insights! #70: personalised search experiences with AI search

Joubin Rahimi:

Great to have you back for a new episode of insights! Today I get to welcome Christina Schönfeld from Algolia. Hello, Christina.

Christina Schönfeld:

Hi. Nice to be here too, Joubin.

Joubin Rahimi:

We'd already had a great pre-conversation just now, and it's about search. Anyone who knows Algolia of course knows you're linked to search. But maybe you'd like to say a few words about yourself and Algolia first.

Christina Schönfeld:

Okay, so as I said, you've already mentioned my name. Christina Schönfeld. I'm currently Director for the DACH region at Algolia, I've been with the company for almost five and a half years now. Every time I say that, I'm surprised and fascinated myself. Because that's really a long time at a scale-up. And yes, Algolia itself, you said search, so we're actually a search, a search and discovery solution platform. That already lets a lot of people in e-commerce categorise us somewhat. But I always try to bring us back to where we actually come from, or what our mission is, which is that we're really here to connect the right products and content with the right people at the right moments. And that's not just the little search box that gets neglected, it can be anywhere structured data needs to be returned.

Joubin Rahimi:

And I'd like to dive further into that topic too. I already mentioned beforehand that we once had an exciting client situation where, after go-live, we sat down with the board, and one board member said, Mr Rahimi, if I type in white sofa, I get sofa cushions in white. That can't be right. Opening bracket. They'd paid for a very basic search, because they didn't want to spend another euro. Closing bracket. And the data bears that out exactly. Sofa cushion white were the terms in there. And for them, sofa is simply upholstered suite and never sofa. The system just couldn't recognise it, but it shows that without proper search, and an adequate search, the shop simply wasn't worth anything. So that changed things, and that's why we thought: search is such an important topic even today. And you just said it's actually so much more than the search box, that we said we simply have to do an Insights episode on it, and thanks for taking the time. When you say it's so much more than the search box, what else is it from your point of view?

Christina Schönfeld:

Well, it's also the category pages, product listing pages as we call them, of course also on the homepage – basically everything that needs to rank. And when you think about it, nowadays all the small screens are getting even smaller, or staying the same size on the phone, but well, people just look at everything on their phone, they no longer have time to work their way through loads of catalogues. Then you also realise how important it is that the first few results are right. Also what you see on a category page like that. But I don't want to give a big long marketing pitch here.

Joubin Rahimi:

No, no, not at all. Yes, about the marketing pitch. But I'd quite like to pass on this understanding of search once more for those who might not yet be that far along in digital maturity, because you say the first three have to fit. Now I know "sofa white", but what actually is it? For me, sofa white might mean something different than for my wife.

Christina Schönfeld:

That's definitely the case especially in the furnishings sector, where it's often about taste and, to some extent, intent too. So where are you currently in your buying journey? But thinking back to your example and the managing director, I'm not surprised that was the reaction, because that's obviously a neglected topic. Were you able to talk to your managing director about how much revenue they actually generate through search?

Joubin Rahimi:

We pulled out the statistics. It was a bit better than before at the relaunch. But the search was actually worse than it had been. But we said: how many of these sessions go via a search, and it's simply 80% or 85%, I don't remember the exact figure. Relatively a lot always went via search, because they had a lot of products and the categories were so diverse. And when you want to be inspired, you sometimes go into categories, but not necessarily with them. But the search was of course always there, or the filter mechanisms within the category.

Christina Schönfeld:

Hm, well quite unusual. Normally the objection we hear there. Nobody on our site actually uses it, everyone's just navigating and browsing and we want to inspire, especially when it comes to furnishing. And then you pull up exactly these statistics. And yes, maybe it's around 20% of sessions with search, but actually, the revenue share is higher than the share of sessions that only run via browsing.

Joubin Rahimi:

The answer was the other way round. You need to sort the categories better.

Christina Schönfeld:

Yes, that was the answer to that. And that's also something we do. So, that's it. The circle closes.

Joubin Rahimi:

But okay, then you say okay, you're in a business too, but search is a complex topic. You said that too, depends on the situation. What exactly do you mean by that? So do you get different search results depending on the situation.

Christina Schönfeld:

The situation, well, that's where the journey should go, right? So that you can already tell from click behaviour what your intent is, where you currently are in your journey. There are lots and lots of things on our roadmap right now, but I can't say more about that yet. What we have of course been doing for a very long time, though, is that you probably see something different on the sofa category page than your wife might, who perhaps prefers more of a bohemian vibe, let's say, while you might prefer more of a leather sofa.

Joubin Rahimi:

I'll check that. We have another product, called brytes. We haven't spoken about that at all. We say it can read and understand digital body language. And that's exactly where I wanted to go, because we've found, if someone puts copy and paste of product name in, then a) wants product name, consumer wants to have product name and see product and probably also buy or see price or availability. But if I enter generic term or wardrobe, chest of drawers or whatever, then I'm in completely different customer journey. Simply based on terminology you can already see that. We always found that exciting, reacting to that, because then you can simply create other offers or you know, they want to buy, I don't disturb them, and other person I then offer advisor or chat. So system does that there. That's why we find search really something. You can read so much out of that.

Christina Schönfeld:

Yes, of course. I mean, it couldn't be clearer — you have no other touchpoint with your users on the website. Neglecting that is a real shame, because then the user probably runs back to Google and tries their luck there. Exactly. Then the data ends up with Google.

Joubin Rahimi:

Then ask a second question. Not a question as such. In this search environment, product data being good matters a lot. But it's never good enough for all these cases within search. How do you deal with that?

Christina Schönfeld:

They will also never be good enough, because if you think about how many ways we have of expressing certain things. It is already difficult enough for two people to understand each other. How is a machine ever supposed to manage that? And you just can. I keep meeting people who want to tag everything. Sure, it should help. You should have good data, yes. But you will never reach that ideal scenario where absolutely everything is covered. And that is exactly what we do. We also have a semantic search engine based on vectors. I could go much deeper into that now. Yes, essentially applying these large language models to enrich queries or enrich them with results. And then much more query-like than what you could achieve with just the keyboard. And I notice I am speaking a lot of Denglish again. But we can come back to that in a moment.

Joubin Rahimi:

Why though — compare it like this, when thinking about tagging, also intranet, but many advocates say you must hide everything, hiding documents doesn't work at all. So system must recognise what I want. That's one side. That's Yahoo, the hiding, so the Yahoo search from 20 years ago, and other side is smart search. So I think with AI you're probably nothing new either way, you've been doing data science product improvements and features for quite a while already.

Christina Schönfeld:

Yes, although you always have to be a bit careful about what exactly is AI or what's just machine learning. So I think AI has of course grown massively in the last twelve months and now we're all starting to get really smart. But there are quite a few, especially in our industry, who throw around the words AI and semantic search a lot. And back then in many sales pitches you were suddenly confronted with: hey, XY does semantic search, and then you dig deeper and look at it more closely and find, no, what they do is a few rules and natural language processing, but not natural language understanding. And those are just such nuances. A really good point, because it's often exactly with search that we compare software with each other.

Joubin Rahimi:

So when it's about replatforming an e-commerce, an e-commerce system, and the maturity level isn't that high yet, you're introducing it for the second time or the first time. Then you say search, and you need search result lists. I want to filter. So you've got a few checkpoints, a few attributes, and everyone says yes, but the way you put it, can it process the language, or do I understand the whole thing and can react to it? That's worlds apart. And that's actually my segue into how we came across you, because within our company we have a culture, we say Drive Delivery Care, and under Drive that also simply means getting things done, and our group name is also TIMETOACT GROUP, and Time to Act means it's time to do something now, and the colleagues said, we've also got pain in other situations, we'll take a look at this, and that's how you ended up on our consultants' radar. That's how we became partners, and I think that's great. I didn't hear anything from you for a very long time. And now? Now you said you've been there five years, and five years ago you weren't even in Germany yet, might sound far off. But actually it's not that far. It was 2019, probably shortly before Corona, and now you're very well known. So many of our customers know Algolia, and the market knows Algolia. How did that happen?

Christina Schönfeld:

Yes. Yes, that's right. Actually five years isn't that long at all. And it's of course lovely that particularly your developers or your consultants had us on their radar, same for our design. And that's also how we originally grew, really out of a developer community. That was certainly the start, over ten years ago. The beginning. We were founded by two French engineers, and from the very start actually with headquarters in San Francisco, because they were in the accelerator at the time and always had the thought: we need to change all these search engines, because over ten years ago mobile networks weren't as good yet. And what we actually wanted to solve was optimising search on mobile devices and everything that existed on the market at the time. So not really with an e-commerce background, but really what you build in terms of open source solutions. That wasn't optimal, was above all very, very slow. And from there they started building it completely from scratch. And that's how Algolia came about. And interestingly, Algolia then had a pretty good and very fast development. Also within Europe. But not yet in Germany. And I had already back then. French, well it's not a French company. But yes, sometimes European roots. I like telling that story too. Of course, because it makes the company truly unique, having such an international footprint. Many want to be international, but you have a US influence in there of course, but the French and European mindset, and I was already in Paris back then, working for another SaaS company, and had it a bit on the radar. And at some point our paths crossed again. And yes, it was mainly our developers at my previous company who had worked with Algolia. And they nearly had tears in their eyes when I mentioned I was talking to Algolia. So hold on, there must be a really good tech fit there, and that's how I ended up there. The only problem was I had no one else. As I said, we had, I don't know. A few hundred customers in the DACH region. Nobody knew us. We didn't really know our customers either, I have to admit, it was all a bit chaotic and had no real investment as you'd expect, like marketing or a partner network, just: get on with it! So a bit naive too, certainly. And then I set off. For that I had a bunch of competitors. That was exciting. The first twelve months, because every single time, no matter who I approached, it was 'aren't you just like them'. So exactly not.

Joubin Rahimi:

Good. But it was great. Then you can already place roughly what it was about.

Christina Schönfeld:

Yes. Yes. So really then somehow the breakthrough. That was actually and that's what sets us apart, because I – I think I said it – we're really an API and API first as a company, and we're not just an e-commerce solution, we can also retrieve content. So many media companies use us in their tech stack too. And at the start, when I began back then, we tried to cover everything, and that's actually never a good idea. We tried to do everything. And at some point I thought, hold on, who does it work particularly well for, and where do we have a business case behind it, and who do we actually need to talk to? And then it quickly became clear that it's no great science to work out that this could work in B2C commerce. But as you said, those objections – well, nobody uses it, it's pointless. Why should I spend money on that? That was actually the challenge at the start, and I was even partly involved with partners myself. I can tell that story. At some point one of our partners approached us, knew Algolia and said, well, I've got a pitch tomorrow and I need an offer. Christina, send it over. Here's the architecture I've worked out, and Algolia's in there somewhere as a small search field within the architecture. I said: right, that's nice. I know you've worked with us a bit before, but we do a lot more than that. Did you know we can also do the ranking on the category pages and return that personalised too? All men. Hold on, that changes everything. And the next day he sent me back an architecture slide where Algolia was really in there as a layer – connected right across the system, CMS connected, shop system connected, and a few other systems hanging off it, and then delivery across the various interfaces: website, app. I can even feed back into the store if you've got tablets there, and that's obviously a completely different pitch, and that's really where the – well, that awakening happened with the e-commerce companies too. That was kind of one of those moments where you say, this changes everything again, because for that partner it probably did change things. I can make the architecture so much better and the customer gets so much more value. Especially because I think up until three years ago, at a lot of companies the mindset was still, right, Covid happened – we need to catch up fast in e-commerce. Now it really matters, and damn, our shop system's outdated. We need to build a completely new platform now with our monoliths. And then people thought about what to do for the next three years. So first, rip out the whole backend. But as a project manager within a company, you want to show results as quickly as possible, and nothing's faster than working out what actually drives revenue and impact in the frontend. And that's exactly what we do – as microservices controlling the ranking in the frontend.

Joubin Rahimi:

Now we're practically at the partner stage, but that probably wasn't the case at the start – it was already a bit further along in the process. When did you have – you basically had DACH as a territory that was simply empty, or you said almost nothing was happening there. When was the point for you where you said, and why? Now things are moving forward. So now I've got a grip, or I'm getting traction, that things are really moving forward for Algolia in DACH too?

Christina Schönfeld:

Well, a lot of trial and error and door-to-door knocking.

Joubin Rahimi:

Simply hard work. Sometimes you really need hard work.

Christina Schönfeld:

Actually, there was, for me I think — I think an aha moment was really at one of these K5 events. I think it was 2019, 2018 or 2019. I went there alone, actually had no network either, and just walked around, approached everyone left and right and simply — yes — started conversations. And at some point I was standing at a stand where a CTO was standing with a bunch of other onlookers. They were all men of course, and rather technically minded, talking about it. I think it was monolith or build-yourself, so the typical build versus buy discussion, and I just listened to all of it. And at some point I simply asked the question — well, because he said, the CTO was a big proponent of 'I build everything myself'. And then I said, well, what about microservices, like a search solution, Algolia? Yes, sure, sure, sure. Microservices are fun, make sense. Yes, we'd look into that. We want to evaluate that more closely. And that was a moment for me, okay? Microservices weren't such a huge term back then, at least not headless and composable either. Now everyone talks about it and it's all old news. But back then it was a bit of a shift in thinking. I need to understand this better, because if someone who's such a staunch proponent is open to it... Because we're not here to — we're not an out-of-the-box solution in that sense. Nowadays we bring everything out of the box and you can leave it as is. But actually, what our philosophy is — we want to be a better way to build things and build faster, and we really speak to the builders. I've never found a German translation for that, unfortunately, because 'Bauern' just doesn't work. Yes, for the builders within a company, and they might sit in development. But they could just as easily sit in product or in marketing. They're the ones who want to change something about the user experience, whether that's on a simple website, in e-commerce, or perhaps even within a company. We also have a lot, a lot of knowledge bases and things like that.

Joubin Rahimi:

Mega exciting, because what you're saying, this buy versus build we also get every time. But if you're in such composable architecture, it's sort of like build. Even if they buy something. So I always wonder if that's maybe play instinct, that's usually men in that context, that they can click Lego bricks together, is bit different. But they can build something, like sandbox. And that's why composable is totally great. Because they keep building, even if function blocks already finished. You know what I mean? So that's thesis I once had at fireside chat. Well, always this buy versus build and then you do composable. Find that funny, you've basically told exact same story. You buy stuff there too, wouldn't it just have to be called completely differently, not buy versus build. That's exactly buy to build. Yeah and you always then took sandbox and Lego. You actually want play Lego and not Playmobil. Or. Lego Technic now?

Christina Schönfeld:

Yeah yeah. I was a Lego person too. But the difference is, I think this whole buy versus build discussion is simply outdated. Back then these were really the two extremes and I think a company, well, you shouldn't start by hiring a search engineering team. Those are very, very rare resources on the market, which are also very, very expensive. And yes, and if you hire them so they can. I don't know, figure out typo tolerance or I don't know. Synonym handling in 15 different languages. So. You think these are trivial topics, but they're not easy tasks to solve. Really. And of course an engineer's heart beats for that, it's just in their DNA, but actually engineering has much, much more potential to get things out of a search solution from an API. We are, as I said, a building block you can build faster with.

Joubin Rahimi:

That is, I think, also one of the totally important quintessences. And with certain things, I think, it's totally valuable to think about whether one can do something better. But it's not always the same for everyone. With search. There's now such a high level of maturity across all products out there and SaaS services. How much better do you have to be to get the extra investment out of it? Well, that's it. I'm not building the thing any cheaper than if I bought it from you. That's already not the case, so for the extra money you spend you have to build a better solution. And there I doubt almost all retailers, because they're not tech companies. Quite bluntly. So okay, you can do it. But are you a tech company? Why do you want to take on a core tech problem? Are you much better at it? No, we're saving money. Wait, you're doing something you can't do and think that's saved you money? No. Probably not. Risk is high.

Christina Schönfeld:

There are many retailers. They've got insanely great tech teams. The really big ones, I'd say. They build things up too. Yeah, as I said, it's only the question. Do I reach the goal faster with a solution that already gets me one step further? And can I then build on that? And that of course depends on the flexibility of a solution. So one should certainly not go with a complete out-of-the-box solution that pushes you into a black box. But let me think again how it came about. You asked me right at the start: how did it happen back then? Yes, and how did I know. Because. That was the moment when I thought, hm, there's something else out there that understands me better. But the story wasn't over yet — I'd actually built my pipeline very laboriously, and at the start of 2020 it actually looked like: right, now it's about to take off. And then Covid came round the corner and wrecked everything again. Wrecked it, pipeline wrecked. That's how you talk as a salesperson. And yes, even contracts we'd already sent out, ready for signature. Of course you can always go back. I think it happened to a great many people, because if we remember, at the start everything was dead. And I think in that moment I thought, this can't be it now, a pandemic. And that was the moment when I considered, as I said, considered what I should focus on now. So no more media companies, that's for the company next door, but really just B2C only, and then within that I built myself a matrix at the time of what actually distinguishes each retail company a little. Because I just mentioned, there are lots of tech, well, retailers. There are some retailers with very good tech teams. Especially the pure online players. They grew out of that, and that's a completely different conversation and a different focus than when you talk to an omnichannel retailer, where before Covid e-commerce made up 1% of revenue. And you're also in a different competitive dynamic with a pure online player. You really have the build versus buy discussion, because open search, open source solutions were the answer, whereas with the big retailers the awareness maybe isn't there yet, but somehow they've already got one of the other competitors in use, or everyone jumps on it and everyone wants to talk to them. And then I just focused on that and also spoke with a huge number of partners. And they tell you what Algolia does. And that's where it comes full circle. Maybe back to what you said at the start, that it was your consultants or our consultants and developers who had us on their radar, because at some point it apparently really did spread like wildfire. That was all I wanted back then. How do I pull off something guerrilla-style without marketing resources? And then the breakthrough came very quickly indeed.

Joubin Rahimi:

And now it's your turn How many staff do you have in sales?

Christina Schönfeld:

We're five now and want to grow to six. We've built everything around that too. Then super fast. So marketing or pre-sales. So when I say five in sales, those are salespeople closing deals. And we also have business developers around that, sales developers, marketing, partner managers, engineers, customer success managers. I think we're 15, 20 people in Germany in region.

Joubin Rahimi:

Yes, great. And that then first for unit generating revenue for you in DACH. Built up within probably two, three years then. If then rather towards end of Covid there was a push upward. Yes, incredible. Time passes quickly, when someone says let me jump back briefly to start. I have a search. Not always so happy? What would you recommend to person besides of course coming to you at Algolia? Yes, so one can make a mature decision. I'm totally in favour of us giving our consumers input so they can make mature decisions. If you were in situation of choosing a search for an e-commerce system, what points would you look out for? Of course always difficult, what kind of e-commerce system, what do you sell, there are many different topics. But I suspect, if you name three things, they'd be that valuable.

Christina Schönfeld:

Three things, actually.

Joubin Rahimi:

Yes, or 5, 7, 33.

Christina Schönfeld:

Okay, let's try categorising this now. Actually, it's clear. You've got the typical product with all its features and benefits and all the stories. Yes, let's leave that as it is for now. And as I hinted earlier, you can go very, very, very deep into detail, and it's important to question what stands behind each term. Because personalisation actually means something different for everyone. That's just stated for now. I think what's really, really important, of course, are the people you work with. Who sits in that company? What are their goals too? Is it simply a culture you want to work with, because at that moment - and this is nostalgic again but back then, when we won our first clients and really big clients in Germany too, those were actually people who ultimately put their careers on the line because they went with us and not with XY, who'd already been very present on the market for 20 years. And you have to keep that in mind, also as a salesperson. That's really important. And the other bracket alongside product and people, so the core pillars, you should pay attention to the things that tend to happen more in the background. So what's actually the architecture things are built on? How do we manage to achieve high availability, for example? What about the documentation? What happens if my search really goes down, or the system goes down? Or can I help myself as a team? Or do I always have to call someone and am I actually dependent on someone? And yes, how easily can I actually get out of it again too? That's just as important.

Joubin Rahimi:

An exciting question.

Christina Schönfeld:

Yes. Because. I mean, that's the advantage of microservices. We can also be built back out again. We don't want that, but that's exactly why. We have skin in the game, so to speak, to genuinely have to prove ourselves.

Joubin Rahimi:

Yes, cool. I find the last point mega important too. There are some that are so intertwined that you don't even want to touch them. And if you say okay, microservices, MACH architecture or composable commerce and I can then swap things out at the endpoints, that's naturally much lower-risk. And of course the point. Yes, the first ones put their job on the line. I find that an interesting perspective too, one you perhaps also need to have as a salesperson sometimes.

Christina Schönfeld:

There you can support in a completely different way. It's not the function, but availability, that it works. Trust, which still had to be built up much more. Yes.

Joubin Rahimi:

Great. Yes, Christina, thank you very much for the insight into how you went from 0 to almost 20 employees. From zero to 20. Wishing you every success going from 20 to 100. And it was great fun hearing the insights and anecdotes about search. Thank you.

Christina Schönfeld:

Thanks to you too. I enjoyed it.

Joubin Rahimi:

If you have questions or comments, feel free to leave them in the comments or write to us, we'll gladly pass them on to Christina. Or you can certainly contact her on LinkedIn too, and we wish you continued success. Bye.

Christina Schönfeld:

Bye.