Optimising pricing with AI

Dynamic pricing in holiday rentals is complex. Operators traditionally relied on rigid, rule-based pricing models, requiring constant manual adjustments. While this provided control, it lacked flexibility and responsiveness to market changes.

Key Pain Points:

• Manual Effort:
  Managing extensive pricing rules became overwhelming as inventory grew.
• Lack of Transparency:
  Users struggled to understand how pricing rules interacted, sometimes leading to faulty prices.
• Slow Adaptation:
  Pricing adjustments were often made weekly based on BI reports, missing sudden market shifts.
• User Skepticism Toward AI:
  Early research showed that revenue managers did not fully trust automation and needed to validate AI-generated suggestions.

User Research & Key Insights

We conducted stakeholder interviews, contextual inquiries, and usability tests with revenue managers from multiple organizations.

Findings:

• Users want control.
  AI suggestions must be reviewable before being applied.
• Trust hinges on transparency.
  Users needed clear explanations of AI-driven decisions.
• More data, not less.
  Contrary to initial assumptions, revenue managers relied on detailed spreadsheets and required granular insights.
• Price optimization is nuanced.
  Businesses wanted flexibility to balance between revenue maximization and occupancy rates.

These insights led us to develop a system that empowered users to guide and refine AI-generated pricing rather than rely on full automation.

Designing the AI-Powered Pricing System

1. Mapping the User Journey

We analyzed the manual pricing workflow and identified areas for improvement:

• Operators spent 2+ hours per day adjusting prices.
• Pricing decisions were based on weekly BI reports rather than real-time data.
• Key friction points emerged during price approval and overrides.

Solution:

• Introduced a "Preview Mode", allowing users to compare AI-suggested prices with manual settings before approval.
• Designed a notification-based system that alerted users to pricing anomalies instead of automatically changing prices.
• Enabled users to configure custom monitors to define what pricing changes were relevant to their business needs.

2. Enhancing Transparency & Control

Initial usability tests revealed trust barriers with AI-generated pricing. Users hesitated to accept recommendations due to a lack of clear rationale.

Key UX Enhancements:

• Dual-Table View:
  Displayed AI-generated recommendations alongside historical prices for instant comparison.
• Price Change Justifications:
  Every AI-suggested change included insights into demand patterns and market trends.
• Override Learning Mechanism:
  If users rejected AI prices, the system learned from their decisions to refine future recommendations.
• 4-Eye Principle:
  Introduced an approval workflow where supervisors could validate price changes before implementation.

3. Usability Testing & Iteration

Through multiple testing cycles, we iterated on:

• Labelling & hierarchy to ensure clarity between AI and manual prices.
• Error handling by adding override explanations to build trust.
• Control settings allowing users to fine-tune AI suggestions rather than blindly accepting or rejecting them.

A revenue manager noted:

"I was skeptical at first, but seeing exactly why AI makes a suggestion gives me confidence. It feels like I’m in control."

Impact & Results

Adoption & Business Outcomes

• Initially, only one user engaged with the system weekly. Further iterations on transparency and control increased usage.
• Among five businesses that fully adopted AI-driven revenue management:
  • RevPAR increased by +12.6%
  • Occupancy Rate grew by +4-6%
  • ADR rose by +11.6%

Unexpected User Behaviors

• Despite automation capabilities, many users preferred to manually validate price changes before applying them.
• Businesses requested a "fair pricing safeguard", ensuring that AI-suggested prices remained within reasonable thresholds to protect brand reputation.

Post-Launch Refinements

• Based on user feedback, we introduced flexible pricing strategies that allowed businesses to prioritize revenue, occupancy, or balance both.
• A Revenue Management Day workshop revealed that users wanted more control over small adjustments, leading to further UX refinements.

Key Learnings & Reflections

UX Lessons from the Project:

Fail fast and adapt.
Frequent user validation helped us pivot away from assumptions and build what users actually needed.

Trust is built through transparency.
Users must understand why AI makes decisions before they will trust its recommendations.

Data-heavy users need data-rich interfaces.
Simplification doesn’t always mean reducing information—sometimes, it means structuring it better.

Challenges & Overcoming Them

Early resistance to AI adoption:
We addressed this by incorporating gradual automation and allowing human oversight at every step.

Balancing automation with control:
We introduced configurable notifications and validation workflows instead of forced AI decisions.

Broader Impact on Maxxton’s SaaS Offering

This project marked one of the first major AI-driven innovations at Maxxton, setting the stage for future data-driven automation in the platform.

What I’d Do Differently Next Time

• Invest more time upfront in understanding revenue managers' mental models to avoid initial UX misalignments.
• Use more role-playing and scenario-based testing to refine workflows earlier in the process.

Conclusion

This project successfully demonstrated how AI can enhance pricing strategies without removing human oversight. By combining intelligent automation with user-driven control, we delivered a solution that improved efficiency, increased revenue, and maintained user trust in the decision-making process.