williams (Director of Product, Elsevier)