Transcript: Live UX Case Study: Fixing Onboarding Drop-Off and a Weak Dashboard with Bar Rachamim

Eyal David: [00:00] Hey Bar, how's it going? Good, you can't hear me... Great, I'm really glad you came. Thanks for having me. Bar, tell us a bit about yourself in a few words. In a few words — so I'm Bar Rachamim, CEO of the company ProgressBar, I have five years of experience in interface design, for a while I was at 1-0, at Login Friendly, and over the past two years I've mostly designed complex systems, including cybersecurity systems, I came from Vulcan, and a feature I designed for them helped raise $34 million, with Cyberloom which was acquired by Algi, and today I work with Gnostic. Great, and you also lecture, right? Yes, I lecture at Shenkar on user experience in Figma, and I've also lectured at Product for Product, for product managers, and at Reichman I also taught the intro course to the field of user experience. A woman of many accomplishments. How did you get into lecturing? What was it like? They just invited me. Oh, really? Just like that... so how long have you been doing it? Something like a year and a bit. Oh, very cool. Okay, so basically the reason I invited you is that we said, hey, let's put your UX skills to use, and you immediately said you could talk about a case — a spontaneous case study, actually — that we'll discuss. So this podcast was a bit different from what we've done so far — we're now going to take a case we built with AI, basically so there's some real "meat" here, and we'll use an example of a fictional company, and we'll analyze it together. Right? Cool. So we're talking about a company that you can look at on your end too, right? No? So it's called Data Insight, it's a SaaS company that has 250... Let me refresh my memory. It has 250 employees, it's in the analytics field, and their vision is basically to empower businesses of all sizes, so that they have analytics on everything happening in the business, and can make informed decisions, and their current goal is — again, according to GPT — their goal is to be the industry leader in this field, and to dominate some industry — it doesn't actually say which here. The data industry. They're theoretically competitors of Mixpanel. No, actually I thought — let's try to narrow it down, maybe even to a specific niche, but I left it vague, okay? Fine. Their audience is really companies with over 200 employees, to give an example, and they give examples here of industries — retail, healthcare, finance and so on. We'll also put a link to this case study in the show notes, and then maybe we can ask questions, look more deeply, and maybe suggest other ideas. So their regular, or typical, users are basically senior managers, heads of departments, and data analysts from those organizations, who need some robust tool to analyze their data decisions, where the data comes from all of the company's data engines. All good so far? Any questions? Okay. In terms of the product's value proposition, it lets you look at the data in one centralized place, and manage it — all in one place. There's basically very simple data visualization, easy to understand, and also a reporting system. By the way, I love that we dove straight into the practical side. I usually do that within an hour. So look, we're impressive and amazing. It's really about diving into the company, diving into the problem, understanding whether there even is a problem. Yes, we already came in with a problem, and then proposing a fix. And as I said, we're also big on very fast onboarding, so we show that too. Bottom line. Yes. The next thing, which moves to their value proposition, is basically the ability to provide predictive analytics and insights. I played around with things a bit here. And to enable collaboration and data sharing, and to be scalable. Okay. So, in terms of the market — any questions up to here? Anything else? So the predictions are some kind of forecast that I look at, so they give me some approximate figure. Do they want to look based on your past data, or based on the data that... Future insights, things that are about to happen, okay? Or where you're going to end up. But can I, for the sake of argument, enter data, and does that change the predictions according to that data? I don't think it would be right for you to just enter information arbitrarily. Let's take a specific case. Let's imagine some retail chain that has all kinds of analytics engines, that basically add information — whether it's physical, whether it's e-commerce, whatever. Everything flows into one place. So it seems to me that this is more data that arrives from a place, and less things that you feed in. Maybe — I don't know — you feed in the staff, which maybe changes, maybe they're split into various teams, and you can attach analytics to them, things like that, I think. Okay? Cool. Alright, la-la-la. And so we can get into the market here — there are all kinds of things GPT gave us, and they seem a bit too advanced and complicated, and just a tangle. Okay, want a user example? A user. We're still running in the background, okay? Give me a user. Okay, so a sample user is Sarah, who's from Nashville, just for the vibe. I love that she's a woman. Yeah, no, that's how we work here. She's a marketing manager, and she works at an e-commerce company with 250 employees. So okay, that fits with small retail for us. Her background is basically 8 years in marketing, so there's marketing that basically wants to see all its data in one place, so she can improve customer engagement, so she can reach her goals; she basically wants to see all her data in one place, and also get recommendations, okay? To date, is she on the Data system? Yes, she's been working for 8 years already, and she's tried a few tools, and basically, I also added for you that the advantage of Data Insight, right, is that it does simplification, there are a lot of English words here because I'm translating on the fly — it basically does simplification, simplifying all the incoming data and the various integrations, and offers a very user-friendly interface, lots of dashboards that lead, as I said, to actionable insights. I loved that you put "friendly" and "dashboard" in the same sentence, because that's not a given. Yeah, I also love that we keep going in and out of parentheses — cool, it feels like tennis. Okay, up to here — any more questions? I'm a bit torn — it feels to me that somewhere she is somewhat technological, it's not that she's disconnected. The question is whether she's tech-savvy, or a bit more — if we're on a scale, then further in. So no, let's take her as someone who's up to — like, we're not expecting her to implement integrations now. Okay, she doesn't implement integrations. Does she have an IT person for that? For sure — a company of 250 employees. That's an interesting point, maybe there's even an IT department, right? Maybe there are even marketing pixels there that actually need to be handled by an external company that supports her, so let's say she's in the middle — she understands the instructions, basically, and if there's something very complicated, she turns to someone who's more, like the company's CTO, say, just as an example, or alternatively some consultant who supports her, okay? Shall we dive into the problem? Which one? Cool, so we have this product, it's a SaaS product — it doesn't say so here, but we'll define it that way, so basically they give a 30-day trial, without — freemium? Yeah, without, just entering a credit card, I'll say this in editing — so just entering a credit card, and basically during that onboarding period, that prospect, that customer, is expected to start feeding data into the system, and that's how you basically see value, okay? So basically, new users like Sarah are really struggling with the complicated and long onboarding process that the product requires, which leads to high drop-off rates, especially at the initial setup stage, of the data integration stages — these are exactly the places we talked about. So hold on a sec, let's create some imaginary flow, a user arrives, signs up, leaves a credit card, says I want in, gets 30 days — right, we gave 30 days, starts up, and then basically hits some wall that says, hey, we need to start receiving data, so here are all the integrations we work with, go ahead and connect, only then can you start seeing value, after data flows in there, okay? And we understand that's where people drop off. I can give you a bit of user feedback, but go ahead and start, if you have any questions at this stage. Regarding the problem itself, because I wanted to go back a bit with you, so give me some kind of insight that I can get from the system in the future — the one we're now tuning toward, the one we're heading into. Okay, simple. So we took a simple example of... of retail. Of retail, e-commerce. So for the sake of argument, the system knows how to identify what the return rate will be for purchases that were made, basically based on historical data and so on, and based on that they know how to deliver insights that would be relevant for ordering new merchandise, and so on. You know? Good example? Great. Nice. So if we also think about this product, you could say it has two main legs, basically, in the flow. One is really getting the user to feed in the data, because once the data is with you, you can really reflect what you know how to do, and the value of the data. You know? Sorry. So... so what does this integration actually look like? Like, why is it complicated? How complicated is it? Is it really... In terms of the effort, which is a KPI, is it at the level of... what does it require? So let's, really... I have a screen that I reach, where, for the sake of argument, there are several integrations listed, like for example Google Analytics, like for example... I'm just making this up. Google Analytics, like for example, some upload from your warehouse, of inventory, of purchases that were made, online orders that were placed, okay? Things like that? But you basically need to connect — you take some instructions; there are ones that are simple, just embedding a script somewhere, or just embedding some key, or whatever — I understand that means something to you — or on the other hand, some instructions that you can take, if the person at your company is more technical, to basically start viewing this data, okay? So you understand, basically, that we're dependent here on all kinds of factors that aren't that same user, because each integration requires a different connection, or a different setup, or different tools, or... I can say that they didn't do this randomly, but probably, actually, you know, there's some cataloging of them — basically taking what's relevant, adapting it as a script, putting them together, say all the marketing ones with pixels, put together, you can basically bundle them and handle them in one place; after that there's another dropdown, just as an example, that contains the others, something like that, okay? So we have here some kind of drop-off, and we can also understand why, right? That it's relatively complicated, and you need a technical person to do it. Yes, and again, the user leaves the screen, has to come back, right? Like, it's some process that happens, that you have to manage. That also takes time. Yes. I can tell you — look, I can tell you this, some more info I can give you — this feels like a test for the product person, but we'll analyze it together. First thing, we see 40% of users dropping off at this stage, which is a ton, 30% leave the process, basically without doing, without actually setting up their dashboard, which means the 30 days we defined passed, and they didn't do it — they just did, left for good. And overall we see — well, this isn't interesting, basically overall we see a very high churn rate there. A shame to get into the numbers. What we do see, just for the point, is that of those who did complete this thing and did see the value, 50% stay on the platform, which is very nice numbers. That's nice. So we have a really strong interest in fixing this thing. So I'm already thinking, well, I won't jump to solutions right away, since we're here, there are two pathways, meaning — either I won't jump, I won't jump, I tell the user, listen, you need a professional here, ideally an IT person should do this, or I somehow manage to simplify it, to a level where the most basic person can go in and do it, the connection. Meaning, give me — like, say, connecting to Gmail, say I want to connect to Zoom through Google. So all I need to do is just authorize that connection. Basically, lean on technology, and simplify the process. Yeah, connect me to this interface and that, I'll already know what to do, I'm already interfacing with it. That's an excellent idea; the question is what it costs, and whether it's feasible. Let's ask another question that GPT helped us with — basically, what are the reasons that today, I really asked for this, what are the reasons that users drop off today, because of them, if we look at the percentages, let's see how good what it returns is. So complex data integration, which is what you said, that I now have to go to some Dave who's technical, and he'll help me — so that's around 45%, I see that's the significant thing. It's because it's complicated. Yes, we should remember — it didn't write this here, but it's interesting — how much value each integration provides, if it's now an integration where, say, 45% of all the integrations are very technical, but if they don't provide much initial value that they can start working with, maybe we don't have to start there. Meaning, the system maybe optionally can start with the things it knows it'll get more value from. Like, yeah, maybe we can... meaning that the prioritization of the optional connections is whatever provides the most value. I think we need to think in two legs, basically always, in RY. Either the thing is very easy, and we can use it to show value, and there's a fast integration; or alternatively, how much value, how much data it provides, how much value the data it brings has. Okay? So how easy, versus how much it advances things. Okay? So maybe that's how we'll think about things. So we said, 45% is this significant thing, of technical knowledge, whatever. The second thing — oh, sorry, this is reasons for why to leave, it's... I confused you a bit. He confused himself a bit. Okay, so let's make things up. Okay? So 45% leave because... because it's a hassle. Yeah. The second thing, let's say, is... 10% leave — I can write down for you — because the data they see doesn't give them enough value. They did do some integration, but it didn't bring enough value. Which is actually a low percentage. Yeah. But, okay, let's try to reduce that too. Right. So, we have on one hand the incentive issue, that there's not a strong enough motivation to stay. You go in and there's nothing there, it's empty. So first of all, we can give dummy information. Okay. Data that isn't real. And we can show insights that aren't real, but I look and I see that if I had data here, I'd get a lot of value from it. Let me take this, I think. Okay. If we're a big enough company, as described here — and it wasn't described here how many customers we have, but let's make up that we have, I don't know, 100 customers — and if we're strong enough in a certain industry, maybe we can even bring some benchmark, some quantitative info from that industry, and say, hey, you're from the retail industry, until you upload your data, this is basically the average data in your industry. The industry average. So yeah, maybe, if we can bring that, it's probably preferable to dummy data, but it's a good idea. But that's kind of sugar, right? What else can we do to basically boost the market incentive, basically his engagement, to get him to take this thing and run with it? Besides showing what I'll get from it. I think showing what I'll get, the "after," is always smart. I also think ease of use — when I see that it's easy for me to look at this, and understand it, and use it, and where to click, and what to do, and then always onboarding or the walkthrough, that's interactive, that's good. That's good, because you can act with it along the way. That's actually clickable. No, look, here we have this, and here there's this, and here there's this. Okay? No. Click here, move here, do here, type here. Really like that, to start, to get the wheels turning, to get the wheels turning, in some certain direction. Maybe also during onboarding, I can ask very high-level questions about the company, and that way maybe start to build some kind of foundation for something, right? To me that's obvious. Okay, mm-hmm. So what would you ask, for example, in a case like this? About the company? What do I want to know? Well, let's say, revenue turnover. Yeah, how many employees. Maybe even, the first upload could be some dumb CSV. What field they're in. Right. Right, really make it easy. That was something I wanted to get to — uploading some Excel. So I think that's a good point, I think what we can also do is look — we really talked about, you can catalog the problems, so start maybe tailoring them, maybe prioritizing the integrations that we know are easier, with the most value, maybe that too. Yeah. I know that onboardings, when they're deep, meaningful, versus short onboardings, do make people feel more engaged, more like you care about me — you asked me questions, you wanted to know, you're going to use this information I gave you, you can even write to me how you're going to use this information I gave — it's not just that I'm throwing data at you, and... I think you're totally right, but that's certainly only up to a certain limit — up to a certain limit, not too heavy. Because everyone has things they need to do, so that's true. What do you think, maybe, if we combined, just, it's not scalable, but if we combined into this some kind of support person — just support, I'm saying, but some account manager who climbs on with you, and basically helps you understand what's wanted from you, what needs to be done. Do it for me. But if that's not possible, but I have to make an effort. But if that's not possible, we're talking about a retail company, maybe there's no simple integration other than an IT person to connect you to the warehouse inventory, I don't know. Amen, amen. But, as much as you can do for me, you know, one of the big, well-known rules in UX — Don't make me think. Yeah, so I actually suggest some person — or just a bot, but a person — who can help bridge between you and the technical person, like telling you exactly what you need to ask them to do, even getting on the line, like already while the user is in the system, it might be able to work. That's interesting; the question is how much it costs. I don't know, okay, I'm not in pricing, I'm in feelings and emotions, and if you tell me, you can't do it on your own, call a technical person — wait, why do I need to call a technical person? Maybe I can do it myself. I think — but, okay, so it's a good idea. I'll route it to the technical person, if I'm not looking. Like how Uber and those do it, you always have two buttons, right? It probably comes from there, I never thought about it. Can I do it myself? No, I need a technical person. Maybe on Facebook, in Facebook Ads there's that. And then when I click on it, yeah, I agree with you. So you click it, it seems to me that helping you actually could be cool. Even if you chose to do it yourself, I'd still leave there the... button A. Some kind of lifeline. And... what other questions would we ask ourselves, basically? For example... what do we do during the loader? What do you mean by loader? Like, okay, I started an integration, it takes forever. I know in a lot of cases they put games, or... it depends how serious you make the system, but something to keep you there. Or that tells you, listen, all good, come back in fifteen minutes, some estimation that gives you a sense of taking control. That's it — I don't really know if that's a problem, like I don't know how long an integration takes, but yeah, if it takes... if they drop off in the middle, it might be that, I don't know why they drop off. I... let's... maybe we hurt their feelings, maybe they're snowflakes, I don't know. We don't know, but probably by the law of large numbers, no one is a snowflake, I don't know, but another thing we should do, which is a good idea — let me grab it a sec, and again I'll reinforce, I think it's good — if there's a situation like that, where a user is waiting, and we don't even know how to estimate how long it'll take, just like, say I send out a query or a prompt to GPT, or to AI, and we don't know — that's interesting, the games are very much, but another thing we should maybe do, too, is ask more questions, right? He's waiting there anyway, we can help him build the data better, give him some value in the meantime, I think. Yeah. Maybe also in a gamified way. Right. Is that something you've had a chance to help with — these aren't things you do in cyber so much, right? So much. It really depends what type of user we have, I'm allowing myself with Sarah. But yes, giving some estimation. Waze did this too. Right. They knew they couldn't solve the traffic jam, but they could tell you, roughly how long you'll be in the jam, and that in itself is a bald-faced lie. I'm joking. But in Waze it's like that too. It gives you back a sense of control. Right, right. Another thing we can look at, after we talked about the onboarding — and we'd probably do a lot of testing there, and dig more into the data, and... and also the copy. Yeah. Copy is very critical. You don't always hear about it — nail it. Yeah, yeah. And yeah, and yeah, and yeah, and yeah, and yeah, and yeah, and yeah, and also the copy. Yeah. Copy is very critical. You don't always hear about it — nail the data. Right. I always think it has enormous significance. So, a sec before I jump to the problem — this second one we mentioned about the dashboard, that it doesn't provide enough value, which is probably also rooted in the data you load and manage to add to the system. Let's talk for a sec, from my side, about KPIs, and then we'll just see how it connects — how we'd actually measure success. So, one last point on the onboarding — yes, I'd check, in terms of data, at which question most people drop off, or at which questions. That's it, so that's exactly the KPI I'm getting to, one of them. Good. I think we got it — just, the example we gave, which was really around her connection, it's at the final stage probably, of the connection, it sounds awful, does it sound, does it sound to you, does it sound to you, does it sound to you, does it sound to you, but, if we need to attach a KPI, then, probably the main KPI is how many actually complete the trial, right? With a credit card, and basically enter into paying. After that, we'd probably want, in our case, to focus on... where they are in the world, and how we reached them. Okay. At what time of day they did it? That, but that's more around — what time did they do it? What time did they do it? That, but that's more around, you know, probably data... interesting, interesting though. Very, but that's basically different dimensions — different flavors, basically of the same KPI. We'll look, like, through the info that came from marketing, just as an example — where it's best to bring users from, and when it works best, but what I wanted to say is, the KPI we'll look at, in this case, is probably the completion of onboarding, right? And then, we can maybe set a more secondary KPI, around the most problematic integration, like you said, or something in that vein, and then we'll be able, all the time basically, to try to improve it. Also check whether they allow Skip, on certain questions. We're adding that, so that's our interface. What do you think about Skip, in general? Great. Yeah, in every situation? When there are mandatory questions, you have to answer them; when there are non-mandatory questions, then yes, then allow the Skip. Sometimes I, you know, as a UX designer, I want to check out interfaces, just to see, how did they do the filter, what does their table look like, how, like, just to see the dropdown, how it's colored. So I go into the system, and then they start asking me all kinds of questions, and I say, Skip, Skip, Skip, Skip, Skip, just to get into the system, to see how the components behave, how they look. Okay, you're not the typical user. I'm not the typical user. And tell me, what kinds of things do you run into — we already opened this up, really, besides Skip, and basic things, that bump up your... in interfaces, where they really don't think about the user. Like Next and Back, maybe, or, even though it's a design company, you can still page through. I'll give you a random example, no, no, no, no, no, I don't know how you can page through. I'll give you an example — not always. Just recently, I'm consulting for an accessibility company, so I know, you won't always be able to, like some huge scanner, on multiples, how do you support that. Are we talking about the onboarding, or in general? Just for example. One of the examples — I can give you two examples. First example, from Zoom. They sent me an email saying, listen, you've exceeded the amount, of the minutes, gigabytes, you've exceeded the quota you can use, and I expected to go to where the link in the email sends me, and reach a page that says, here's your overage, and it's some message in red, with a big number, and as I delete it, it shrinks and then disappears, turns green or something. So I arrive at a page that has nothing in it — no message, no option to filter the table, no option to sort it, meaning, even if I want to just order it from the heaviest recording to the lightest, I don't have that. I don't even know how many recordings are on the page before I clicked Next — it just says, you have a total of 52, shocking, and you know, like, the filtering they allow is by name, and who names recordings, everything is "recording," come on, Rachamim, drop it. I went in one by one, deleted what wasn't needed, it was a nightmare, and you know, I tried to generate some optimism or hope for myself, like, okay, how many do I have left, and you have to calculate the amount of server space it takes, start connecting this, and this, and this, and this, where do I stand. You know what, I'm laughing, this is exactly the kind of thing we see with clients, right, when you come to look at a product for the first time, or you see some dark corner in the product, and you see it like that, and just, you run into these things. It's really maddening, but, it's ultimately a matter, I think, of trade-off, in the end. Like, you can't, as a company, focus on every aspect of the product. You probably put your attention in certain places, and then it grows and you neglect this. And then people like you come along. They also have no incentive — but they want me to use more space so they can charge me more money. I thought about this while you were saying it, and I'm not sure. Okay. I mean, I'm not sure a company this size does all those tricks. It could be — Amazon, for example, does all those tricks. Dark UX. Yeah, exactly, we talked about that. But actually, in the aspect of of Zoom, just, I don't know how their pricing works in a way that's different from... I don't really know if... it sounds more like a genuine dark corner. They lean toward good, I can vouch for them as a company. And another example, it's from... I have a Google Pixel, that's my phone. When you get an alert, say, in the morning, of an alarm clock, or anything else you set, then right now on mine the option to stop it is only by sliding, or saying stop. You know, and then I'm there for an hour in front of the phone, stop, stop. Working on my English accent, stop. Stop. And then the neighbors think my boyfriend is abusing me. Stop. Or I do a slide. Now, I just woke up in the morning, I'm not into slides yet, okay? And there's a phase where they turned it into two big buttons. So what? And it's much easier to press a big button than to do a slide. A slide forces you to wake up. Of course. And is it effective? Waking up? Hang on. Well. And then they swapped those wonderful buttons back to a slide. And I went into settings and flipped the settings and tried to find like where... how do I get it back to the previous state? It's a component they have, and it doesn't bother them to keep it. See what that means. What does it mean? Either they ran a test and you don't represent the majority. Which is a bummer, maybe. Right, but you keep this component. It doesn't cost you anything. It's in the design system. But you're saying... and there are users for whom it's... you're saying, in the Zoom example, that maybe it's not worthwhile for the business. Ahh. Maybe it's... no, they have, they keep both, I know. No, but in this case, maybe they identified that people don't wake up, get lazy, I don't know. The other thing that could be is that people complain. It wasn't good. Something happened there, right? Listen, afterwards I really saw on the phone of my... I'm sure you actually researched this, what's-it. I saw on my boyfriend's phone, that he does have buttons, I don't know, I flipped the settings, I couldn't manage it. On my phone, and we have the same Pixel. Whoa, that's fascinating. Is that fascinating? Is that fascinating? Is there something wrong here? Just, no, because on iOS this thing never existed, so it's actually maybe good to be a bit behind in this aspect. Come on, I'm bringing us back a sec from the parentheses, from the meat of nothing, we're coming back. Really, we had a different dimension, I don't remember anymore where we stopped, but basically we kind of wrapped up the onboarding stage, we talked about KPIs, we talked about how we'd measure it, we didn't talk at all about the technical aspect and the implementation of this thing and how much it costs, because, I don't know, is that a conversation you usually have, that you manage, or is it more in the direction of you do a session, think about things, the product person goes off, comes back with the pricing, and then you start — how does it actually usually work? I dealt with the KPI because that's the furthest I got, and I know that integrations are very branched, meaning, for example, at one of the cyber companies I worked at, yes, there were a lot of fields you had to fill in, and it takes time. And it's really a process to achieve all that. I brought a good example, okay. Yeah, so, meaning, I plan it design-wise, if they consult with me, it's how it'll look, or, meaning, okay, what do we do now that there's a loader that takes a long time? How do we frame that for the user? Does he meanwhile open another tab, or can he use this tab? Meaning, on the functional aspect. You don't actually manage the discussion of okay, well, this thing costs too much, the solution we thought of, let's take a step back, think for a sec according to the boundaries that we can, how we basically shoot for a different solution. Just, I'm curious to know, but that's very deep thinking, to go for the biggest or broadest solution you want, and then suddenly do this shrinking. Right, you need to look at the data, meaning, to give to users, talk to them, card sorting, A/B testing, you know, whatever's needed. Usually it's the product managers I work with, who already come, look, 1, 2, 3, 4, how do we approach this, what do we do? Okay, so I ask them lots of questions, usually they'll go back to the developers to ask them what's possible, what's not possible. Is that a discussion you enjoy having with dev teams? Of course. That's it, I think. But everything fascinates me, that's not wisdom. I think not, but from the efficient angle, and also your abilities, it sounds to me like this could just carry you a lot. Every UX puzzle whatsoever, I can stew over it, the more, you know, complex it is, a few days, and it'll stay with me and I... in the shower suddenly. In the shower suddenly an idea comes to me, oh, this could look like this and that. Two buttons, a slider. Yeah. And like, yeah, and there'll be a side sheet here. Cool. No, it just made me curious, and now let's really dive into the second problem, let's say and we managed to feed in all this data, we managed to improve the experience, our KPI is skyrocketing, blah blah blah, and basically, we won't even get into the technical aspect here, which we said we wouldn't, so okay, now we have a second problem, which again, I say, is probably fed by the first problem of lacking data, but basically we said — how much did we say? A lot. A hundred? I don't know, let's do 20 percent, I don't remember what we said. 20 percent of users, or 10 percent, sorry, of users, stay because the data satisfies them. All the 90 percent. We said 10 leave because the data doesn't give it to them. So let's say 90 percent. 90 percent. Yeah, you have no room for improvement, it's easy. Okay. So basically you proposed, hey, let's put some dummy data, so even if there's not much to see, we can show you how the final result will look, right? Yeah, types of insights we gave, and what it led to, and how they used this knowledge. Like a real case study, maybe even. How would you approach this problem now? So still, only 10 percent stay. Which might be good numbers, I don't know how much it costs to bring them in, but let's say it's not good. How would you approach it? You need to redesign this whole dashboard. Understand how it's built, prioritize it. Whether it gives it a speaker, how, like. Yeah, maybe we even need to understand the problem in depth. Why is this data we collected, and we're giving you, making accessible to you, not good enough enough. Meaning, dig deep, deep into the problem and talk to people about what they expected and what they didn't get. Meaning, really into their mental model. What did you want to see? What isn't showing up? Really understand Sarah, ultimately. Understand Sarah. Sarah from Nashville. What questions would you ask her? In some interview? And what would you arrive at? A lot of times people ask what are you missing, or what don't you have, and that's Ford — that's really asking what you want, faster horses. Or say on Netflix I want to work on a certain feature, which is skipping violent scenes. So I can ask people, would you skip a violent scene in a movie? If you had such a button? Yeah, they don't know. But if I ask the question the opposite way, if I ask, in a series you watched, was there a violent part and you fast-forwarded, then I'll get an answer that's much more precise for the feature I want to develop, whether there's justification or feasibility to develop it. Yeah, of course. So with Sarah, first of all I'd ask what she gets already from other places. What did she get from other places that was good for her and gave her value, or not. Or that didn't give her value at all, and what did she do around that? Or what insights did she get in the past and used? Which data points need to be connected for her? And I think that if I progress toward a so-called solution, then really allowing a dashboard that's given with drag-and-drop, that she can assemble herself. What a joy, a world where nothing has a price. Nothing has a price, yeah. I'm joking. In an imaginary world like that. I think. Or alternatively, that AI arranges it by nature. Wow, how scary that is. We talked about this earlier. I'll add — I think you're totally right. I'll add that what's also interesting is, in the end, we probably want to guide her to take actions. That's what she basically said she wants. That's a very good point you're raising, the actions. We — I'm saying about one of the companies I work with right now — to reach a level of a digest like this of, here, just confirm, or here's what you need to do. Like at the level of AI being so advanced that forget it, you have no need for a dashboard. That's what needs to happen now — that's stunning. I don't know if it's stunning, honestly. Depends what Sarah wants. Depends on her persona. But wait, I think I'm in the end. She wants insight in order to do something, right? She doesn't want the information just to sit in it and there are a few things she wants to do — one is probably that she uses this platform because she wants to see insights in order to improve her marketing costs, like you said, who came from where, blah blah. So we need to improve the CAC. Exactly, for example from these areas. Maybe she wants to order less inventory against the aggressive marketing. Like, adapt it to the seasonality of the marketing, and so she wants to look ahead. Maybe she even wants to now take a screenshot from this thing and send it to her manager so he sees what amazing insights she's pulling out and what an amazing product she's using. There are probably all kinds of things like that — action items she wants to take, and we need to understand her really deeply. Right. And you can just derive from that, maybe, backward, basically what — you know what the things are we need to show her. What's the story we're telling her, basically. Bottom line, she wants to make the business... yeah, she wants the business to bring in more and spend less. That's it, so we basically went to the job she comes to do, and those are the things you need to pin down. Exactly, that's the research. How does the business spend less, bring in more. Also to give — also to give marketing suggestions to develop toward. Say, Labubu for teens, Labubu, kids, Labubu. Whatever trends there are now, oh, right. Yeah I haven't experienced them in the world and I'm like, I didn't know what would break out. Ba(?)... it was, yeah. No, so Baby... no, but right now I think it's all the fashion companies too. I don't know if all of them. I saw two ads from fashion companies that made an ad with Labubu, and I think also adapting the clothes to them so they feel comfortable in them oh, interesting. I'll bring us back just to our example, so I think it's interesting if I want to look at the company level. I always do this, by the way, at the company level. Or at Sarah's specific department level and what she's trying to achieve. You maybe need to understand who this product is aimed at and who pays for it, right? Right. Who needs to basically receive this value? Usually that's two different ads, and okay, I'll ask you one more question — when you come to an idea like this with Sarah or the Sarahs of the world, how would you — how do you arrive at a meeting like this? And how many iterations, just for the sake of argument? We had a first meeting, understood the incentive of the action she wants to achieve blah blah. When would you go back to her again to show her again for feedback, or not? Like, what does a process like that look like for you? And what would you come with? It depends what we managed to formulate from this data. Again, of course you need to talk with a few Sarahs, not just one Sarah. The moment you show a user some screen, he fixates on it, meaning, that's the point he'll return to, that's his framing So you don't show screens until they're fully fixed, as much as we believe they're fixed. I think I'd ask her in-depth questions, because the moment you talk to 10, 11, 12, you're already, you start getting a certain idea, because you ask them at a high level, and then you dive even deeper, into Sarah's personal needs, and then if I manage to find in Sarah's deeper needs, and in people's deeper needs, points, then I can start connecting them and assembling the screen, the layout of the screen, in a way that's much better and cool, and there's then... by the way, also in terms of the dashboard I already said, to show them more audiences and more, just, thoughts of where you can develop and where else you can take this. You're saying... and what's trendy now, I don't know, trendy spinners, how do we do spinners for an H&M company, I don't know, again, you need to understand their incentive, whether it's something interesting, but probably in the end I also want the bottom line — to give her a KPI of her own that she'll show her manager at the end of the year exactly. Like, here, we were here, now we're here, this is the percentage that changed. Exactly, again, that's probably also going to push her to take some actions, everything probably connects, so I'm curious — is there some fixed dashboard layout you use? I don't mean at the metrics level, but just, I see a lot of dashboards that, say, start from... KPIs? KPIs, or yeah, something high-level and then basically dive toward the action direction do you have something like that? You have a hierarchy, but every dashboard will always be different, and my suggestion is first to draw it on paper even before you go to Figma or reMarkable — do you know it? Huh? You know reMarkable? What's that? Wow, it changed my life, okay. It's kind of a notebook, basically. With dots? A tablet no, but it feels like a pencil, oh, okay, okay. Nice. So yeah, as I said, we'll try. We opened parentheses, so she draws it basically to understand, hold on, to prioritize what's most important for the user to see, which user he is because we have several users in the system. Am I now creating several dashboards for several different people? That's possible, okay?

Bar Rachamim: [41:12] and then to understand what's most important to him and order the items, the cards, by prioritization from most important to least important. The most important — what is it in English — is at the top left, the less important is at the bottom right. It can include graphs, it can include a table even when you do graphs you need to do graphs properly and understandable, so that at a glance I can understand what's going on here, yeah, it's a whole world, graphs by the way, and I've had my fair share — I hate — I'm still carving my path there learning how to improve, and that, well, probably really to draw enough graphs and see if enough people find them clear and to consult with a data analyst

Eyal David: [41:54] that's for sure, Bar, so we're approaching the end of the episode — how was it for you? Hey, it was fun

Bar Rachamim: [42:00] it was fun for me too — so how do people get in touch, if they want? I'm on LinkedIn should I give my number? No, no, but that would be a shame — but let's give the LinkedIn, pleasant

Eyal David: [42:13] so Bar, thank you very much — I'm also on Facebook — thank you very much, thanks for having me, bye bye