Ivan Martin
Election polls are usually consumed as a scoreboard: who is up, who is down, and what happens next.
But Morgan Parnis, CEO of market intelligence firm Esprimi, argues that Malta’s election surveys also offer a practical lesson for local businesses, because the same forces that make voters hard to predict are reshaping consumers, employees, and stakeholders.
Esprimi conducts the election surveys published by Times of Malta, with three polls released across this year’s campaign.
In public, these polls are about politics. Behind the scenes, they are also a demonstration of what modern research has to do to stay useful: update methodology, quantify uncertainty, and turn “messy” responses into insight rather than noise.
The new signal: refusal and uncertainty, and smaller parties
Three features of recent Maltese election polling stand out, Parnis says. First, refusals to answer are rising. Second, “don’t know” responses remain high. Third, support for smaller parties is rising, and the survey trace suggests it is increasingly a settled choice rather than a one-off protest, with a meaningful share of respondents who voted Other last time intending to do so again.
In a traditional mindset, both are treated as a nuisance, responses that reduce sample efficiency and blur the headline number. Esprimi’s view is that they are data points in their own right. Refusal can indicate mistrust, fatigue, reputational fear, or disengagement. “Don’t know” can signal genuine indecision: a voter (or customer) who is waiting for one event, one message, one price change, or one controversy before making a choice.
For businesses, the translation is immediate. If a growing share of the market is undecided or unwilling to engage, it is not enough to ask, “What do people prefer today?” The more useful question becomes, “What would make them decide, and what would make them switch?”
That is where the predictive element of surveying matters: treating uncertainty as something to model and track, not something to ignore.
From telephone-only to mixed methods
Polling methodology has had to evolve because the market has. Telephone surveys were once the primary route to a representative sample. Today, behaviour is fragmented across channels and attention is harder to capture.
Esprimi has moved toward mixed methodologies that combine telephone outreach with online approaches to improve coverage and achieve a better spread across demographics. One of its differentiators is a proprietary online community: a curated panel of thousands of people designed to be representative of the general population. Participants opt in, respond to surveys, and are rewarded through points they can convert into vouchers.
“That is where the predictive element of surveying matters: treating uncertainty as something to model and track, not something to ignore”
For election work, that approach helps mitigate the distortions that can emerge when a single channel over-represents certain groups. For businesses, it offers something equally valuable: speed and repeatability. Companies can measure sentiment, test messaging, and track shifting preferences without waiting for slow, one-off research cycles.
The Obama-era tool now used in Malta
Once the data is collected, Esprimi applies an ensemble of machine-learning techniques; running several models in parallel and looking at where they agree and disagree. The approach has become widely associated with modern campaign analytics since the Obama era, when US campaigns moved from single-model forecasts toward multi-model approaches that quantify uncertainty as well as point estimates.
At a high level, the value of running several models is that it surfaces both consensus and disagreement. Where models agree, the signal is robust. Where they diverge, the divergence itself is information about which assumptions matter most. Instead of treating ‘the market’ as a single average, this approach identifies different possibilities and the conditions under which each becomes likely.
That matters because both elections and markets are often decided at the margin, the persuadable minority, the soft supporters, the late deciders, the quiet churn risk. Machine Learning techniques help identify those areas of conformity as well as areas of tension, which in its own right is an indicator of a changing political environment.
For Maltese businesses, the application is straightforward: segment customers by drivers of behaviour, not just age or location. Segment employees by what retains them, not just job title. And most importantly, build strategy around movement, which cluster is growing, which is shrinking, and which is close to tipping.
From research to “decision intelligence”
Esprimi’s next step is to combine predictive survey work with real-time signals. The company has announced a strategic partnership with CARMA, an international media intelligence provider, bringing enhanced media monitoring to Malta.
The logic is to merge Esprimi’s survey and behavioural insight with continuous monitoring of public sentiment across broadcast, digital, print, and social media—powered by AI analytics and complemented by predictive modelling and human interpretation.
Parnis describes it as a shift away from retrospective measurement toward decision intelligence: tools that help organisations navigate product rollouts, manage crises, and refine market positioning using live feedback loops rather than delayed post-mortems.
In resilience terms, the takeaway is clear. In volatile environments, the winners are not those with the strongest opinions, they are those with the best instruments.
Election surveys show what happens when you track uncertainty properly, modernise your data collection, and use analytics to anticipate change. Maltese businesses facing tighter competition and faster reputational cycles can borrow the same playbook: measure earlier, segment smarter, and act before the market forces a decision.
