The data company Schober Information Group has become Capaneo – and with udo, the company now relies on a data intelligence platform with AI agents. Sven Waldenmaier, has been supporting tourism customers since the 2000s. In an interview with turndown.de, he explains why “Hosted in Germany” is not just a marketing slogan, what tour operators like alltours get out of a clean database – and what advice he gives hoteliers before their next CDP project.

What is the difference between a tourism client and an industrial client in day-to-day consulting – and what attracted you personally to the industry?

The biggest difference: tourism customers sell experiences. This sounds like a marketing phrase, but it fundamentally changes how data must be read. An industrial customer thinks in terms of delivery times and repurchase cycles. A hotelier thinks in terms of emotions, seasons and life events. The question is never just “Who bought last?” but “Who will book next summer – and what motivates them right now?” This makes the work analytically more complex, but also more honest: if the segmentation is right, you can see it immediately in the conversion rates. What gets me personally is that hardly any other industry has so much data and uses so little of it. That’s a real lever – and it’s fun to use it.

What is the question you will be asked most often in 2026?

“How do we use AI?” – this is the question I will be asked most frequently in 2026. Almost word for word, across all industries. What’s behind it is usually still unclear: Should AI write texts? Manage campaigns? Understand guests? The awareness is there, but the target image is often not. What I then explain: AI is not a switch that you flip. If you simply load customer data into an LLM, you get superficial outputs – and in the worst case, a compliance problem. What makes AI really powerful is the underlying database: consolidated, cleansed, enriched with third-party data and potential data – and ideally anonymized – customer information. Only then can AI say where market potential lies, which segments can grow and how to tap into new target groups based on data. Many companies want this result, but do not yet have a plan on how to get there. Our job is to develop precisely this plan.

From Datenhaus Schober to Datatech Capaneo

What role has tourism historically played in Schober’s data universe?

Tourism was never the loudest segment at Schober – but one that I personally got to know very deeply. As an analyst, I created customer profile analyses for almost every tourism segment in the 2000s: Package tour providers, wellness hotels, family resorts, luxury cruise lines, even tourism associations. The task was almost always the same: to create statistical twins in order to optimize catalog distribution for new customer acquisition. What this has given me is an industry knowledge that I still use today – I not only understand the data, but also the motivations behind it. People who book a cruise think differently from camping holidaymakers, and people who go to a wellness hotel react to different stimuli than city travelers. This understanding of target groups makes a difference when it comes to building data models that are not only technically clean, but also meaningful in terms of content.

Schober has become Capaneo. What remains of the DNA legacy – and what had to be left behind?

What remains: the depth of data and the seriousness in dealing with customer data. Schober never played for quantity, but for quality and relevance. This is a cultural value that Capaneo has taken with it. What has changed is our self-image as a pure data and solution provider. We used to supply addresses and selections – that was it. Today, we are just as deeply involved in the question “What do you do with the findings?” as we are in the question “What data do you need?” We have always taken the data processing side seriously – but today it is on an equal footing with data intelligence. This requires different skills, a different approach and a different product. udo is the visible symbol of this change.

Capaneo advertises with 70 million consumer and 4.6 million company data. Which features are particularly valuable for tourism customers?

Travel affinity is the obvious thing – but that alone is not enough. What really sets tourism customers apart is that they need contextual knowledge. Who travels how, when, how often, with whom – and above all: why does someone change brands? This is why our tourism customers are particularly relevant: Lifestyle segments (travel experience, preferences), household composition (family with children vs. best agers), purchasing power and creditworthiness characteristics as well as geographical mobility patterns from movement data. The latter are comparatively new, but highly relevant: They show which places someone actually goes to – not just what they say they go to. And they are particularly suitable for reactivating old bookers in the knowledge that a high volume of travel is still taking place.

udo: What makes a CDP with AI agents different

What is udo – and what is it explicitly not?

udo is the system that understands, evaluates and translates your customer data into actions – without the need for a data engineer. It’s not another email tool or a classic CRM that manages contacts and writes tickets. It’s also not a BI dashboard that shows pretty graphics but doesn’t provide any recommendations for action. udo connects: Data consolidation, analysis, AI-supported decision-making logic and campaign management in a system that talks to existing tools instead of replacing them. In concrete terms for hotels: you finally know which guests can be reactivated, which are at risk of leaving, and what you should offer them and when. And very importantly: udo brings its own data with it – features that can be used to profile customers and identify new potential. In addition to the market potential view, concrete measures can be derived immediately.

The CDP market is full – Salesforce, Tealium, Bloomreach, Adobe. Where do you draw the line to the big US platforms?

Salesforce, Adobe, Bloomreach – they are built for the enterprise and for enterprise budgets. If you’re a medium-sized tour operator introducing one of these platforms, you’re also buying a multi-year implementation process, external consultants and a dependency that is difficult to break. We draw the line where time-to-value begins. A customer should see the first results within weeks, not after quarters, usually years. Also: data sovereignty. With the large US providers, customer data flows into global platform infrastructures. At udo, it stays where it belongs – under the customer’s control, in Germany, in compliance with the GDPR.

A key promise is that sensitive business data will not flow directly into an LLM. How does this work technically?

This is one of the most important technical points, especially for hospitality customers. Guest data is highly sensitive – booking history, payment behavior, personal preferences. All of this must not flow unprepared into a Large Language Model that is operated on external servers. udo solves this as follows: before an AI agent becomes active, udo has already prepared and anonymized the raw data – in other words, udo puts the data into a form that ensures that LLMs work based on facts and without hallucination, without knowing personal data. The agent therefore does not work with the data record “Mr. Müller, room 214, has cancelled three times and is a VIP” – but with an anonymized form: Facts on cancellation behavior and customer status, without personal reference. The intelligence lies in the database, not in the prompt.

How autonomous are AI agents in practice – where is the marketing manager pilot?

The marketing manager remains the pilot – always. The agents do what takes time and does not require creative decisions: Analyzing target groups, outlining campaign ideas, formulating content in a first draft, researching contacts. What the human does: check, decide, approve. I like to say: udo makes you faster, not superfluous. The real benefit is focus – instead of three hours of list work before every campaign, you can use the time for the question that only you can really answer: What do we want to achieve with this contact?

You talk about “control agents” that are supposed to prevent hallucinations. What exactly does such a control mechanism look like?

In concrete terms, this means that udo works with a defined, validated database. When an agent makes a segmentation recommendation or suggests a subject line, this is not based on free association, but on patterns that we have previously derived from real customer data. A control agent checks in the background whether the output matches the parameters – target group size plausible? Tonality suitable for the segment? No contradictory statements? This is not a panacea against errors, but it is different from a bare ChatGPT prompt. The system knows what it doesn’t know – and tells you so.

Tourism use cases in practice

alltours is your prominent tourism case. What was the decisive lever there?

At alltours, the starting point was classic: booking data was distributed, communication was largely broadcast, and the cross-selling of additional services – travel insurance, transfers, upgrades – was done via undifferentiated mass mailings. udo was customized: Data consolidation, data cleansing and refinement, and enrichment with the integrated data universe of 70 million consumers. The result: an AI-based target group selection based on real customer behavior. Booking figures and sales increased after just a short time. The decisive lever was not a single feature – but the combination of a clean database, precise segmentation and fully automated campaign management. Dr. Georg Welbers, Managing Director of Marketing and Sales at alltours, sums it up like this: Customer management has been noticeably optimized, sales and customer satisfaction sustainably increased.

How do CDP setups differ between package tour providers, hotel chains and destinations?

The data logic is fundamentally different. A tour operator like alltours has transactional data in abundance – booking paths, cancelations, additional products. A hotel, on the other hand, often has fragmented data: PMS, loyalty program, booking platform – three systems, three realities. A destination has no direct booking data at all, but has to work via detours – transaction data, tourist cards, partner feeds. This means that the CDP approach must be calibrated differently in each case. What remains the same is the core question. Who is my guest really, what motivates them – and what is the next meaningful point of contact?

Which tourism data sources do you typically connect to – PMS, IBE, Loyalty?

Essentially, there are three layers. Firstly, the transactional systems: PMS such as Protel or Apaleo, booking engines such as OPTIRES or Nemo, channel managers. Secondly, the communicative systems: email marketing tools, CRM. Thirdly, external enrichment sources: Transaction data, purchasing power data, lifestyle characteristics from our own database. Loyalty programs are often the most exciting starting point because there are real opt-ins. The trick is not to connect everything – but to prioritize the right sources in the right order.

Do you have a specific example where a tourism customer had an “aha experience” thanks to udo?

The most striking example is when a hotelier sees for the first time how many of his “lost” guests could actually still be reactivated – segmented by reactivation probability. One of our customers discovered in the first step that a third of the contacts listed as inactive were not inactive at all – they were just not recognized correctly because they had rebooked under a different email address. Solving this duplicate problem was not an AI moment – it was data hygiene. But the result was that the reactivatable base was suddenly 21% larger than expected. Only then could we talk about personalization in a meaningful way.

How tourism customers work with CDP data

Which three to five use cases pay off most quickly in tourism?

In my experience, there are these five, in approximate order of time-to-value: Firstly, reactivating inactive customers – low contact costs, known people, high conversion if the offer is right. Secondly, cross-selling of additional services – insurance, upgrade, transfer – because patterns can be modeled very well here. Thirdly, newsletter personalization: fewer mailings, more relevance, measurable in opening and click rates. Fourthly, win-back after cancellations – an often underestimated contact point. And fifthly, lookalike audiences for digital channels: If I know who my best customers are, I can mirror them digitally. This is not a new concept, but it is much more accurate with a clean database. What sets us apart from generic platform lookalikes: We work with real brand attribute and location scoring. This means that we can not only identify “similar people” for a hotel, but also those who have a proven affinity with this brand and this destination – a decisive difference in conversion quality.

Booking behavior has changed permanently since Corona. What do your data patterns show?

Shorter lead times are real and here to stay – that’s no longer a surprise. What interests us is the segmentation behind it: Not all last-minute is the same. There is a type of customer who structurally books at short notice and is still loyal. And there is the guest who used to book early and is now hesitant – a sign of declining brand loyalty or economic pressure. If you can distinguish between the two, you can communicate in a more targeted way. What we are also observing: Flexibility as a booking feature is a clear predictor of churn risk. Those who only choose offers that can be booked flexibly are more sensitive to comparison.

How exactly is a lookalike audience built in udo and activated in Adform, Google or Meta?

The starting point is always the existing customer base: who are my most valuable customers – measured in terms of CLV, booking frequency, additional sales? From this core, we create a profile based on a neural network of real characteristics: socio-demographic, geographic, behavioral. This profile is then mirrored against our external database of 70 million consumers – and the most similar profiles are selected. The result is a lookalike audience with a real data anchor, not just algorithm-generated approximations as with Meta. For activation, we have established interfaces to all digital platforms – DSPs such as Adform or Google, but also large audience platforms. The transfer is pixel-based or via hashmatching – GDPR-compliant.

Keyword first-party data strategy: Which tourism customers are particularly far ahead of you?

The pioneers do one thing differently: they treat data not as an IT issue, but as a business strategy. That sounds banal, but it’s not. In practice, this means that data strategy is part of the marketing plan, not just the technology budget. Opt-in acquisition is an active goal, not a by-product. And the question “What do we do with this data point?” is asked before it is collected – not after. What these customers actually do differently: They invest in clean permission structures, build loyalty programs as data sources and actively use booking routes to enrich profiles.

Concretely measurable added value

You advertise with 30 % more sales potential, 90 % more freedom, 25 % less process costs. How do these figures come about?

These figures are based on aggregated experience from our customer base – they are not a guarantee, but a guide. We see the 30% sales potential most often in the reactivation and cross-selling context, and it is usually the result of the first year, not the first month. The 90% freedom in marketing resources describes something real: when campaign selection, segmentation and content design are automated, the work shifts from doing to deciding. The 25% process costs are the most difficult to measure – but are often even conservatively estimated for customers who previously had manual reporting and external service providers.

How does a hotel manager or CMO recognize in 12 months that an udo project was successful?

The key figures we talk about in Quarterly Reviews: Reactivation rate of inactive customers, open and click rates compared to the previous year, cross-selling rate for additional services, data cleanliness in the inventory (duplicate rate, deliverability) and – if available – CLV development of the best segments. This is just as important, but is measured less frequently: How much time does the marketing team save per campaign? If the answer after a year is eight hours per week, then udo has already paid off. The soft KPIs are often the most honest indicators of whether a system is really being used – or just running.

GDPR, EU AI Act, data sovereignty

“Hosted and made in Germany” is often an empty phrase in marketing. What does this really mean at Capaneo?

In concrete terms, this means that all data is processed and stored exclusively in German data centers – under German law and GDPR-compliant order processing in accordance with Art. 28. No data transfer to third countries without an explicit legal basis. Order processing in accordance with Art. 28 GDPR is standard for us, not optional. Pseudonymization takes effect at several levels before data flows into analysis or AI processes. What this means for hotel customers: Guest data – including booking history and personal preferences – does not leave the defined processing framework. This is not a marketing argument for European hotels, but a legal necessity. And one that we fulfill.

The EU AI Act is gradually coming into force. What exactly will change for tourism customers?

The directly relevant category are systems that generate recommendations with an impact on individual decisions – i.e. targeting, scoring, price recommendations. These fall under transparency obligations: The user must know that AI is involved. For marketing automation in tourism, this means in practice: documentation requirements for the models used, explainability of decisions (why does segment X get offer Y?) and clear responsibilities. udo addresses this through the architecture itself: Because AI agents work on interpreted, comprehensible data – not on black-box algorithms – explainability is structurally anchored, not added later.

Outlook

Where do you see the tourism industry with CDPs and AI agents in three years’ time – standard or competitive advantage?

Standard – but not for everyone at the same time. What is changing: The term CDP will disappear, the function will remain. In three years’ time, no tour operator will have to explain what customer data consolidation means. The question will be who does it better. The pioneers who start today have a structural advantage: better data quality, more historical patterns, trained models. AI will not be the differentiating factor – the company’s database and data culture will be. That is actually the difficult homework.

If you could give a marketing manager in a medium-sized hotel a single piece of advice – what would it be?

Start with your inventory. Not with new technology, not with a CDP tender. Ask yourself: How many of my contacts are actually reachable today? How many are duplicates? How many have no permission status? These are figures that you can have in four weeks – and that will immediately show you where you stand. The most sensible next step almost always emerges automatically from this analysis. In most cases, it is data hygiene, then segmentation, then activation. If you start with activation, you are building on sand. If you start with inventory, you are building on knowledge.

What is on the roadmap for udo and Capaneo in 2026/2027 that should be of particular interest to tourism customers?

What I can say: We are working on further standardizing the interfaces to PMS and IBE systems – this is the most common point of friction when onboarding hotel customers. We are also further developing udo’s agent capabilities, specifically in the direction of automated campaign control based on real-time triggers – in other words, no longer just scheduled mailings, but context-sensitive triggers. For tour operators and hotel chains with loyalty programs, we are working on deeper CLV models that combine churn risk and upgrading potential. The goal: less manual model maintenance, more automatic recommendation. What I would like to emphasize here: Of course we have a classic roadmap – but Capaneo’s real strength lies in the speed with which we implement individual customer requests. It is therefore not yet possible to say in full what udo 2027 will be able to do – because the best further developments often arise directly from collaboration with our customers.

An interview with Thomas Harnisch, founder and publisher of turndown.de

Has been working in tourism and digital marketing for over 13 years – with positions as online marketing manager at a luxury cruise line and in the top hotel industry. Founder of TURNDOWN, the digital magazine for the hotel, restaurant and tourism industry.