Transcript: Sigalit Sadeh | Lusha | How to Do Smart Opportunity Sizing

Intro and meet Sigalit

Eyal David: [00:00] Hi everyone, today on the episode our guest is Sigalit from Lusha. Sigalit is a super smart, experienced product person with a lot of wins under her belt at companies like ClickTale and Dropbox, and today she leads AI-based recommendations at Lusha. I really enjoyed talking with her, and I'm sure you'll get a lot of value and really enjoy the conversation too. Sigalit brings rich experience building smart data- and AI-based products, and above all a very structured way of thinking about problems, prioritization and decision-making. We talked about building a product in a saturated market, especially now that the pace has become much faster, about moving quickly from problem to solution with a business focus, about using AI well in day-to-day practice, and especially about how to do sizing of opportunities in the most practical way. So, as I said, a very practical episode. Enjoy listening. Hi Sigalit, good morning. Good morning, how are you? Really great, really great that you came. Great to be here. Really, right? Really. Let's see if it's going to be fun. So tell us a bit about yourself so we get to know you better.

Sigalit Sadeh: [01:10] Yes, I'm Sigalit Sadeh, Director of Product at Lusha, and I lead the AI and data recommendations area. Overall I bring more than fifteen years of experience in product, and eight years building and leading product teams at different companies. I've worked at Dropbox and HP. I think my forte is really the combination of smart platforms, data, AI and machine learning solutions, and personalized user experiences.

Eyal David: [01:39] So I, I mean the forte, so one second, let's go in here for a moment before we continue to Lusha. In your view, what's the most important product skill for a product manager?

The most important skills for a PM

Sigalit Sadeh: [01:49] Great question. I think of three things mainly. One, product managers who are obsessed with finding the perfect problem or opportunity. That packs a lot of things into it: customer obsession, data obsession. Basically the ability to look and understand which problem is worth solving or which idea is worth building. Because in the end we're flooded with problems, we're flooded with opportunities, the market is spinning around us, and you really need to do some kind of synthesis between what we want to do and what's worthwhile and profitable to do. The second thing is the ability to execute. We live in a world of uncertainty, and many times we need to lead our teams and execute the strategy even when not everything is clear, and put all kinds of small stations along the way that give us an indication we're on the right path.

Eyal David: [02:46] Like milestones, but also, like we talked about in the prep call, right? Right. About the need right now to move fast in the current period?

Sigalit Sadeh: [02:54] Right now it's really crazy, with all the AI happening around us and the market moving terribly fast. You can develop fast and you can validate terribly fast, and we also have to move fast, not wait until we have the perfect vision, but really move fast, set small milestones, and keep checking that we're on the right path, that we defined success the right way and that we're on track. If we're not, that's the time to pivot and rethink it. What's the third thing? The third thing is accountability. We're here as product managers to deliver impact, so we need to see that we're progressing in the right direction, that our metrics always reflect success, and we make sure the business is profitable. We don't build ideas or aspirations; we really solve a real problem for customers and deliver value to the business.

Eyal David: [03:52] So that also includes knowing the area really well, right? Right, knowing the area, knowing the users,

Sigalit Sadeh: [03:59] knowing the company's main metrics, knowing how to define success metrics well and precisely.

Eyal David: [04:06] What do you expect, say, from someone who comes to you for an interview, what do you expect in these areas, what level of readiness do they arrive with?

Sigalit Sadeh: [04:14] The truth is I always do product intuition exercises with them, so I expect them to come curious. I'll probably give them a problem that's not well defined, but on purpose.

Eyal David: [04:26] Like what? Now I'm going to reveal to you the... Get people better prepared for interviews with you.

Sigalit Sadeh: [04:34] I'll ask you what your favorite app is. And then I'll ask you what you'd improve in it and why. But to answer a question like that, you basically need to go through some thought process. Because I will ask you, during the interview, who you're solving the problem for. And if the problem is yours, do you represent all of the app's users? And if you tell me yes, then I'll ask you, okay, what are your underlying assumptions? How do you know that people similar to you use the app, or at least what's the hypothesis guiding the thought process? And how painful is it really? Is it something that's a must have? That if I solve it, everyone will use it? Or is it a nice to have? Meaning, it's really nice that it's here, but it's really not essential that it be here, and then not everyone will use a feature like that. And if it's a nice to have and not a must have, then they probably won't pay for it either.

Eyal David: [05:32] So we can't. Let it go into the backlog. Exactly. Okay, so let's move on to talk about Lusha. Describe Lusha in a few words so we get an idea.

What Lusha does and where AI is taking the sales market

Sigalit Sadeh: [05:39] What does the company do? So Lusha is a company that sells to B2B companies, to their sales organizations. It helps them reach prospects much faster, with much more effective data, understand them, and helps them act much more precisely. How do we do it? We provide very accurate data on both business profiles and companies. The transition we're making now is basically to person and AI. And why are we doing it? We discovered that salespeople aren't really busy selling. 28% of their time they're actually selling; all the rest of the time they're doing research, they're fixing their CRMs and their spreadsheets and their lists. And then they sync all the systems and go back and check. They use 9 to 10 tools on average to collect all this information, to put it in one place, to understand who to talk to and what to talk to them about, but it's not a one-off, it repeats with every lead. And here we stopped and said, what can we do for them? We also discovered that they really, really don't like searching and doing this research. They're salespeople. They're salespeople, they want to sell, they want to talk to people, they want to do their job. And that's our goal: to bring AI in so it takes all the tedious tasks off them, the ones they're not supposed to be doing, because in the end AI can understand data much better than they can. It can also look at all the data in the world, and we'll simply bring them: who do you need to talk to? Why do you need to talk to them? Now, why is it relevant? Why will they actually talk to you and close the deal with you? And what do you need to talk to them about, so that it will draw their attention? We'll do all of that for you. You'll talk to them, you'll focus on the good side of selling, and that's what the AI will do for you.

Eyal David: [07:26] Amazing. And tell me, all this shifting to AI is basically happening around the fact that there's AI too, but that's clear here, I think, to the customer, but what other trends are happening in the market right now that made you decide to go for AI, decide to save time for salespeople, only from an AI standpoint or has the nature of the work there changed too? What does it look like? First of all, the nature of the work is changing. Tell us a bit about the market, about what's going on in it, it's really interesting. Right.

Sigalit Sadeh: [07:55] AI has entered all sales organizations very strongly. The traditional cycle of salespeople in general is that you do some discovery, look for the relevant accounts or companies, then do deep research inside each of the candidate companies, move to a stage of scoring and prioritization, find who to talk to, and then send out some outreach. That stage keeps getting shorter and smarter as time goes on. People are no longer looking for static data. Tell me who I need to talk to. What are the signals that show me there's very, very high relevance here for me? There's a kind of transition from volume to relevance. I want to talk to the right people. I don't want to send generic emails and work at terribly high volumes, or meet and do 100 demo meetings a day. No, I want to do three and know for certain that those three will close with me.

Eyal David: [08:50] Quality, basically? Exactly. And when you talk about static, explain the difference for a second.

Sigalit Sadeh: [08:56] Static means I don't see data changing over time. I know what's true now, I don't know what was in the past, and I certainly don't see what will happen in the future based on all the signals happening around me.

Eyal David: [09:08] What other trends are happening in this industry, say in terms of competitors? You probably have a lot of strong competitors.

Sigalit Sadeh: [09:14] Right, we have a lot of competitors: ZoomInfo, Cognism, Apollo, Amplemarket, Clay, they're all building solutions in these areas. I think we see the same trend in all of us; we're all looking at the same problem from the same angle, we're all heading toward recommendations and personalized outreach. We took a somewhat different approach here: we don't want to do only recommendations or only personalized outreach. We want to be the little copilot next to you that always tells you what your next action is. You're never stuck, you always have the relevant lead by your side, you always know what happened with it and what you need to talk to them about so it'll be relevant and convert.

Eyal David: [09:59] Maybe tell us a bit about what's ahead, your plans, whatever you can share, or things you launched recently, so we understand what's going on inside the product?

Sigalit Sadeh: [10:07] Okay, so we had two new features we launched very recently: recommendations of business profiles you should talk to, and new accounts you should discover. And in parallel we also launched the playlist, which is basically your ability to create a list that's entirely based on recommendations, and that list keeps growing day by day with new leads. It also connects to personalized outreach capabilities. Very soon we'll add the ability, for every lead we recommended, to understand who to talk to inside the account and what to talk to them about, so the outreach comes out personal,

Eyal David: [10:50] all of it in two clicks, almost effortless. That's what you mentioned, and it really connects from data to actionable insights, which you mentioned earlier, because what? You measure the salesperson's time, so how do you actually know, how do you collect the feedback beyond the 28 percent they told you? Do you measure like, say if I'm Gong,

Sigalit Sadeh: [11:14] then I know because the person logs it for me in the tool. How do you know how to close that loop? We still don't know how to close that loop well. We talk a lot with customers, and they really tell us how tedious this process can be when it's manual. Just think about the process they go through: they need to understand what the account is and then go and collect lots and lots of data on it. On average they use between 8 and 10 tools to collect all the information they need. Then they need to understand, analyze, who to talk to, because many times sales organizations will talk to different people, in different industries and of different sizes, and then they need to understand that with this specific persona, in this specific department, I need to talk about X, and with someone else I need to talk about Y. And this whole process is terribly manual. Our estimate is that we save them 6 hours on average, but right now we really don't have 6 hours a day, and we don't really have a way to validate it, because, like you said, we're not Gong-ing the call transcript right now. Right now, you said it.

Eyal David: [12:19] Right now. Okay, so what challenges are you dealing with? Because you're basically describing a situation, we said, there's a saturated market, there are shifting trends, a saturated market and everyone's kind of racing to get there. Really a red ocean. And we talked about it, also what's expected from PMs in general, you, now, in this whole situation, deals also with very good spelling, you talked about defining the problem really well, and deciding it's impactful enough to move on it at all, and then planning it really well, and being, that Master of the Domain, I don't know what we called it, like really knowing the, how do you do all of this while you need to run terribly fast, how does that happen?

Working at speed: quarterly planning and user conversations

Sigalit Sadeh: [13:02] Right. A challenge. A challenge. We all live it, but how does it happen with you? If I once had the luxury of planning a year ahead, I don't have it anymore. We plan a quarter ahead at most, and like I said, every two weeks we always stop and ask, are the problems I'm solving still the problems that need solving? We do a very deep pass over the metrics, go over user feedback, and all the time, at least twice a week, we talk with users. All the time. It's a routine you can't give up in this world. And of course, keeping your eyes open to everything happening in the market, not just with competitors, certainly what's happening around us in AI, to understand which tools, which abilities were added to these capabilities, that can accelerate us and let us deliver even faster and provide higher value. How do you stay, like, how do you stay on top of things? Curiosity. What works, at least? An hour a day, at least, I read.

Eyal David: [14:01] What, on Twitter? No, on LinkedIn. Got it. Okay. And let's talk, really, about the matter of prioritization, super interesting in my view. Again, in the prep we also talked about the book, Teresa's book, Continuous Discovery Habits. Right. So when you talked just now, I understand, about talking with customers every two weeks, that's really also one of the takeaways in the book. Right. But let's talk, how do you do it? How do you really approach it, and how do you guide the product folks under you, or who work with you, to work on the most important things? How do you do that prioritization?

Prioritizing and diagnosing problems with business impact: the Opportunity Solution Tree

Sigalit Sadeh: [14:40] I'm getting into it, but one second, it's important to me, since you already mentioned Teresa, to talk about one sentence she wrote in the book, and it goes like this, I'm going to read it: Stop falling in love with solutions, and start falling in love with problems. And that's so, so true. She has a method she calls OST, which is the Opportunity Solution Tree, which is basically a framework you can use. You really don't have to draw it, or strictly follow the methodology, but simply practice the thought process. You always start with the outcome, with a business goal. What do I want to improve? Do I want to improve engagement, retention, do I want to do acquisition of a new type of customer, do I want to enter a new market? You start there, and from there you go one level down and map all the problems, or opportunities, that exist, the ones blocking us and preventing us from reaching that goal. After you map all the problems, you take each problem and find at least two different solutions for it. This stage is very important, so we don't fall in love with solutions, and so we practice very, very diverse thinking. Yes, there's never just one way to solve a problem. And sometimes there's a really easy way to solve a problem, and you don't have to... That you understand very well. Exactly, because you understand it very well, and you look for the Pareto, you go 80-20, you don't have to achieve the full hundred percent. And now, when you look at these problems, you can also suddenly connect problems,

Eyal David: [16:08] that's also something that happens when you break them down. Absolutely. So you work this way, I understand.

Opportunity sizing: three simple questions that lead to a smart choice

Sigalit Sadeh: [16:14] I, and certainly my product team, yes, work this way. I encourage all the PMs to do this process themselves all the time. I never define for them what to research and what to check; it's the opposite. I want them to come to me with the problems, with their hypotheses, with the success metrics, and explain to us, explain to me, why this is a problem worth solving. Why this is an idea worth developing. And yes, I have to admit that not every idea or every hypothesis always succeeds. There's always a margin I give for mistakes, and I don't even call it a mistake, I call it guided learning, because you have to learn hands-on, and I don't always have all the solutions either. What I'd do, what I learned in the past, certainly today, isn't always right, so it's important to give that space to experiment, because we'll always discover new things, and if we don't do, we won't learn. Absolutely. And how do you do that sizing? Okay, that sizing is basically about answering three main questions. One is quantitative: how many people does it affect, who is the segment I'm going after, how many people does it affect, is it inside my customer base, do I need to go and acquire it, what's the business value, is it a feature I can monetize. Then I get into the level of pain, like I said before: is it a nice to have or a must have, how painful is it, how much are they trying to solve it themselves today and failing. And then finally, the third lever is: what's the frequency, how often do they encounter the pain. Because say I found a terribly, terribly painful problem, but it only happens to very niche customers, or terribly, terribly large ones that don't represent my customer base. Or even, it's a terribly painful problem, but if you encounter it once every six months, then we probably shouldn't solve it. There are exceptions. There are exceptions, yes, but you need to be very mindful of all these measures to understand whether this is a problem that should be solved. After you answer all these questions, I think you have like a solid answer in hand,

Eyal David: [18:24] and the PM already knows how to answer it themselves. You know, no, that's very true, and there really are exceptions too, if it's enough that, just, your power user needs that dashboard to take a screenshot for the quarterly presentation, then probably, I don't know if it's painful enough. But there's always the aspect of, okay, I really have a few initiatives, I've probably researched them very well, now go compare between them, and then always, if there's no dollar sign, which isn't always, not in every product is it a given, that's a point that really interests me. So we wanted to go methodically, let's go methodically, but in my view that's the place I'm most curious about. So say you're working with your PMs, there are the company metrics that are the most important at that moment, right. So basically you look at it once a quarter and derive the initiatives from there. If I now come and think of something else that hits the dollar sign, but that probably isn't connected to what's currently important to the company, does it also carry weight in the prioritization? What does that look like?

Sigalit Sadeh: [19:22] Absolutely. I think at Lusha we give, not everywhere, I have to admit, have I encountered this, but at Lusha, if there's a dollar sign, and if we manage to attach value to a feature, even if it's not within the company's or the group's metrics, we do enable the ability to move forward and make an impact. Because in the end, product owners are here to deliver business value, and even if it's not in the ways we as management thought of, I think it's really important to let them do it.

Eyal David: [19:51] Yes, say all those that... there are technical features in general that save the company money. Right. Those are actually easier to quantify, usually, I think. Right, because then you really know how much money you saved,

Sigalit Sadeh: [20:04] but every feature can be too, if your assumptions and your ballpark are good enough. You know what the business value is, you know what the business value is, you know how many customers you'll win back, and then you can quantify their subscription. You can look at conversions and signups and see how many new customers you brought in, or how many customers did upsells following features you launched. In the end, with that same framework, you can really understand how much business value you brought to the organization, also in hindsight,

Eyal David: [20:32] both before and after. Yes, after is always possible. So okay, so I went, I went, and along the way I really understood the problem well, I have a list of problems, I understand who they affect, we probably gave them some measure so we can compare between... What are the metrics you work with,

A bit of practice: from an opportunity score to ICE

Sigalit Sadeh: [20:52] what does it look like? Now there are two measures. There's one at the opportunity level, like I said: how painful it is, how often it hits me, and quantitatively, meaning, who the segment is and how much of it. It really is a mathematical formula that at the end produces some score, and then we've chosen the problems. Now every problem has one solution, sorry, two solutions or three. Now you need to rank the solutions within. And there we get into ICE methods, meaning you need to compare impact versus effort.

Eyal David: [21:25] Okay, let's slide over there in a moment, we're still on problem prioritization for a second, right? Right. My question is this: you basically give your two cents already at the level of prioritizing the problems, before we even talk about solutions, and then we go and think about solutions? Absolutely. Okay. And then we go and do the big prioritization, and then we see how we weave all of it in. In, right. And in terms of zooming in, like, with the hypotheses themselves, you say there's a formula, but that formula, how much does it change per company, for that matter? What does it look like? How does it go in, if I, say, to my teams, to my organization? What does it look like? The thing is that it changes...

Sigalit Sadeh: [22:05] It's tricky. The sizing will probably be different for almost every problem you come to solve. That's why the thought process and the questions are what matters here. Let me give an example. If I want, for instance, to be able to sell more data at Lusha, I'll probably go in and check what my strongest regions are, or where I have the most users. And if I come and build a product that's zero to one, my hypothesis will be based entirely on assumptions. How many users complain about a certain problem, or have a certain need, based on what they do today, how often they'll do it, how strong the pain is, how they solve it today with other tools, and how often they repeat this action day to day, and then... I basically can't compare the prioritization of the two, and that's how I'll break down every problem. And every problem really has its own opportunity sizing,

A simple recipe for an organization: derive everything from the metrics

Eyal David: [22:59] because it's basically a thought process. Let's, for a minute, really focus on SaaS companies, of a size of, I don't know, over 100 employees, which already gives us some criterion for company size in terms of revenue, and basically we now want to focus on... So let's say we now focus only on retaining people, and look at a segment. We probably also need to divide the segment by how much revenue it brings us, or how active it is, right? What's the simplest way, basically for any organization, because there are so many organizations that don't do this. Right. What's the simplest way to take an organization of this size, focused only on retaining the existing base and keeping it, how can it take this magic and bring it into an organization in a simple way? Because I don't want to give some big recipe, or send people to read a book, because cool, we'll add a link, but I'd like us to come away with some simple recipe

Sigalit Sadeh: [23:54] that you can apply tomorrow. Very simply, it's working with a set of metrics. Meaning, the VP needs to know what their metrics are and where they're driving the teams. Then you need to understand what the group's target metrics are, and from there you need, you can start deriving it downward. That's the first stage. From there, from defining the goal, everything basically derives. Let me explain. If we defined retention as our top-level goal, we go to the existing customer base, we understand how many of those we have, we understand what drives repeat usage, what they're missing, and then we start building the problem tree. Here it's very easy to do sizing, because we already know the customers, they use our systems, we have all the metrics, we see where they get stuck in the funnel, and we can do really precise sizing. We see which funnels get stuck, where we have a drop, and then we can really get into that process. It's a bit harder to bring this in, as I said, for new features, for zero to one, because then it's a kind of muscle, some research we need to develop. Let's give an example: we want to do expansion and enter another stage in the user's funnel, not our product one. We want to expand usage. It may already be that we're talking about a different persona we need to go and interview. They're no longer with us. We need to really understand their process. We need to develop tools that help us quantify the sizing. And take, in the end you can quantify it, make a real formula of how many times a day, or a week, or a month, depending on our product, they repeat it, try to rate the level of pain. Is it high pain, high, just like you do in T-shirt sizing, low, medium, high, and give it its score. And try to estimate what the TAM is, meaning, what the total addressable market is. You can project based on our customer base and based on how much we can penetrate. You can go really wild with this data, and understand what the expansion base is based on the customer base we have now, and arrive at a very solid sizing formula. Wow, and there's something else here

Eyal David: [26:10] that I thought about along the way. If until now, really, probably, this calculation of, well, probably their solution will be, maybe they'll open some hours in it, maybe they'll develop something good for themselves on the side, some thing, maybe Excel, today it's terribly easy, like, to also build something, a bomb on the side, so that suddenly, suddenly they can deal with their problem much better, maybe outside your product, so, like, that's something you might quantify and understand from them in the conversation. That too. So what does it look like? I think you also have a lot of insights from the book around customer conversations, I hear it, because it seems you do it very well. What are the best takeaways? There, you're also smiling in agreement, so I'm happy. What are the best takeaways you took from the book in this area,

Customer conversations and The Mom Test

Sigalit Sadeh: [26:52] or in general? Also from the book, I don't know if, the book doesn't really put an emphasis on it; there are other books that emphasize customer conversations. I was hoping you'd say that. The Mom Test, for example, an excellent book, I really recommend reading it to anyone setting out to do customer conversations. The advice: never ask about the products you're building; always understand the process, understand how they work, what tools they use, how many such tools, what the pains are, what works better for them, what works less well, what they'd pay to get rid of in their day to day. Why do I recommend the book, The Mom Test? Because it really gives very, very structured practices for how to question your mom about a product you're building so that she gives you objective answers. That's terribly, terribly important, because many times in customer conversations we're looking for confirmation of a feature we thought of, we want to build, because we really want to get past the validation stage. So be very, very careful about that: always talk about the problem we're coming to solve, and not about features, and not about solutions. And here we'll run into it. From my experience,

Eyal David: [28:00] I very much agree. From my experience, the "what would you pay for" question, I'd ask, first I'd let them list the top pain points, and then, of those, which of them would you pay for. That's more, usually the question

Sigalit Sadeh: [28:13] can also kill a conversation if it isn't asked. Yes, definitely. It needs to come really, really at the end, maybe even in the second conversation, you don't have to hit it in the first conversation, and it needs to be very, very planned, because we can't just ask them, would you pay for it? We can also get a very insincere answer; they can say yes, and in practice we'll meet a very, very different reality. So customer conversations need to be planned, they need to be structured, so that we also have a basis for comparison between customers. If it's freestyle from customer to customer, we won't be able to make a decision. It has to be the same set of questions, they have to be detached from the solution and focus on the problem and the workflow, and in the end we need to come away with an understanding of, one, whether they'll use it, not from them, but from the conclusions,

Eyal David: [29:02] and whether they'll pay for it. You asked how they progress, it also depends a lot on who goes with them all the way, it's also how much they're willing to pay. That's really, that's a really good measure for that. Right. But I already have customer conversations,

Sigalit Sadeh: [29:14] I can say, even from the last month, when a customer who really identifies, thought process, because we didn't talk about features there, something that solves a problem for them, they say, I would pay for it. They actually say it themselves, we don't even need to ask. Which is the most fun thing to hear. So tell me, I'll ask now,

Eyal David: [29:35] we have a new AI corner, so really around AI, recommendations you'd like to share. You can think in the direction of opportunity sizing, you can think about user interviews, any recommendation for a tool you discovered, news, news that happened, what would you like to share?

Using AI to shorten discovery and understand the market

Sigalit Sadeh: [29:50] I'll share. I'll connect to opportunity sizing, since we already started, so let's continue. I'll say that if it was once very, very, very hard to do opportunity sizing and build some product, today we have AI tools. We can do prototyping with an effort that really amounts to an hour or two, it doesn't have to be perfect, and go out to validate with customers very, very fast, with tools like, I can mention, Base44, Lovable, even with Claude, it has all kinds of integrations. And we've already done it; we really managed to explain very complex concepts, because when you come to build data products, recommendations, agents and so on, it's sometimes terribly hard to explain what the value proposition is. When people see it, it's very easy to communicate it inside the organization, also to executives, and within two hours you can be in a customer conversation and start getting feedback, even seeing them do initial interactions. For me, that's a gift, because before, if before we had to plan, and go down to design, and invest in prototypes, and do usability testing, then between an idea and usability testing it took us about two weeks; today it's a matter of two hours. Great suggestion, also for researchers who need to,

Eyal David: [31:07] let's focus on that, let's focus on that, because it really has become easier. So how do you do it? Very easy. I build myself an agent,

Sigalit Sadeh: [31:14] or use deep research, and start asking it questions. An agent in GPT, like. Exactly. What is the total addressable market of a specific thing? Describe to me what the workflow of an SDR looks like, of an AE, of whatever persona you're working with, how painful this thing is, what tools exist in the market, who my competitors are, who built a similar solution. The research, the research for opportunity sizing becomes so easy. You can get at least 60% of the data through AI, very fast.

Eyal David: [31:46] Lovely. We also do something, yes, write things down on the side, because between one answer and the next you can lose things. Right. So I write things down all the time. And tell me, what other uses do you make of AI in general? Say around discovery, around conversations, the things you do?

Insights from analyzing customer conversations with AI

Sigalit Sadeh: [32:05] What? Of course. First of all, customer conversations and always extraction of insights go through AI. If once, really, going back a few years, we had the comparison matrix and we'd do the extraction, today we come out of conversations, run them through AI transcription, and we have all the insights we need. I also feed in the whole customer base and start getting patterns of things that were said, pains, opportunities, things that are repetitive, and I get them quantified. The opportunity sizing sometimes comes out of these analyses all by itself. You can also ask a question, meaning, some opportunity I assumed was an opportunity, after thirty customer conversations, do you see that it actually exists there? It'll probably immediately tell me that it encountered it ten percent of the time, fifty percent of the time, or eighty percent of the time, and that's really, really, really good input, really, really, really good input, into opportunity sizing. I also do this a lot on customer feedback: I simply slice three months, six months back, and analyze. Really nice things come out. It also knows who reported the same problem more than once, and how many of those there were. And that refines

Eyal David: [33:19] the opportunity sizing in a different place altogether. Cool. I think that's also interesting, because suddenly it becomes more dynamic, because say you did your opportunity sizing on day one. During conversations insights come up, right. It may also change there, and you're very eyes on the prize about what you're asking right now, and you don't notice that maybe some opportunity has now dropped in priority, because people talk about it less. So that's really interesting. Right. It's like detecting trending, right? Absolutely. Cool. Tell me, how many PMs do you have on the team? Six. And they all work with AI? Is there such a thing as a day without? There's no such thing. I think,

Sigalit Sadeh: [33:55] with us everyone works with AI, the six on my team all the time, and also worldwide within Lusha, because it simply makes us deliver better, deliver faster. It's not a replacement in any way, but it shortens our discovery processes, the research, the opportunity sizing, the conclusions from customer conversations. Just as we got, or are trying to get, salespeople to focus on selling, we really need to focus on the product management side, really do this opportunity sizing, and keep understanding two things: that we're solving the right problems, and that we're on track. And that you're working on, yes, the right problems that bring in money, that can be delivered in a relatively simple time, and with an easy solution. And you measure afterwards too. Cool,

Eyal David: [34:45] so if you can tell us a bit about what you launched recently, or are about to launch, so we get an idea? We recently launched recommendation engines, basically,

Lusha's recommendations, and how Sigalit got into product through HP

Sigalit Sadeh: [34:53] recommendations of business profiles we think you should contact, and new accounts you should discover in order to promote sales. You can find it in two forms in Lusha today. It's inside prospecting: when you come to search, before you come to search, we tell you, wait, come see recommendations that are tailored for you, before you go searching. So we've already saved you that side of things. Exactly. We'll tell you who you need to talk to, don't go searching, don't work hard, and we'll also tell you why. And you can find it in our new playlists too. Why a playlist? Inspiration from Spotify. Basically what we wanted to create here is, just like Spotify builds us daily playlists based on my mood, based on what I heard yesterday, we'll build you the perfect stream of leads you need to focus on today. That basically keeps updating based on the past and... Exactly. We look at what you do, how you use the system, and we build you what fits your business right now. Tell us about how you got into product. Oh, this is really interesting. Okay, so I came to the world of product from the world of development. There was a profession called data architect. And I was among the first data scientists, when this field was really, really new; I even filed two patents. But during my work as a data scientist, I noticed there were lots and lots of repetitive reasons that were holding us back, causing us to deliver, to answer a business question, in nine months, not really understanding the value of machine learning inside organizations, what value it can provide, what problems it can answer, mainly because the business is always impatient. Answering a business question after nine months means I answered it and it's no longer relevant, and you also need a very, very unique skill set to solve problems of this kind. When it all started there were so few of us. And then I was working at HP, and we had a special accelerator program: anyone could come with an idea and advance it, to pitch it to management. And I really came with an idea for ML automation, a no-code environment, something we're really used to seeing today, it's really standard, but back then it was really new, to significantly shorten, to bring development time down from nine months to two weeks at most, so we could focus on preparing the data for the algorithms, and not constantly run and figure out the algorithms and tune for accuracy, which the machine can do on its own. So I really entered the accelerator program, and after two weeks I stood in front of management in a baseball suit and explained to them, I pitched them the predictive analytics power, and I also demonstrated it in practice on win-loss prediction for HP's internal sales team, and how we got them to convert 30 percent more. And I got funding to develop that product back then, on one condition: that I lead the product as a product manager. How ironic. It really was... Deadlines too, like deadlines? A plot twist. I asked them, why? They explained to me that as a data scientist I managed to sort of lift my head above water and say, no, we're doing this wrong, let's do it differently, and much faster, so product probably suits me. So I said okay, I'm willing to try. I went into it and fell in love, because it really combines everything I love: data, technology, forward looking. And that's it, since then it's love, since then and 15 years forward, it really is an ongoing love. And since then I stopped sleeping. Also. Yes, you can't, product managers don't sleep. And I think that now, actually,

Eyal David: [38:50] with AI, then, it's supposed to make a lot of processes more efficient, I think, and it can also simply help with priorities, I don't know what to do, or do it better.

Dropbox: metadata and search at scale

Sigalit Sadeh: [39:02] And how did you actually get to Dropbox? I had two hops along the way. First I went through two startups, and then a bigger startup dealing with behavior analytics, and then I moved to e-comm, where I built AI solutions for luxury e-commerce. And after about two years Dropbox approached me and offered me to join. And Dropbox was always a place that interested me. The suite of products, basically: everyone in Israel knows it as the app that backs up files, but no, Dropbox has a really big line of products, especially for image and video, and that's what fascinated me. I joined, and there I really led the whole area of metadata and tagging for images and video, at very high scale. It's a fascinating challenge: basically tailoring a personalized search experience for media people and the creative market, who search for very, very specific things. And then enabling them also to do visual collaboration on that, in a way that's personal and also helps them cut time in editing, in organizing, and in communication with clients around the images

Eyal David: [40:24] and the videos. What was the biggest impact you made there that you remember?

Sigalit Sadeh: [40:30] Wow. In search, it was an acute need for everyone who consumes images. Meaning, there's metadata that exists on the image, and they get paid when this metadata is well organized, and we at Dropbox basically didn't support displaying the metadata, or editing the metadata, or searching based on the metadata, and that was... that was really... that was critical. So basically everything we added, the support for reading, displaying and editing that specific metadata, and search capabilities on top of it, made a crazy impact. Later we also added manual tagging, when they really, really asked for it, because, for example, car companies, all their images were around cars, so they really wanted to organize them by models, by years and so on and so on. So that really was the biggest-scale impact we made at Dropbox. Later we also did all kinds of zero to one in the world of personalized visual collaboration, but from there I moved on to Lusha.

Eyal David: [41:42] Is there also some downside to using AI throughout the process, especially regarding the scale process,

The risks of GenAI and how to bring it into production properly

Sigalit Sadeh: [41:48] I mean, do you want to say a word about it? Right. It's very easy to build things with AI: prototypes, products, writing code, making estimates, research, doing deep research. Yes, sure, it saves us a lot. But to get it to production you need to pay attention, you need to manage the risks around it. If you ship AI models, especially GenAI, without context, without clear instructions, it may make things up. You need to make sure we're managing it, that we're mindful of what comes out of the GenAI, that we have accuracy checks, that we verify the structure: did the structure we expected come out, are the words we expected there, are the topics we expected there. We do send AI to check AI, but it's very, very important to check what comes out, so we don't get surprises. You also need to take into account that everyone is learning to use AI, and you can argue with AI. And if your products are offered for a fee, but you offer them through conversational services, whoever can argue can get it at a discount, or get it for free altogether. Pay attention, simply put guardrails in place, understand what, state what you're not allowed to ask your AI. That's especially true if you don't fully support conversational, for example, and there are certain, very specific things people can ask of you. So there, like, the system of protections needs to be bigger, and always, always, always

Eyal David: [43:19] pay attention to cost effectiveness. Exactly. And just one more thing, which is always that fine line, right, where the conversational is somewhat limited, and then on the one hand, like, you understand from the other side that they limited it because they wanted to minimize risks, and on the other hand, ugh, right, you could do it. I think that at some point this too, like, will go toward some kind of regulation. All this risk management has to have something around it. Absolutely, right. Cool, I think, I think it came out a great episode, I'm very pleased, thanks for coming. I'd like you to leave us, though, with something positive, with some recommendation, or something good you read recently, or some advice for young PMs, or PMs on the way, that you'd tell your employees, or people who come, see you, I'd really love for you to give something.

Closing advice: why curiosity is a product manager's strongest tool

Sigalit Sadeh: [44:07] It's a new world, everyone is learning it from scratch. We have AI, we have accessibility to lots and lots of data, but in the end there's no substitute for our own heads. The one who's able to do the prioritization I described, with opportunity sizing, the one who ultimately is able to make a decision, is us. The more curious we are, the more we learn from AI, listen to customers, know the market, really, the better decisions we'll be able to make.

Eyal David: [44:37] That kind of takes us to the place, that came out well, that kind of takes us to that very cheesy place of really understanding where you continue and where AI stops, right? But that's it, at the end of the day.

Sigalit Sadeh: [44:49] Right, right. AI here is supportive, and there are things that can't be taken from us. That really came to life.

Eyal David: [44:55] And maybe in the next five years, we'll see. Okay, Sigalit, I'd really love for you to come again, it was really interesting. If people want to get in touch with you, ask questions, and so on? You can on LinkedIn, I'm really available. We'll add a link. Alright, bye for now. Bye, see you. Hi everyone, thanks for listening. If you found this podcast valuable, you can subscribe and follow us, of course, for more episodes on Spotify, Apple Podcasts or any other app. Of course, if you didn't find us on some app, I'd love for you to write to us. I'd really appreciate five stars on every platform, and for you to follow us so more listeners can discover us and find the podcast. You can also find the previous episodes on any app, or on the YouTube channel; we have links in the description. Until next time, come on, be effective, and bye bye.