Harnessing AI: Understanding sentiment scores in NRLA reporting
Introducing our new AI platform: Wordnerds
With hundreds and sometimes thousands of individual comments submitted each quarter, how do we ensure that every voice is heard? Coinciding with the launch of Landlord Eye is the full onboarding of a new AI-powered analysis tool, Wordnerds.
With Wordnerds, we can quickly process large volumes of landlord comments, identify key themes across different landlord groups, and understand not just what landlords are saying - but how they feel about it.
Our new platform interprets context and nuance – giving us a clearer, more accurate understanding of landlords’ concerns and priorities.
This blog post is an introduction to how the NRLA are using artificial intelligence to make sure every landlord comment is being considered.
How NRLA Research uses AI:
- The NRLA is committed to using new technology in a responsible and ethical way: data security and privacy remain paramount.
- All analysis takes place within a secure, closed system, and is used to solely amplify landlord voices and support our advocacy work.
- Landlord comments are never used to train external AI models or shared outside of the NRLA. They will always remain anonymous and cannot be individually identfied by any AI tool or platform.
Understanding sentiment
Our new platform understands not only what landlords are saying, but their sentiment towards the topic they are commenting on.
It does this by comparing comments against a library of millions of sentences. The model adapts to landlords’ comments and learns how positive or how negative a comment is.
- Each sentence then receives a sentiment score from 0 (“very negative”) to 100 (“very positive”). This scoring provides clear insights into landlords' feelings and attitudes.
- The platform then averages these sentence-based scores across the whole comment to determine an overall “sentiment score” of that individual response.
One further advantage of this approach is that comments recorded in a survey have their sentiment assessed against a huge “back catalogue” of comments and sentences, not just the comments collected in the same survey.
The approach enables the NRLA to track shifts in landlord sentiment over time – whether optimism is growing, frustrations are rising, or new concerns are emerging. Figure 1 below shows how the platform classifies these sentiment scores:
Figure 1: Sentiment score scale
How this works in practice: Measuring sentiment in landlord confidence
In previous iterations of our research consultations, the NRLA would ask landlords to rate their confidence in the sector and then choose from a lengthy pre-selected list of policies, taxes, regulations, and other factors that they felt were most influential.
This approach was useful in identifying the most frequently selected topics – but gave us no indication of the strength of feeling or the scale of influence these factors has on landlord descision making.
Now however, the NRLA have replaced the tick-boxes, and now invite landlords to comment on the influences on their landlord business, government policy, as well reflect on their own ambitions and goals.
An example: sentiment around the Renters' Rights Bill
In the first Landlord Eye 1, the Renters’ Rights Bill was the most frequently mentioned issue: it appeared in 39% of comments. The average sentiment score of those comments was "41". Some specific aspects of the Bill – which landlords specified in their comments - drew more negative comment and thus a lower sentiment score.
“Without court reform it will be ridiculously difficult to achieve prompt action for possession of property and rent arrears.”
The overall sentiment score of 41 (identical to the sentiment attached to the RRB) was also classed as “negative”, but is closer to “neutral” (45) than “very negative” (24). This is because of other, more neutral comments:
“Demand for property is very high in my area. I am confident that I am providing good quality housing, and my margins are good. Therefore, I don’t see much changing with the new legislation”
Fewer mentions, lower sentiment scores
At the same time, the topics with the lowest sentiment scores - those that landlords felt most negatively about – reveals a different pattern of concerns: Although mentioned less frequently, topics like energy efficiency (36), general costs (33), references to selling (32) and Capital Gains Tax (29) all received much lower sentiment scores than the Renters' Rights' Bill, but were mentioned much less frequently.
Figure 3: Typical landlord comments included
What's driving high(er) confidence?
The platform does not just order negative comments! Many landlords took the opportunity to highlight factors which drive optimism and confidence in the PRS: after all, the most recent Landlord Eye indicates almost two-thirds (62%) of landlords anticipate remaining in the sector at least until the end of 2026.
Three main themes with higher sentiment scores emerged:
- Property as a long-term investment (63 sentiment score)
- Strong tenant demand (59)
- Good relationships with tenants (56)
It is always important to note whether the positive drivers outweigh the negative, or as here, the other way around. This is the value in an “overall score” which reflects the full picture of landlord sentiment, not just individual concerns.
Figure 3 below summarises key drivers of sentiment – both positive and negative
Figure 3: Top positive and negative drivers of landlord sentiment
Summary
NRLA analysis of the comments made in the first Landlord Eye tell us that:
- The Renters’ Rights Bill remains a key concern which draws a lot of pessimistic comment.
- There are however a set of other issues which AI analysis shows are prompting the most negative sentiment, albeit from fewer landlords.
Harnessing sentiment scores allows the NRLA to move beyond simply identifying the most frequently discussed topics. While highlighting common themes remains valuable, understanding how landlords feel about these key issues - and pinpointing precisely what drives negative or positive sentiment - is a crucial step-change in the NRLA’s approach to ensuring ALL landlord voices are heard.
This additional insight enables us to (i) respond more effectively, (ii) advocate more powerfully, and (iii) better support landlord interests.
