J. Paul is a London based designer and researcher with expertise in Speculative Design, Service Design, Design Research, and Strategy.
A signal is a specific, concrete thing that has already happened and suggests a new possibility has become real: it has a who, a when, and a particular outcome. A trend is an abstraction drawn across many signals — a direction of travel that no single event demonstrates on its own. A megatrend is either the same trend operating at much larger scale, or several trends converging into one larger shift. These are three levels of abstraction, not three synonyms, and the difference between them decides how early you can see a change coming and how much advantage seeing it gives you.
A signal has a date on it
The test for a signal is whether you can name who did what, and when, and what specifically resulted. If you cannot, you do not have a signal.
In August 2022, at the Colorado State Fair fine arts competition, Jason Allen entered an image called Théâtre D'opéra Spatial in the digital art category and won first place. The image had been generated with Midjourney. The judges said afterwards that they would have awarded the prize anyway.
That is a signal. It has a person, a date, a venue, an outcome, and a public record. Note what it is evidence of. Not that software can produce images — that had been true for years. What became newly real that August was narrower and more consequential: a machine-generated image could be judged against human work by people applying human criteria, and win. The category of competence that the argument had been resting on quietly moved.
"AI is getting good at images" is not a signal. It is a summary of a mood, and there is nothing in it anyone can check, date, or disagree with usefully. That distinction sounds pedantic until the first time a sceptical executive asks where a piece of futures work came from. A dated event survives that question. A mood does not.
Signals are also, importantly, historical claims rather than predictions. Nothing about a signal asserts what happens next. It asserts only that something has already happened which was not previously possible, or not previously ordinary. Everything after that is interpretation, and interpretation is where the disagreement — and the value — lives.
Trends are abstractions, and they arrive late
Take a different set of events. Norway crossing the point at which the large majority of new cars sold are battery-electric. A national government fixing a date after which new petrol and diesel cars may not be sold. Volvo announcing it will make only electric cars. A manufacturer converting an assembly line that has built combustion engines for forty years.
Each of those is a signal: dated, attributable, verifiable. None of them is the trend. The trend is the thing you infer by holding them together — electrification of the car: rising electric share of sales, shifting consumer preference, and manufacturers reorganising production around it. No single event proves that. The abstraction does work that none of its members can do alone.
Scale it, and it becomes something else again. Heat pumps replacing gas boilers. Induction hobs replacing gas rings. Electric heavy vehicles, electric agricultural equipment, grid-scale storage, industrial heat. The car trend is now one instance of a much wider substitution of electricity for combustion across the whole economy — the electrification of society. That is a megatrend, arrived at by scaling.
Megatrends also form the other way, by convergence. Ageing populations, the cost curve of diagnostics, and the normalisation of remote consultation are three separate trends; together they compose something larger about where care happens and who provides it. Both routes are legitimate. What matters is that a megatrend sits at a level of abstraction where it is nearly always true, nearly always large, and almost never sufficient to decide anything with.
The three levels, and what each is for
| Signal | Trend | Megatrend | |
|---|---|---|---|
| What it is | A specific event that has already happened | An abstraction across many signals | A trend at much larger scale, or several trends converging |
| The test | Can you name who, when, and what resulted? | Can you name the signals underneath it? | Does it hold across sectors and decades? |
| Timescale | Now, and the recent past | Years | Decades |
| What it gives you | Specificity and credibility | Direction | The underlying pressure that makes a future plausible rather than arbitrary |
| If you only have this | A pile of anecdotes pointing nowhere | Consensus you could have bought | Something true, enormous, and useless for a decision |
You need all three, for different reasons. Signals make the work checkable. Trends give it direction. Megatrends explain why a future is plausible rather than invented. The error is not using trends — it is starting there and never going down a level.
Why most organisations start at the wrong level
Almost all organisational futures work begins at the trend layer. There is a straightforward commercial reason: trends are what consultancies sell. A trends report is a sellable object — bound, illustrated, presentable to a board, comfortably abstract. A signal is not a sellable object. It is one strange event that somebody noticed and thought about.
The problem with buying at the trend layer is not quality. Trend reports are often very good. The problem is timing. By the time a trend has been named, evidenced and printed, it is consensus — that is what naming a trend means. And consensus is already priced in. Your competitors have the same report, drew the same conclusions, and are making the same three moves. Being right about a named trend is table stakes; it is not an advantage.
Starting at the signal layer is harder and better. Harder because signals cannot be purchased in a useful form — they have to be noticed, and noticing is a practice rather than a procurement. Harder also because the interpretation is yours and is not underwritten by anybody: nobody has pre-agreed what the event means, which is exactly the condition under which you might see something first. Reading the cultural material where many of the most interesting signals surface is a discipline in its own right, and one we teach separately in Decoding Culture.
The trade is honest and worth stating plainly. Working at the signal layer means holding things whose meaning has not been settled, some of which will turn out to mean nothing at all. That is the price of being early. If that discomfort is intolerable to the organisation, it will keep buying trends, and it will keep arriving at the same time as everyone else.
What makes a signal worth keeping
Specificity is built into the definition. Beyond that, three principles, which we would defend anywhere:
It points somewhere. Something is increasing, decreasing, fragmenting, converging, moving from the margin to the centre. An event with no vector is news, and news is not a signal.
It does not depend on a single source. If the whole case rests on one company's press release about its own product, you are holding marketing. The claim should survive the removal of any one piece of evidence for it.
Thoughtful people disagree about what it means. This is the one most often skipped and the one that does the most work. If everybody who looks at the event draws the same conclusion, the interpretation is already consensus — you have found a trend wearing a date. Disagreement is the indicator that the meaning is genuinely unsettled, and unsettled meaning is where the useful futures work is.
The failure mode: the library nobody uses
The predictable way signal work goes wrong is that collection becomes the point.
Collecting is pleasant, and it is legible. A wall of two hundred signals looks like progress, is easy to show a stakeholder, and never requires anyone to commit to an interpretation. So the collection grows, gets tagged, gets a shared drive and eventually a taxonomy — and converts into nothing. The organisation has built an archive: well curated, regularly added to, referenced by no decision anyone ever made.
The number worth watching is not how many signals you have. It is how many have ever been placed at the centre of something and worked through against a specific focus. Ten signals genuinely used are worth more than four hundred filed. The cost of collection scales; the value does not.
There is a quieter version of the same failure, which is collecting signals as confirmations — accumulating evidence for a conclusion the team reached in month one and has been decorating ever since. Nothing in the definitions above protects against this. Only someone in the room who disagrees does.
Where to take this next
If you take one thing from this, make it the habit of dropping a level. When a trend comes up in a strategy conversation, ask what the specific dated events underneath it are. Either they exist, and you have found the interesting material, or they do not, and you have found out something more useful still.
Our Speculative Design Basics course covers signals and how they feed the rest of the process — consequences, scenarios, and the design work that turns a future into something specific enough to argue with — at your own pace. The wider workflow is described here if you want to see where this stage sits.
Key takeaways
- •A signal is a specific event that has already happened, with a who, a when and an outcome. A trend is an abstraction across many signals. A megatrend is a trend at much larger scale, or several trends converging.
- •The August 2022 Colorado State Fair win by an AI-generated image is a signal — dated, attributable, checkable. "AI is getting good at images" is not.
- •Rising electric vehicle share, shifting consumer preference and manufacturers converting production lines are signals; electrification of the car is the trend; electrification of society is the megatrend.
- •Most organisational futures work starts at the trend layer because trends are what consultancies sell. By the time a trend is named it is consensus, and consensus is already priced in.
- •A signal is worth keeping if it points somewhere, does not rest on a single source, and thoughtful people disagree about what it means.
- •The common failure is hoarding: a growing, well-tagged signal library that never gets used against a specific question. Count signals used, not signals collected.