Nottingham Trent University: Turning site search into instant answers for prospective students

The challenge

Nottingham Trent University (NTU) is focused on attracting prospective students and helping them make confident decisions about where (and how) they want to study.

But as prospective students evaluate universities, they're looking for more than course lists. They want a clear sense of student life – accommodation options, clubs and societies, facilities, wellbeing support, and what it actually feels like to be on campus.

In the context of increasing cost pressure across the higher education sector, NTU was looking for cost-effective ways to address this, without adding significant operational overhead.

Their objectives for the project were:

• Make accommodation and student life content easier to find and act on for prospective students
• Reduce the volume of routine enquiries to staff
• Enhance the digital experience cost-effectively, without adding significant operational overhead

The solution

Already a Squiz DXP and Funnelback customer, NTU introduced Conversational Search – a tool that would allow users to ask natural-language queries related to student life, open days, and the campus experience.

What is Squiz Conversational Search? It enables students to ask questions in natural language and receive direct answers drawn from verified university content. Unlike traditional keyword search, it’s designed to understand intent and return clear, content-grounded responses.

To start, NTU took a focused, incremental approach, launching Conversational Search on a single content topic rather than across the entire site. Implementation ran from November 2025, with go-live in February 2026. The primary focus during that period was ensuring the underlying content was accurate, well-structured, and robust enough to withstand interrogation by the LLM – so that when students asked questions, answers were grounded in verified, trustworthy university content. By scoping responses to a defined section of the website, NTU could test value quickly, learn from real user behaviour, and interpret early performance in context before expanding.

Rather than launching a separate “chatbot” on the site, NTU embedded Conversational Search into the existing site search bar. That means students typed questions into the same search box they already used, and then either received a conversational answer (when the query was in scope) or Funnelback Search results (when it wasn’t).

Following the initial launch, NTU moved into a production phase and kept the use case tightly focused on its first topic. The pilot focused on a defined section of the website covering student life, accommodation, campus experience and related content.

Governance and configuration

Squiz worked closely with NTU to configure how Conversational Search would behave. This included defining the instructions that govern how the tool responds, what topics it should and shouldn't trigger on (including appropriate guardrails around sensitive subjects), and aligning on tone of voice.

To ensure quality before go-live, Squiz ran a structured review and tuning process. Responses were assessed across three pillars – accuracy, completeness, and tone – with a clear threshold of acceptability required before the tool moved to final testing and launch. Where responses fell short, the content and configuration were refined accordingly.

The results

The early data show strong engagement, with meaningful usage immediately after launch. To keep results simple, we’re using the most recent complete month of production analytics.

Conversational Search performance (April 2026)

• Questions asked: 4,237
• Off topic: 1,218
• “Within scope” rate: 71.25%
• Questions answered: 3,555
• Questions unanswered: 682
• Over 80% of questions answered successfully

What these metrics mean (and why they matter)

• Questions asked indicates demand for natural-language Q&A and helps show whether prospective students are using the experience.
• Off-topic and “within scope” rate indicate how well the current content topic matches what students are trying to ask. Off-topic queries can help prioritise future slices.
• Questions answered vs unanswered indicates how often the experience can generate a verified answer from the current content topic. Unanswered questions help pinpoint content gaps and tuning opportunities.
• Answer success rate indicates how consistently students receive a full answer rather than the default response.

What’s next

The next phase will focus on refining the existing experience, improving analytics, evaluating alternative interaction models, and using real user behaviour to determine where Conversational Search can deliver the greatest value before expanding to additional content areas.