Spotting Change Signals
How to run a horizon scan that reaches a decision, and the tools that help with the parts machines are good at.
Photo by Linus Sandvide on Unsplash
Almost anyone can find something interesting. Send a curious person off with a brief to look for weak signals and they will come back with a folder of fascinating links, and everyone in the room will enjoy reading them.
The trouble comes later. Horizon scanning rarely fails at the collection end. It fails at the far end, where a pile of interesting things is supposed to turn into a judgement somebody acts on. I have seen scans running for years inside organisations that could not name a single decision the scanning had changed.
So this is written backwards from that problem. What counts as a signal, where to look for them, how to run the process, and how it connects to a decision. The tools come last, because they solve the easy half.
What counts as a signal
A lot of scanning goes wrong at the definition, so it is worth being blunt about the vocabulary. The one I use runs weak signal, signal, trend, megatrend, in ascending order of how visible and how large a change is.
A weak signal is an early indication of something that might matter, with thin evidence behind it. A signal is the same thing with more corroboration. A trend is a direction of change with enough evidence to be described confidently. A megatrend is a large, slow, structural shift that shows up across many domains at once. Read more about that in this article.
Three questions place most things well enough:
Scale. If this became common, how much would change and for how many people?
Evidence. How much is there, and how good is it?
Novelty. Is this actually new, or an old thing wearing new clothes?
The practical test for whether you have a signal at all has three parts. Something specific has already happened. It is not yet widespread. And it would matter if it were. Miss the first and you have a speculation. Miss the second and you have a trend that everyone already has in their deck. Miss the third and you have a curiosity.
Where to look
The most common failure in a scan is that every item comes from the same place: this quarter, this sector, this language, this political tradition. A scan assembled from the trade press is a very expensive way of finding out what your competitors are already reading.
Some practical correctives:
Read one sector over. Change frequently arrives in an industry from an adjacent one that solved a similar problem first.
Read one country over. Regulatory and social experiments run at different speeds in different places, and somewhere is usually five years ahead on your particular question.
Read the people you disagree with. A political tradition you find uncongenial is noticing things yours is not.
Look at the edges of your own organisation. Frontline staff, customer complaints and the questions nobody can answer are all signal-rich and free.
Watch what changes in law and procurement, which is slower than technology and much more predictive of what will actually be built.
How to run it
The design matters more than the effort. Four things make the difference between a scan that survives and one that fades after two quarters.
Name the intended outcome first. A scan with no owner and no decision behind it is a hobby, and it will be cut in the first budget round, correctly. Before starting, write down which recurring decision this is meant to improve: the annual strategy cycle, the innovation portfolio review, the risk register, the board's horizon item.
Fix the rhythm. Monthly works for most organisations. Fortnightly if the field moves quickly. A standing slot beats an enthusiastic burst every time, because the value compounds only when the record is continuous.
Capture in one shape. Every entry gets the same fields: what happened, source, date, why it might matter, and what would follow if it became normal. That last field is where the thinking lives, and it is the one people skip. An entry without it is a bookmark.
Cluster, then interpret. Scanning produces items. Sensemaking produces meaning, and it does not happen automatically. Set aside a session where the group sorts recent entries into clusters, argues about what the clusters mean, and writes down the two or three implications worth carrying into the decision you named at the start.
Getting from a scan to a decision
This is the step that gets skipped, and it is the reason so much scanning loses its funding.
Keep the scan output short. Three or four clusters, each with a plain statement of what appears to be changing, what would have to be true for it to continue, and what it would mean for the organisation if it did. Attached to each, an indicator worth watching, so that next quarter's scan has something specific to check rather than starting from a blank page.
Then the connection has to be structural, and enthusiasm will not substitute. Someone owns it. It lands in an existing meeting with an existing agenda slot. It produces a recommendation that a named person can accept or reject. Scanning that depends on one motivated individual to carry insights around the building by hand will stop when that person changes role.
This is the part we spend most of our time on with clients: building the foresight system end to end, from where the signals come from through to the decision they are meant to improve, so that scanning becomes an operating capability rather than a project.
What a collective scan looks like
When at the Copenhagen Institute for Futures Studies I oversaw the Global Scanning Network, a distributed group of contributors from around the world and across disciplines who fed signals into a shared pool. Amazing in scope and ambition and it worked well.
The value of running it collectively was mostly about coverage. Any individual scanner, however good, reads a finite and fairly predictable set of sources. A network with different geographies, languages and professional backgrounds catches what a single perspective structurally cannot, and it does so cheaply, because everyone is already reading something.
Two things I would pass on from it. The contributions get much better when contributors are asked for the "so what" rather than only the link, because that is what turns a person from a feed into an analyst. And the pool needs curating: an unfiltered stream of everything anyone found becomes unusable at surprisingly low volume.
You can run a version of this internally without any special infrastructure. A dozen colleagues from different functions, a shared capture template, and one hour a month is a functioning network.
Tools, and the question to ask of all of them
Software helps with capture, storage, deduplication, tagging and retrieval. It is now very good at surfacing candidate signals from large source sets. It does not do the sensemaking, and a team that buys a platform hoping it will supply the judgement will be disappointed at renewal.
The landscape has also changed shape since this article first appeared. In 2023 there were a handful of serious tools. Now every person and their uncle can launch a platform in a weekend, because the hard engineering has become a prompt. So the question worth asking has moved. It is no longer what the tool can do, since they can all do roughly the same things. It is:
Who verified this signal? Somebody, or nobody?
Was it made sense of, or only collected? A feed is not intelligence.
What is the process behind it? If there is no editorial layer, you are buying a search engine with a radar skin.
Is the signal actually interesting? Volume is easy now. The scarce thing is a signal that makes a room sit up.
Three ways to approach it
Build your own. Tempting, and more achievable than it used to be (thanks to AI). The usual argument is control, particularly of your own data. That argument is thinner than it looks, since the AI provider still controls the infrastructure underneath whatever you build. The harder question is about advantage: if your signals are generated by a model reading public sources, what is your distinctive perspective? Everybody's model reads the same internet. Building your own tool does not give you a point of view.
Buy acces to a platform. The established players are properly built and their libraries are verified by professionals, which is the thing you are actually paying for. The trade-offs are real. You do not own the platform, it is a budget line that has to be defended annually, and there is some degree of lock-in. Check the export terms before signing, because several of them do allow you to take your signal base with you, and that changes the risk considerably.
Also, a catch is that shared infrastructure does not create shared habits. It needs agreed processes, a system for who contributes what and when, and humans in the loop doing the interpretation. Without those it becomes a shared folder nobody opens.
Scan by hand. The least glamorous option, and much better than nothing, provided enough people pick up the habit and keep it. A spreadsheet maintained by fifteen curious colleagues who have been shown what a good entry looks like will beat an unused licence every time. As long as it feeds into the work, there is real value in it.
Some futures intelligence platforms worth knowing
FIBRES offers AI-assisted sourcing, a central signal and trend database, and clustering and sensemaking support. One of the more customisable options. Its foresight capability also reaches the market through Valona Intelligence, so treat the two as related when comparing.
Newness is a collaborative cloud platform built for scanning at scale, with automated scanning, curated source packs, radars and scenario tools. The most accessible entry point for a small team.
Futures Platform is visual and collaborative, with a curated futures library that does a lot of the early work for you. It merged with the market-data company Statzon in April 2025 and has since launched Synapse, an AI-native futures intelligence workspace.
Strategic Intelligence by the World Economic Forumprovides expert-curated briefings across several hundred topics, with transformation maps showing how issues connect. There is a free browsing tier, which makes it the easiest place to see what curated futures intelligence looks like before committing to anything.
ITONICS is the heaviest enterprise option, with structured signal-to-action workflows and AI trend recommendations across a very large signal base.
Start smaller than the tooling suggests
Almost every organisation that asks us about scanning platforms would be better served by running a manual scan for two quarters first. A shared document, a fixed monthly hour, a named decision to feed. That tells you what you actually need before you buy anything, and it surfaces the real constraint, which is almost never the software.
We help organisations design and run foresight systems end to end, from horizon scanning through sensemaking to the decisions the intelligence is meant to serve. If you have a scan that produces interesting reading and no decisions, that is a solvable problem, and a common one.
Originally published in 2023. Revised and all platform information re-verified in September 2026.
Frequently asked questions
What is horizon scanning? The systematic search for early indications of change that could affect an organisation's operating environment, together with the interpretation of what those indications might mean. It is a continuous practice rather than a one-off exercise, and its output is intelligence that feeds decisions.
What is the difference between a weak signal and a trend? Visibility and evidence. A weak signal is an early indication with thin corroboration. A trend is a direction of change with enough evidence behind it to be stated confidently. A signal sits between the two, and a megatrend is a large structural shift visible across many domains.
How do you tell a signal from noise? Three tests. Something specific has actually happened. It is not yet widespread. And it would matter at scale. Anything failing the first is speculation, the second is already a trend, and the third is a curiosity.
How often should you scan? Monthly suits most organisations, fortnightly in fast-moving fields. Consistency matters more than frequency, because the value of a scan comes from the accumulating record.
Do you need a horizon scanning platform? Not to start. A shared document, a fixed monthly slot and a named decision will take an organisation through its first two quarters and reveal what it actually needs. Platforms earn their cost when volume, contributor numbers or retrieval become the bottleneck.
Should we build our own scanning tool? It is easier than it used to be, and the control argument is weaker than it sounds, since the AI provider still controls the infrastructure underneath. The question that matters is what makes your signals distinctive. If they are generated by a model reading public sources, everyone else's model is reading the same sources. A tool does not give you a point of view.
Why do horizon scanning programmes get cancelled? Almost always because they never connected to a decision. A scan without a named owner, a standing slot in an existing meeting and a recommendation someone can accept or reject will be read with interest and cut in the next budget round.
A hands-on guide to foresight frameworks, tools and process
There are dozens of foresight tools available. Which one do you reach for, and when?
Doing foresight means choosing the right method for the question in front of you and sequencing those methods into a process that leads somewhere. Foresight is a strong asset for exploring future possibilities, but without guidance on which tool to use when, it becomes overwhelming quickly. This masterclass is about making those choices with more confidence.
In this practical masterclass, foresight experts Mathias Behn Bjørnhof (ANTICIPATE) and Chloé de Ruffray share how to unlock the strategic value of foresight for your clients or your team by choosing, applying and mastering the right tools. The session runs from scoping and framing, through horizon scanning and the futures wheel, to scenarios and backcasting.Foresight is a great asset to explore future possibilities, but without guidance on which tools to use when and how, it can quickly become overwhelming.
What you will learn
✺ How to scope a foresight process and pick a framework that fits the question
✺ When to use horizon scanning, the futures wheel, scenarios and backcasting, and when to leave them out
✺ How to sequence tools so a process moves from scanning to sensemaking to scenarios to decisions
✺ How to keep foresight tied to strategic value rather than to method for its own sake
Methods and tools covered
✓ Scoping and framing the foresight question
✓ Horizon scanning - gathering and organising signals of change
✓ The futures wheel - mapping first, second and third order consequences
✓ Scenario planning - building plausible and useful futures
✓ Backcasting - working back from a future to the steps that get you there
Who it is for
New and seasoned foresight professionals, strategists, innovation teams, and curious minds eager to sharpen and expand their practice.
What you receive
Your purchase is a downloadable zip file containing:
✓ The full session video replay
✓ Slides and workbook
✓ Full transcript
Buying for a team or through your company?
If you would like several seats, an invoice, or a purchase made through your organisation rather than by card, fill in our contact form and we will arrange it. We also run this masterclass as a live, tailored session for teams.
By purchasing this digital product, you confirm that you agree to the following statement: "I agree that delivery begins immediately upon payment, and I understand and accept that my 14-day right of withdrawal expires once delivery has begun.” (see Return Policy for more information)
VAT (25%) is applied according to the EU place of supply rules for digital services. Danish VAT number: DK45996689
