UFOQ.AI

02What are weak signals?

A signal is one thing that happened which suggests something might be changing. Each part of that sentence does work.

One thing. A single event: one occurrence, in one place, at one time. Not a pattern, a direction or a summary.

That happened. It's in the past tense. A signal is evidence about the present, not a claim about the future. It has a date and a source, and someone else could go and check it.

Suggests something might be changing. It puts pressure on an assumption you hold and raises a question about it.

Put together, a signal is the smallest unit of evidence about change. Weak signals are simply signals that are easy to miss or dismiss, usually because they're small, early or come from somewhere unfamiliar.

03Weak signals vs trends, drivers, megatrends and wild cards

Signals and trends are easy to confuse, and the confusion runs through everything that follows. Here's the whole family, from smallest to largest:

A common failure in scanning programmes is to collect trends and believe they're signals. The difference matters for a simple reason: by the time something is a trend, everybody can see it. A trend is a measurement, and measuring needs history. If you have that much history, you're not early; you're informed, like everyone else.

Drivers of change, megatrends and wild cards all matter for strategy and scenario planning. But for spotting change early, signals are where the advantage is, precisely because they're too small to be obvious.

A cluster becomes an emerging risk when it passes the four tests in our definition. Something has changed; it can reach something you depend on; you haven't yet measured your exposure; and it could change a decision.

04Four things a signal is not

A forecast

"Experts predict the market will double by 2030" isn't a signal. It's somebody's opinion about the future. A prediction tells you what one person thinks; an event tells you what the world did.

Forecasts trip people up because they're everywhere and they feel like information about the future. Mostly they're information about forecasters.

An opinion piece

An article arguing that something is important isn't a signal, although the event it describes might be. Strip out the commentary. If there's a dated occurrence underneath, that's your signal. If there isn't, there's nothing there.

The news

If it's in the news every day, it isn't a signal for you. It may be true, important and worth tracking, but you gain nothing by noticing it, because everyone else has too.

This is the hardest discipline in scanning, because news is the easiest thing to find. Scanning files fill up with big stories that everyone already knows, and the programme slowly dies of it.

A list of scary things

"AI, climate, geopolitical instability." Anyone can write that list, and it tells you nothing. A real signal is specific enough that you could turn out to be wrong about it.

05Weak signals examples: what good signals look like

Some real and typical examples make this concrete.

A police case. In February 2024, Hong Kong police described a fraud ring that had used eight stolen identity cards to make 90 loan applications and register 54 bank accounts [1]. Deepfakes fooled facial recognition checks at least 20 times. One case, one date, one source. It tells you a method worked well enough for someone to build a business on it.

A regulator's alert. On 13 November 2024, the US Financial Crimes Enforcement Network (FinCEN) issued an alert on fraud schemes using deepfake media [2]. It asked banks to mark related suspicious activity reports with the key term "FIN-2024-DEEPFAKEFRAUD". A reporting code means data will build up, which means the risk will become measurable, and pricing and controls will follow.

A job advert. A bank advertises a role that didn't exist two years ago. Someone made the case for that budget, and that decision is the signal.

A conference agenda. Compare this year's session list with last year's. The new sessions are a scan someone else has put together for you.

A patent filing. Companies patent what they intend to build.

A funding round in an unexpected place. Money moves before capability does.

A near miss. Something fails and nothing breaks, so everyone relaxes. That's a signal too, and an especially dangerous one, as we'll see below.

What these have in common is that they're small, specific, dated and checkable. None of them would make a headline.

06How to judge a signal: six criteria and one test

Score each signal against six questions and keep the strong ones. (Our horizon scanning article shows where this sits in the wider process.)

• Is it new? Have we seen this before in our area? If it's familiar, it's probably a trend you're late to.
• Is it real? Is the source solid? Is it an event, or someone's description of a feeling?
• Does it challenge something we believe? This is the most important question. The more central the assumption it puts pressure on, the better the signal.
• Could it spread? Is there a plausible path from one case to many? If you can't describe the mechanism, you have a curiosity, not a signal.
• Would it matter? If it did spread, how much would it affect us? Interesting and relevant aren't the same thing.
• How soon? Months, years or decades? All three are valid; you just need to know which.

The single best test

If you only ask one question, make it this one: what would have to be true for this to matter?

It forces you to describe a mechanism instead of a feeling, and the answer usually tells you straight away whether to keep the signal. If the answer is "a lot of unlikely things", let it go. If it's "two things that are already happening", you've found something.

07Why weak evidence is the point

A good signal is weak evidence, by definition. This is the heart of the discipline, and it is the hardest part to accept.

If the evidence were strong, it would be a trend. If it were obvious, it would be news. If it were certain, it would already be on your risk register.

So you're always working with material that falls short of your organisation's normal standards of evidence. Everything about a signal invites dismissal: one case, an unusual source, unclear relevance, no numbers. That isn't a flaw in the signal. It's what early looks like.

The skill, then, isn't finding strong evidence sooner. It's handling weak evidence well: recording it properly, holding it without overreacting, and waiting until independent pieces point the same way.

08Weak in three different ways

The idea of weak signals goes back to Igor Ansoff, who argued in 1975 that strategic surprises are preceded by faint early information [3]. The Finnish futures researcher Elina Hiltunen later refined it in her "future sign" model [4]. It shows that "weak" can mean three separate things that people tend to blur together:

In our experience, the third is the most common. Often the signal wasn't weak at all. It was loud and ignored, because it came from an unfamiliar source, contradicted a senior commitment, or fell outside anyone's job.

One habit helps a great deal: keep the observation separate from your interpretation of it. Record what happened in one field and what you think it means in another. When your interpretation turns out to be wrong, the observation is still good, and someone else can use it.

09The most dangerous signal: the near miss

Something happens, nothing breaks, and everyone concludes they're fine.

In October 2010, floods spread across 38 of Thailand's 76 provinces and killed at least 94 people [5]. Ayutthaya, just north of Bangkok and home to some of the country's biggest industrial estates, was among the provinces badly hit. Yet there was no global supply shock.

A year later came Thailand's worst floods in 50 years [6]. Seven major industrial estates were flooded [7], in places to a depth of about three metres [8]. Thailand made about a quarter of the world's hard drives, and the floods disrupted that production [9]. In December 2011, Intel, which makes chips rather than drives, cut its quarterly revenue outlook by about USD 1 billion [10]. The reason it gave: hard-drive shortages led computer makers to buy fewer processors.

Anyone raising flood risk in late 2010 would probably have heard: we've just had serious floods and the factories were fine. It sounds reasonable, but it gets the lesson backwards. The factories weren't fine because they were protected; they were fine because they were lucky.

An event that causes no damage looks like proof that a control works. More often it's proof of luck. Treat near misses as warnings, not as tests you passed.

10Five common mistakes

1. Counting coverage as evidence. Three articles about one event are one signal. It is an easy mistake to make, and it corrupts every count that follows.
2. Logging the date you found it. Record the date the event happened. Something from 2023 that you found last week is a 2023 signal, and its age tells you something.
3. Only reading sources you like. Your regular reading is where your existing assumptions came from. The signals that matter most tend to come from somewhere you don't normally look.
4. Collecting the spectacular. The dramatic case gets the attention, but the boring one often turns out to be the bigger problem. A file full of exciting signals will mislead you about where to spend.
5. Never throwing anything away. A signal from three years ago either became a trend or was noise. Retire it. A file that only grows is a file nobody opens.

How to know you're doing it right

After a few months, your file should look like this:

• Most items come from sources you didn't read before you started.
• You could hand any record to a stranger and they'd understand it.
• Several items make you slightly uncomfortable.
• You've thrown away more than you've kept.
• At least one item has changed somebody's mind.

And the real test: you can name an assumption you used to hold and no longer do. If nothing has changed your mind, you're collecting confirmation, not signals.

11Conclusion

A signal is one thing that happened, with a date and a source, that makes you question something you assumed. It isn't a forecast, an opinion, a trend or the news.

It will always be weak evidence, because that's what early looks like. In our view, the organisations that get ahead of emerging risks aren't the ones with better data. They're the ones that record weak evidence carefully, hold it without overreacting, and act when it starts to add up.

If there's one habit to take away, it's this: collect the mechanism, not the headline.

UFOQ collects signals like these every week and traces them through their first-, second- and third-order impacts. Read the latest Weekly Risk Foresight.