- Home
- /
- By your needs
- /
- Monitor populations and places
Understand what is happening to a population before it becomes a headline.
Monitoring is not waiting for something to happen. It means detecting the pattern, the variation between areas, the widening inequality or the signal that is starting to move, with methods that tell the real trend apart from the noise.
Changes in a population can leave signals before their scale becomes evident: in the geographical distribution of a health event, in the differences between social groups, in what people search for online, in what is published on a topic. The challenge is not a lack of data, but distinguishing the signal from the noise within them, and doing so methodically rather than by eye.
Monitoring and exploring populations, places and information flows always follows the same logic: starting from data that already exist (registries, surveys, searches, document corpora) and extracting an interpretable pattern from them, with the uncertainty quantified and the limitations on the table.
It covers three strands, depending on what you want to monitor: how a phenomenon evolves over time and varies across areas; the needs and inequalities that separate some groups from others; and the digital, documentary or scientific signals that help describe changes, interests and information structures. You can start with whichever applies to you.
This is for you if…
You need to monitor how a health phenomenon evolves in a population and detect changes in trend with statistical judgement, not just by checking whether the figure goes up or down.
You suspect that a problem is not evenly distributed across areas or social groups, and you want to quantify that variation and those inequalities rather than guess at them.
You want to know what people search for, ask or share on the platforms analysed about a health topic, and use that digital signal as a source of population information.
You work with a large volume of text (scientific literature, publications, document corpora) from which you need to extract topics, trends or structure systematically.
How we can help
Surveillance and geographical variation
Monitor how a phenomenon evolves over time and describe how it is distributed across areas.
The systematic monitoring of a health event in a population: defining indicators, estimating prevalence and incidence, standardising to compare between groups or periods, detecting changes in trend. The statistics that turn a flow of health data into an interpretable surveillance system.
View service →When where something happens matters as much as how much of it there is. Analysis of the geographical distribution of a health phenomenon: risk maps, detection of spatial clusters, models that account for geographical neighbourhood. It makes it possible to quantify the variation between areas rather than assume it.
View service →Estimating how much of a nutrient or food a population consumes from food consumption surveys: adult-equivalent adjustment, handling of individual variability, estimation of the population intake distribution. The basis for monitoring the nutritional adequacy of a population.
View service →When what matters is evolution over time: decomposing a series into trend, seasonality and irregular component, detecting change points, modelling temporal dependence to assess whether the observed change exceeds the expected temporal variation. The tool for reading how a health indicator changes over time.
View service →Needs, inequalities and determinants
Quantify the differences that separate some groups from others and the factors associated with them.
Measuring how much, and how, some social groups diverge from others in health: inequality indices, socioeconomic gradients, concentration curves. Putting a number on an inequality is the first step towards discussing it with evidence rather than impressions.
View service →Studying which social, economic and environmental factors are associated with a population's health outcomes, and how strong those associations are. Models that help to understand what lies behind a health pattern, beyond the strictly clinical, with the caution not to confuse association with causation.
View service →When monitoring relies on a survey, the sampling design determines whom its results represent. Stratified, cluster and design-weighted sampling so that the estimates describe the target population, provided that selection, data collection and weighting are carried out correctly.
View service →Quantifying the burden of a disease or risk factor on a population by combining mortality and morbidity into comparable indicators. A summary measure for gauging the magnitude of a health problem and comparing it between groups or areas.
View service →Digital, documentary and scientific signals
Extract population information from what is searched for, published or written, with reproducible methods.
Internet searches can provide an indirect signal of interest or concern around a topic. Analysis of search series (relative volume, seasonality, correlation with events) to use search activity as a source of population information. These are observational data, not designed for research, and they are interpreted with that caution.
View service →The analysis of what circulates on social media about a health topic: volume, evolution over time, structure of the conversation. A complementary source for studying how a topic circulates and is represented among the users and platforms analysed, treated with the rigour of an observational source, not as gospel.
View service →When the volume of text exceeds what can be read, natural language processing extracts structure from it: latent topics, trends, sentiment. Topic modelling, sentiment analysis and document corpus mining, applied with validation and not as a black box.
View service →The quantitative analysis of the scientific output on a topic: what is published, by whom, how it evolves, how it connects. Mapping the output, evolution and structure of a research field with reproducible methods, useful as the basis for a review, a research strategy or an assessment of a field.
View service →When what matters is the structure of relationships (between authors, between terms, between actors in a system), network analysis makes it explicit: who is connected to whom, which nodes are central, how they cluster. A lens for reading the structure that underlies a corpus or a system, beyond counting isolated elements.
View service →Not sure which of these fits your situation?
Tell us what stage you are at and we will tell you what work is needed, or whether none is. No commitment, and to the point.
Let's talk about what you need to monitorHow we have approached it in other projects
Before you take the step
What kind of data can I use to monitor a population, and do I need to collect them myself?
You can often start from data that already exist, although some surveillance systems do require specific data collection or broader sources. This applies when you want to monitor how a health phenomenon evolves, the differences between groups or areas, or the signals left by searches and publications on a topic. The usual sources are health registries, health surveys, indicator series, internet searches, social media activity or corpora of scientific literature. The challenge is not usually a lack of data, but distinguishing the signal from the noise within them methodically, rather than by eye. You don't need to know which of these sources is yours; the work is organised into three strands (geographical variation, inequalities and determinants, and digital signals) so that you can start with the one you recognise.
How can I tell a real change from a one-off fluctuation?
Distinguishing a real trend from a passing variation is exactly what methodical monitoring provides, as opposed to checking whether the figure goes up or down. It arises whenever an indicator moves and you have to decide whether that movement means something or is expected noise. To do this, the normal variation of the phenomenon is modelled (its temporal dependence, its seasonality, its irregular component) and the observed change is assessed to see whether it exceeds what would be expected by chance, rather than reacting to every fluctuation. An isolated figure that rises is not a trend; a one-off peak is not a structural change. What we avoid is the double trap of the false alarm and the real change that goes unnoticed, so that alerts respond to changes that the expected variation does not explain.
Can different types of data (geographical, social, digital) be combined in the same monitoring?
Yes, and combining sources often gives a more complete picture than any of them separately. This is the case when the same phenomenon leaves traces in several places: its distribution across areas, the differences between social groups, what people search for or share about it. That said, combining sources requires treating each one for what it is: a health registry, a survey with its sampling design and a series of internet searches do not have the same status, and mixing them without that care produces fragile conclusions. Digital signals, in particular, are observational data not designed for research, and they are interpreted with that caution. The aim is monitoring that integrates different sources while respecting the weight and limits of each one, not an aggregate that treats them as equivalent.
How reliable are internet searches or social media as a health signal?
They are an indirect and valuable signal, but an observational one: they tell you about interest, concern or conversation around a topic, not about the clinical reality of a population. This arises when you want to use search activity or what circulates on social media as a complementary source of population information. Their value lies in capturing early, and at low cost, how interest in a topic is moving; their limit lies in the fact that they reflect the people who search or post on the platforms analysed, not the whole population, and that they were not designed for research. That is why they are treated with the rigour of an observational source (characterising their temporal pattern, their correlation with events, their biases) and not as gospel. What we avoid is turning a volume of searches into a claim about prevalence or causation.
I work with a volume of text I can't read in full (literature, publications, documents). Is that monitoring too?
Yes: extracting structure from a large corpus of text is one of these strands, that of documentary and scientific signals. It arises when the volume of literature, publications or documents exceeds what can be read by hand and you need to identify systematically the topics, trends or structure it contains. For this, we use natural language processing (topic modelling, sentiment analysis, corpus mining), bibliometrics to map the scientific output of a field, and network analysis to make explicit the structure of relationships between authors, terms or actors. All of it with validation, not as a black box that returns results without any control. The outcome is a reproducible map of what a corpus contains (what it talks about, how it evolves, how it connects) without depending on a complete manual reading.
Do I need to be clear about what I want to monitor before getting in touch, or can you help me define it?
You don't need to bring it fully defined: narrowing down what to monitor and with what data is part of the work, not a prerequisite. This arises when you recognise that something is happening in your population, your area or your topic that is worth following, but you have not yet defined the indicator, the source or the scale. With the general question and the context, the decision is made on what is worth monitoring, which existing data make it possible (or what additional data collection would be needed) and which method separates the signal from the noise. What does help is to be clear about the phenomenon of interest and the scope (geographical, population or information) even if the specific indicator is still to be defined. If what you already have is a detected pattern and you need to communicate it, the natural next step is to **communicate results and build tools**.
Monitoring produces findings that almost always need to be communicated: a geographical pattern, a quantified inequality, a trend. When you have the result and need others to understand it and use it, the natural next step is to communicate results and build tools they can work with. And if you are still defining what you want to monitor and with what data, let's talk first.
Shall we talk?
Tell us what you are working on and we will give you a straight answer on how we can help.