From interviews with office managers through a usability test, a Business Model Canvas and a user story map to a clickable prototype: how a travel tool for office management turned into a two-sided marketplace that pools demand and price risk.
The brief for the final exam: we are product managers at a travel company that wants to push the digitalisation of business travel with a new, sustainable product.
Early research showed real demand, above all in one group: office managers. They plan trips for other people, align budgets with management and chase receipts afterwards. Interviews with this group gave us first problem statements from the customer’s point of view and a persona to stand in for them.
Fully automated booking and expensing of business trips by AI agents: people decide instead of administrating.
Together with Yannik Munter I developed the product (called “TripDesk” in the exam, Tabby today) through the whole concept process: problem selection, solution sketches, storyboard, prototype and usability test, then business model, business case and the user story map for the MVP.
After the exam I kept going on my own. The business-case numbers raised a different question than the one we had started with, and the answer became the marketplace model shown in the prototype at tabby.peter-spiegel.com. I built the concept and the prototype with AI.
How might we make travel planning and expensing so efficient for the office manager that the company saves resources?
We worked agile and customer-centred rather than waterfall or stage-gate, so real needs would shape the product early and mistakes would surface early in tests. Concept development ran in three phases: understand, sketch, decide. Each phase ended with a clear result.
Office managers from the target group test the concept on the prototype.
Finance, marketing, customers, technology and design, so desirability, feasibility and viability are judged together.
The decider makes the critical calls (six super votes); the facilitator keeps the team on track.
The interviews surfaced three core problems for the office manager: a lot of time spent, a dense workload and high travel and hotel costs. Before looking for solutions we questioned the problem and ranked the questions as how-might-we notes.
Why does this problem exist in the company at all?
Trips are pieced together across many portals, approvals happen by email and receipts arrive after the trip, often incomplete.
How might we make travel planning and expensing efficient for the office manager?
Top priority: this became the concept hypothesis.
How might we make approvals and budgets so transparent that nobody needs to ask?
Became the approval flow for management.
How might we capture receipts and expenses without manual effort?
Became the receipt scan with automatic matching.
We collected best practices from travel and expense apps, sketched solution ideas and dot-voted three favourite components. These three carried the prototype:
Travellers photograph receipts during the trip instead of collecting them afterwards.
Location and time window assign every receipt to the right trip automatically.
Every trip is checked against policy and budget automatically, with a traffic-light signal.
In the decide phase we isolated the winning solution (from art gallery to heatmap) and turned it into a storyboard. The CEO discusses upcoming company trips with the office manager. She is already at her limit, googles for a tool and finds ours. She signs up with her company address, invites the CEO and brings every employee in during onboarding.
individual observations from the tests, clustered into eight themes
Information density, layout & filters
Roles, perspectives & permissions
Entry point, statistics, AI & notifications
“The traffic-light system is great”: the most frequent positive comment from both participants. The need to act was obvious at a glance, and the clear dashboard and receipt scan landed too.
Recommendation: keep it and link it consistently, so every signal leads straight to a fix.
The most open questions: Who books for whom, who approves, what is the office manager allowed to do? Receipts looked like the traveller’s view, and the perspectives were mixed up.
Recommendation: an explicit role and permission model, a visible approval flow and a self-service area for travellers.
“Too much at first glance”: the entry screen was overloaded and role-specific filters were missing. Questions about the company credit card and data privacy came on top.
Recommendation: slim down the entry screen and filter by role, put active bookings first, separate bonus programmes from the credit card.
The prototype doesn’t work, but delivers essential insights.
Some ideas met the needs, others didn’t. So: improve and test again in another iteration.
The concept fully meets customer needs.
The value proposition in the Business Model Canvas follows directly from the test feedback: a web and mobile application that lets companies (specifically office management and travellers) plan business trips efficiently in time and cost, and expense them inside the app. It covers saving resources, AI receipt capture, clarity, archive & compliance, CO₂ transparency and a mobile wallet that needs no download.
global business-travel revenue 2025, back at the pre-Covid record level of 2019
business trips in Germany 2025, worth €48.6 bn
bookings and 1,800 subscriptions in the business case
Purchase decision with management, main use in office management
Sources: VDR Business Travel Report 2025; Fortune Business Insights (2026).
| Revenue stream | Model | Volume | Price | Revenue year 1 |
|---|---|---|---|---|
| Subscription | Base fee, paid by the company | 1,800 | €288 (€24/month) | €518,400 |
| Licences | per travelling user | 500 | €120 (€10/month) | €60,000 |
| Commissions | share of booking costs | 18,000 | avg. €6 / booking | €108,000 |
| Add-on services | e.g. SAP integration | 20 | €12,000 | €240,000 |
| Total | Base scenario | €926,400 |
| Metric | Worst case | Base | Best case |
|---|---|---|---|
| Revenue year 1 | €499,600 | €926,400 | €2,397,000 |
| Total revenue | €2,721,863 | €6,318,490 | €21,413,349 |
| Total costs | €2,574,346 | €3,205,849 | €6,356,884 |
| NPV | €3,953 | €2,477,076 | €13,162,382 |
| Growth rate | 10% | 30% | 40% |
| Payback | 4.2 years | 0.9 years | 0.7 years |
base scenario
per year
in year one, about 42% of year-one costs
in the base scenario
We translated what the test and the business case taught us into one central user story. Three role stories hang below it and together make up the system. That also answers the biggest open question from the test: who may do what.
Approves trips and budgets visibly and spots overruns immediately.
Manages trip, tickets and receipts on the phone. When something breaks, an alternative arrives automatically.
Books and monitors trips, rebooks and matches receipts on its own.
Scenario: The AI agent proposes a complete, resource-optimised trip plan.
Epics follow the user’s process, with features and user stories below them, ordered by business value. The lines separate the walking skeleton from the MVP scope.
The business case holds up, but it also shows its limits. More than half the revenue comes from subscriptions. On the booking itself the tool earns €6 on average. In the worst case the NPV is barely €4,000. And a pure time-saving tool hardly touches the third core problem from the interviews, high travel and hotel costs: at a trade fair you pay the peak price however efficiently you booked.
What if we didn’t just simplify booking, but took on the price risk ourselves?
Tabby secures fixed allotments early, for inventory that is volatile and peak-priced but predictable (trade-fair hotels, mobility, venues).
Occupancy and price risk are pooled across many clients. The pool absorbs the swings, not the individual customer.
Travellers and dates stay changeable until shortly before the trip. When something critical changes, the agent re-optimises.
Customers (office management, executives, travellers) get a predictable fixed price and full flexibility. Partners (hotels, mobility, event providers) get pooled demand secured early instead of last-minute hope, and keep control of their terms and brand.
Freelancers: planner, traveller and budget owner are the same person. There is nothing to delegate, the volume is tiny and there is no ERP. Tabby is a tool for companies with separate roles.
The market climbs towards the fair while the customer’s price stays flat. The pool manages the difference (example figures per night).
Tabby buys a fixed room allotment early, cheap because it is early and in volume.
PoolThe customer books at a guaranteed fixed price without committing to details.
CustomerChange travellers and dates without rebooking fees.
CustomerMarket price. No cancellation fee: the room goes back into the pool.
CustomerTabby resells the freed-up room at peak price: +€305 per night.
PoolThe 2025 hotel peaks double as the go-to-market order. The best markets for the pool spike high, come back every year and are easy to forecast. Six of them are in the home market.
Oktoberfest, IAA Mobility, bauma. Peak ADR ~€415, ~50% of hotels sold out on top nights.
K, drupa, MEDICA. During K 2025, up to ~77% of hotels were unavailable on the peak day.
CPHI, Automechanika, Light+Building, Book Fair: a tightly packed trade-fair year.
ITB, IFA. 92% occupancy during ITB.
Hannover Messe, Gamescom, Anuga: reliable major events with high demand for beds.
MWC, CES. International benchmarks for later.
The catch on the supply side: in exactly these cities, hotels often reserve their peak inventory exclusively for official trade-fair housing partners.
Resale needs two things at once: a room that has become free and a market that is still hot. Tabby bets on individual, independent dropouts, not on a recession.
Recession or a wave of insolvencies: many allotments come free, but nobody wants them. The capital is tied up in rooms nobody books.
Individual travellers drop out while the fair is booming. The room sells at the top price.
A quiet market. The pool covers its costs but hardly earns.
Business as usual. The margin comes from the planned spread: the bread-and-butter scenario.
Correlation is the crux. The pool is only robust if dropouts are unrelated to market demand. A dominant client, a single fair or a single industry ties the two together, and you end up in the red quadrant. So the whole margin depends on how well the pricing engine forecasts. That engine is the real moat, not the interface.
To close the presentation we ran the model through an enterprise scenario: the German Football Association’s travel agency (DFB-Reisebüro) as a possible first client, with around 18,000 travellers, more than 1,400 team trips a year and over 140 destinations. That adds multi-tenancy, group trips as the core object, an allotment engine, SAP/ERP and rail integration, and traveller tracking for emergencies.
booking volume through the platform
per year
in the scenario
at a 12% discount rate
The risks are named openly too: the custom-software trap, one anchor client tying up the whole team, and the conflict between traveller tracking and the promise not to track employees.
The prototype runs on its own address: the office dashboard with wallet, occasions and trip planning, plus the reasoning as interactive pages. There you’ll find the revenue calculator with worked examples, the data page, personas you can question, and the system architecture with its roadmap. The pages are in German.
Open the prototype ↗ tabby.peter-spiegel.com · designed for desktop (1366 px)I saw it as a mandatory part of the exam. In fact it asked the most important question of the project: what does the product earn on, and which customer problem gets left behind?
Most sticky notes in the test weren’t about pixels but about roles. In B2B tools with several parties, the permission model has to come before the interface.
The marketplace model lives on a bet. Only once I had written down clearly when it loses could I defend the model honestly.
I built the prototype, the calculators and the personas with AI. That let me test every hypothesis as a clickable page instead of describing it on slides.
“I would interview the hotel and mobility providers as early as the understand phase. The second side of the marketplace only came in through the numbers, and it is still the least validated.”