A review is not the end of marketing – it’s where the next conversion can begin.

# Bookerz on the Success of 123watches — #trstd connect Amsterdam

Marijne van Luijk: Thank you. Thanks. I was actually quite curious to see who I had in front of me, because I don't know anyone here yet. Are you all from marketing agencies? Could everyone from a marketing agency raise their hand?

I heard around 60%. Okay, that's actually a little less than I expected. I based the presentation partly on that, so be prepared.

I was also curious how many of you have customers with a lot of data that you feel you're still not using enough. Yes? Great. Then at least I'm not standing here for nothing.

That actually brings me straight to the point. We see that companies have been able to collect enormous amounts of data, but in many cases still don't do enough with it because some systems simply don't communicate well enough with one another.

Review data is part of that as well. So I want to talk about that, about the solutions we offer to get more value out of it, and briefly about who we are and what we do.

What if a review isn't the end point of marketing, but actually the beginning? That's what I want to talk about today.

Fortunately, this presentation is quite short, so we may even make up some time.

First, a very brief introduction to Bookerz: who are we and what do we do?

Then I'll take you through three steps. Reviews: how do we collect them intelligently? How do we use them intelligently? And how do we get the most out of them?

After that, I'll show you a customer case where we've put this into practice and already achieved some very good results.

So, Bookerz. As I was introduced already, I'm Marijne. I've only been at Bookerz for about three months. I don't come from the data world at all; I came from the wine industry. But I've been able to learn and pick up a lot very quickly, which has been great.

Bookerz itself has been around for 19 years. We started in 2007 as a boutique agency. That means we work very personally and can deliver a lot of custom solutions.

We have backend developers in-house who can basically build anything themselves. One of them is here with me today, so if you have difficult technical questions, there should be a good answer available.

Our founder and director has also been with the company from the beginning, and we're based in Ravenstein.

What do we do? We've been building our data platform for 19 years.

I thought it was interesting to see Factor Blue speaking just before me, because I was already thinking: this overlaps somewhat with what we do.

We build our own data management platform to which we can connect all kinds of systems and make them communicate with each other. I'll explain more about that in a moment.

And we build custom solutions where the standard solution stops.

This is probably a recognisable situation. Everyone knows where their data comes from and what they would like to do with it. The question is often: how are we actually going to make that happen?

In an ideal world, all systems communicate with one another. Here are some examples of systems data can come from and the places where you want to use it again.

Our solution is the data management platform we've been developing for years. I always describe it as a kind of central roundabout in the middle to which we can connect everything, because we have all the backend development expertise in-house.

You can connect the webshop, a CDP, CRM, review system, loyalty platform, POS system, an ESP — whatever you need.

Ideally, you want all that data available in almost real time in one place so you can build segments and serve the right people in the right way.

Once you can connect everything, the possibilities become enormous.

That's what we've been working on for years. We make sure data is deduplicated. We make sure it's available almost in real time — perhaps with a delay of a second or so.

And every year we take another step in the level of customisation we can provide.

Of course, fully custom development would become very expensive. That's why we work with an approach of roughly 80% standard and 20% custom.

The same applies to marketing agencies. You may have many different customers that don't all do exactly the same thing, but if you look closely, you can often reuse about 80% of the integrations and solutions.

The remaining 20% is the customer-specific question. That's where we add the custom work.

We handle that through our data management platform.

So how do we collect reviews?

I'm genuinely curious, because everyone here probably collects reviews in some way. Otherwise, I don't think we'd all be here. But the way companies collect them is often fairly similar.

So how can we make that smarter?

The first thing is to make sure a review request is relevant.

Imagine I buy five pairs of running shoes. Is it really useful to send me five separate requests for reviews?

Maybe two of those pairs are being removed from the assortment tomorrow. Why would you ask me to review those, rather than prioritising a new product that still has no reviews at all?

You can filter that kind of thing and prioritise what matters.

For us, that's relatively straightforward because we can create segments around it.

So the goal is: ask the right customer, at the right moment, through the right channel.

That's really one of our core principles: reach the right person, through the right channel, at the right time.

Relevance matters. Without relevance, the data doesn't help you very much.

What else can you do with Bookerz?

One example is an integration with Spotler Activate, which many of you may know. Website behaviour is extremely interesting data to work with, because it can help determine which review request is most relevant to send.

You can also connect review collection to a loyalty system.

You don't want to force people to write a review when they don't feel like it, because then there's a good chance you'll mostly get negative reviews. People who want to complain are usually very motivated to leave a review.

So how do you make sure positive experiences are also represented?

Imagine you have a loyalty programme with points. Customers collect points and can later exchange them for a discount.

You can also use that for reviews. Leave a review, earn a point, and later receive a discount. That's just one simple example.

I'll move through this quickly, because the next point is even more important.

Everyone already has so much data, but if you don't use that data in a relevant way, you're missing a lot of steps and a lot of opportunities to increase conversion.

So how do you use review data properly?

You actually have gold in your hands.

A review is a customer evaluating something. As a customer myself, if I want to buy a pair of running shoes, I would much rather know what other people who bought those shoes think than hear what the manufacturer says about them.

The manufacturer wants to sell the product. Other customers have actually used it.

That's why reviews are so valuable.

The most obvious thing is to show those reviews on the website. That's probably what all of you are already doing: using social proof to increase conversion.

But you can go further.

Think about an abandoned cart email. If I have running shoes sitting in my basket but haven't completed the purchase, I may receive a reminder saying, "Those running shoes are still in your basket."

Now imagine that message also contains several relevant reviews: "The comfort is excellent," for example.

That immediately becomes much more persuasive.

You can do that on the website, but also in email. The channel you use is up to you. Almost anything is possible.

Pop-ups and forms are another option.

Then there's personalisation and marketing.

Review data enriches the customer data you already have.

Take the running-shoe example again. Suppose someone writes, "The comfort is amazing."

The interesting part is the word comfort. That tells you something about what that customer values.

You probably have many other products on your website positioned around comfort. So how can you use that insight to show that customer more products from a comfort-oriented range?

A negative review can also help improve your website.

Suppose many customers complain that a certain running shoe fits much smaller than expected.

You can use that information automatically. For example, you can change the sizing guidance on the site or flag that customers may want to choose a larger size.

Those are all examples of how review data can improve the customer experience.

And then we get to generative AI.

This is an area we find particularly interesting, and we've connected it to a customer case.

The idea is to use the content of your reviews as input for something similar to a chatbot.

A customer can ask a question on your website.

As I said earlier, I'm personally much more interested in what other customers who bought the product think than in the answer from the company trying to sell it.

And the question doesn't even have to be about one specific product.

It could be something broader like: "What are the best running shoes if I have knee problems?" or "What kind of shoe is best for my situation?"

The system can then generate an answer based on the reviews that customers have actually written.

We implemented this in a customer case, and it contributed to a substantial increase in revenue over several years.

One of the things we built there was this generative-AI layer based on review content.

The example on the slide is difficult to see from here, so we'll make sure the presentation can be shared afterwards.

The idea is that a customer asks a question, and the answer is generated from the body of reviews collected on the site.

The customer in this example is a Dutch webshop selling lighting and sound equipment.

Someone might ask a very specific question about a smoke machine for a very large stage: what kind of machine do I need, what dimensions are relevant, how much output is necessary, and so on.

The system then looks for a targeted answer based on the reviews and can say, in effect: this is the product we recommend, because customers say the mist distribution works well for this kind of space.

Again, the point is that your reviews already contain a lot of valuable information.

You don't need to invent the answer.

And frankly, you shouldn't want to invent it. You want something closer to the truth, and that truth often comes from customers.

That's why review data can be so powerful.

There's another interesting point: if you started with Trusted Shops or Bookerz today — or both — you probably wouldn't be starting with empty hands.

We can already see which orders have been placed.

Maybe you've had a webshop for three years and accumulated a lot of orders, but you've never asked those customers for a review.

Those historic orders can often still be used.

You can send customers a message such as: "You bought this from us some time ago. We're doing some research — what did you think of it?"

That gives you another opportunity to grow the review base.

Build the database, because there is an enormous amount you can do with it.

For marketing agencies, the division of roles is actually quite straightforward.

You define the strategy and what you want to achieve.

The technical people are at Bookerz, and we build it.

Almost anything is possible.

We actually like it when a question is too difficult for someone else. That's a challenge we enjoy.

So ask yourself: which customer still has an unresolved question, or data they aren't doing anything with yet?

Today we're talking mainly about review data, but there are many other sources: order data, website behaviour, and so on.

Send those questions our way and we'll gladly look at them.

Thank you.

I'd be happy to take questions, and we have technical support here for the difficult ones.

Audience member: I have a question about the review data you're using to give customers advice on the webshop.

How do you deal with information in reviews that is factually incorrect?

A customer can write something in a review that simply isn't true.

You see that quite often. If I go on holiday, for example, everyone may say the food is terrible, while I actually think it's perfectly fine.

So how do you deal with that kind of thing in the data?

Bookerz: That's a good question.

My first response would actually be that, in practice, this problem doesn't really exist in the way people often think it does.

And I'll explain why.

It's basically the same issue as asking: why can AI produce a good answer when the internet also contains an enormous amount of nonsense?

It's about volume.

If you have enough reviews, the outliers and nonsense tend to be outweighed by the broader pattern, and the underlying truth starts to emerge.

We've tested this ourselves quite extensively.

At the moment, this particular customer case isn't live anymore because the customer still finds the technology somewhat uncomfortable. We see that quite often with these kinds of systems.

Technically, it works. The tests look good. But AI still feels risky to some organisations.

In this case, we had around 33,000 reviews available as source material.

In practice, that volume provided enough relevance.

If everyone says a certain smoke machine is terrible, then generative AI will of course communicate that the smoke machine is terrible.

Is that a problem?

I don't think it's any more problematic than having one poor review among a collection of good reviews. It still gives a realistic picture.

What we also do is not only generate an answer through a large language model, but also show which reviews the answer is based on.

That gives you an additional layer of social proof on top of the generated answer.

We do that precisely because we know people ask: where did this text come from?

So hopefully that answers your question.

In short: if everyone says something is bad, then perhaps it really is bad.

Audience member: Are there particular industries where you see much larger revenue increases from the solutions you provide, or is it fairly evenly distributed across all the customers and industries you work with?

Bookerz: That's a more difficult question.

There are certain niches where we naturally have a lot of expertise.

Online sports retail is one example. We have many customers there, a lot of data and a strong understanding of what works and what doesn't.

Naturally, that makes it easier to generate results.

But we also work for hospitals, for example.

I think social proof works almost everywhere, and the value of reviews applies across industries.

With almost every customer I visit, I say the same thing: you already have so much data, and I think the next step is review data.

It's essentially content generated by your customers.

In the past, everyone understood that user-generated content was valuable.

Oddly enough, I think companies still do surprisingly little with review data today.

One place where I immediately see clear commercial value — and you can confirm it very quickly with A/B testing — is the abandoned-cart email.

What we do is show highly relevant reviews next to the products sitting in someone's abandoned cart.

What counts as highly relevant?

First, obviously, a review of the same product.

But if we're sending you that email, we probably know more about you as well.

Maybe we know you live in Nijmegen, and we happen to have a positive review of that product from someone else in Nijmegen.

Then that review may be more relevant to show because there's a better chance the content will resonate with you.

For me, that's an obvious next step.

In the past, it became standard practice to have an abandoned-cart email.

I think it's becoming equally obvious that review data should be added to that email.

Audience member: You already have 19 years of experience, but AI must also be changing a lot for you.

What are the things you're now advising clients to do differently with data management compared with one or two years ago?

Bookerz: That's a very good question — and a difficult one.

What really changes for us?

Many of our customers are slightly cautious. I understand that. We're cautious as well, in the sense that whatever we implement needs to work and needs to be reliable.

At the moment, I don't think that much changes except in places where we're dealing with textual data.

Are we suddenly using AI everywhere? No.

I'd rather say we're using large language models much more.

We were already translating reviews before.

Now we're increasingly extracting conclusions from review text.

We're combining product descriptions and review text.

Those are things we weren't doing in the same way before.

So one thing I advise now is: don't just collect the structured data you've always collected — order data and so on — but also make sure you collect and retain the textual data.

That textual data is extremely relevant to have inside a data management platform.

The more sources you have, the more directions you can use the data in later.

I think that's the biggest change right now.

For the rest, I think we're still waiting for AI to become genuinely reliable in more operational use cases.

I had actually expected more progress over the past year.

And I don't know what the coming year will bring. You also hear people in the media saying that we need to slow down.

If we do that, then perhaps we simply stay where we are.

Audience member: Do you already provide an MCP for marketing agencies to talk directly to the data?

Bookerz: No, not yet.

What we do already is make sure that, where it's relevant, structured data such as JSON-LD can be added to a page.

That's typically something you can generate from data you already have.

One interesting example is that you can generate FAQ data fairly easily from review data.

There is a specific JSON-LD schema for FAQ content.

You can use that to make the content more machine-readable and provide AI agents with the information you specifically want them to pick up.

That may actually be an even better answer to the previous question.

What matters now is applying GEO in places where we used to think primarily in terms of SEO.

Where does that data come from?

Often, you already have it in your CMS or your product data.

But you now need to expose it in a way that these systems can actually index and use.

Michel: Hi, I'm Michel.

Two things caught my attention. One was the negative reviews, and the other was the fear around AI.

On the first point, how much have you designed the system to turn a negative review into something useful for someone else?

Take the shoe example. I have fairly narrow feet. If someone writes, "This shoe is far too narrow," that's actually fantastic information for me. That's exactly the shoe I may want.

Do you make concrete choices around that?

And the other question is more about AI adoption.

A lot of people are afraid of it.

Do you have some kind of workflow in your platform where a human still approves things — for example, the FAQs that have been generated?

Bookerz: Yes, I agree with that, and those are several questions at once.

Let me start with the approval process.

When we're publishing content automatically on behalf of a company, for example through a content management system, then yes: with all the customers where we do that today, there is still a manual step.

Someone has to click and say: this text can go live.

A human reviews the content.

We don't do that with review data itself.

And we didn't do that in the customer case I just described either.

People are going to ask live questions. You can't manually review every question and every generated answer, because then you're no longer automating anything.

So on some parts of the process there is always a brake.

My rule of thumb would be: if it's content being published in the company's name, there should still be a human approval step.

In other areas, there isn't necessarily that brake, and then you have to accept that occasionally you'll see an odd answer and need to investigate it.

As for your narrow-shoe example: yes, that's exactly the kind of thing review data can be useful for. Something that sounds negative to one person can be highly relevant and positive for someone else, depending on their needs.

Roy Malaiaars: I think there are probably more questions, but considering the time and the afternoon programme — which we want to start around one o'clock — and because we don't want to take away your lunch, I'd like to ask for another round of applause for Marijne.

Thank you.

For the final part of the morning programme, we've already talked quite a lot about the different things changing because of AI.

We now want to look at that from three different perspectives in an interactive panel discussion.

So I'd like to invite Michel, Jannique and Ludo to come forward. The microphones are here.