UFOQ.AI

02What is horizon scanning?

Horizon scanning is looking for early signs that something is changing, before the change is obvious.

It's easy to confuse with two things it isn't. Forecasting tries to say what will happen; horizon scanning only records what has already happened and asks what it could lead to. Reading the news tells you what everyone else already knows; horizon scanning looks for the things that haven't made the news yet.

The idea goes back to Igor Ansoff. In 1975 he argued that strategic surprises are usually preceded by faint, early information, and that normal planning systems filter it out because it doesn't look important enough [1]. Many organisations still do exactly that.

Horizon scanning is the deliberate habit of keeping those faint signals, usually called weak signals, long enough to see whether they add up to something.

03A signal is not a trend

If you take one idea from this article, take this one. Much of what goes wrong in horizon scanning starts with mixing the two up.

Horizon scanning collects signals. Trends come later, once enough signals point the same way.

A lot of scanning programmes collect trends and believe they're scanning. But a trend is visible to everyone by definition. If your scanning file is mostly trends, you've built a news digest, and you're already late.

04What counts as a weak signal, and what doesn't

A signal is evidence about the present, not an opinion about the future. These usually qualify:

• A first-of-its-kind product or service
• A draft regulation or public consultation
• A court ruling that reinterprets an existing rule
• An unusual funding round or acquisition
• A patent filing
• A research result that contradicts something you assumed
• An incident or near miss with a new mechanism
• A new player entering your market
• A job advert for a role that didn't exist two years ago

These usually don't:

• Forecasts. They're someone's view of the future, not evidence of what's happening now.
• Opinion pieces with no event underneath them.
• Restatements of something already well known.
• Anything that's in the news every day. If everyone knows it, you gain nothing by knowing it too.

That last point matters more than people expect. Scanning files fill up with things everybody already knows, because those are the easiest things to find.

05The horizon scanning process in eight steps

1. Set a boundary

Decide what you're scanning for, and write down what you're not scanning for. Without a boundary you drown: "scan for anything important" produces a pile of interesting articles and no decisions.

A good boundary is specific: changes affecting how we verify who someone is. A bad one is a topic: technology. The best boundaries come from what you depend on. That is the Anchor step of the Horizon Loop: list what the organisation relies on, and the assumption behind each, before you start scanning.

2. Collect

Go through your sources on a fixed schedule and pull out anything that fits the boundary. Aim for quality over quantity. Five to ten good items per category beats a hundred weak ones; you're picking the best, not trying to catch everything.

3. Record

Use the same format every time. Each record should take about two minutes.

The "if this scaled" field does most of the work. It makes you describe a mechanism instead of bookmarking something interesting. If you can't fill it in, it probably isn't a signal.

4. Screen

Score each signal and keep only the strong ones. Six questions cover it:

• Is it new? Have we seen this before?
• Is it credible? How solid is the source?
• Does it challenge something we believe? And how central is that belief?
• Could it spread? Is there a path from one case to many?
• Would it matter? If it spread, how much would it affect us?
• How soon? Months, years or decades?

The most valuable signals usually score high on challenging a belief and come from sources you don't normally read. Those are exactly the ones your existing filters are built to throw away.

5. Cluster

Group signals that share a cause. UFOQ's working rule: you have a cluster when three or more independent signals share one mechanism.

The word independent matters. Three articles about the same event are one signal, not three, and this is the easiest mistake to make in clustering. Once a cluster has three independent members, give it a name and write down the driver behind it. It's now a theme rather than a pile.

Before a theme moves on, check it against the four tests in our definition of emerging risk. Something has changed; it can reach something you depend on; you haven't yet measured your exposure; and it could change a decision. A theme that fails any test stays in the file or is retired.

6. Interpret

For each theme, answer one question: why is this happening? You've already recorded what happened. What matters now is the driver underneath, because the driver tells you whether it will continue.

7. Hand it over

A theme that stays in the scanning file has achieved nothing. Give it to someone who can act on it, name them, and attach a short list of things to watch. A theme with no owner is still just an observation. In the Horizon Loop, this is where Trace, Size and Own take over.

8. Maintain

Retire signals that are more than two years old and never joined a cluster; they were noise. If you don't retire anything, the file grows until nobody reads it. That is how many scanning programmes quietly die.

06Horizon scanning template: a signal record, before and after

The difference between a useful scanning file and a reading list shows up in the records. Here's the same story logged twice.

Before: a bookmark

It's a topic, not an event. There's no date that means anything, the source is someone's view, and the two fields that matter are empty because there's nothing specific to say.

After: a signal

The second record names an event, a date and a source someone else can check. More importantly, it says which belief is under threat and what would happen if this became common. That's what lets it join a cluster later, and what gives an owner something to act on.

07Horizon scanning in practice: the signals before the Strait of Hormuz closure

On 2 March 2026, an adviser to the commander of Iran's Revolutionary Guards declared the Strait of Hormuz closed [5]. To many organisations it felt sudden. The signals had been on public record for about eight months.

Closure threats were not new, so another threat on its own would have scored low on "is it new?" What changed in June 2025 was that different actors started treating closure as a real option. A team scanning with a boundary like disruptions to our imported inputs could have logged these:

These pass the independence test: three different actors, three separate decisions, one shared mechanism. Closure was moving from rhetoric to something people were planning around. The driver, the answer to "why is this happening?", was open military conflict. Israel began strikes on Iran on 13 June 2025 [8], and the United States struck Iranian nuclear sites on 22 June [9].

The "if this scaled" field is where the value lies. For a food producer, it might have read: if the strait closes, Gulf fertiliser stops reaching us on normal routes. Check how much of our supply comes from there. That's a question you can answer in a week, and one that turned out to matter. When the closure came, the United Nations Conference on Trade and Development (UNCTAD) estimated that about a third of the world's seaborne fertiliser trade passed through the strait [10].

None of these signals predicted the date. What they offered was eight months to find out where your fertiliser came from, line up alternatives, and name someone to watch transit counts and supplier lead times. The Horizon Loop follows the same case through every step, to an owned and dated call.

08Where to look: five tiers of sources

Sources differ mostly in how far you can trust them and how early they are. Ranked by trust:

A few of these deserve more attention than they get:

• Court cases. A conviction tells you a method worked well enough to be worth prosecuting, and the judgment often describes exactly how it worked.
• Job adverts. When companies start hiring for a role that didn't exist before, someone decided it was needed.
• Conference agendas. Compare this year's sessions with last year's. The new sessions are a ready-made scan.
• Conversations. Ten good conversations a year are worth more than most subscriptions. Ask one question: what have you seen in the last six months that surprised you?

With vendor research, read the method page first, separate what was measured from what was projected, and ask what the vendor is selling. A report that concludes the vendor's own product is urgently needed isn't independent evidence.

Source habits

• Record the type of source, not just the link. Regulator, court, academic, vendor or press. You'll want to weigh them differently later.
• Date the event, not the article. One event has one date, however many times it's written about.
• Don't let a single secondary source carry a material claim. If it's going to drive a decision, find the original.
• Prefer the boring source. The findings that change minds usually come from dull datasets, not headline cases.

09How it runs: a simple cadence

Horizon scanning works as a routine, not a project. Most of the effort is small and regular.

Start with a shared spreadsheet. As a rule of thumb, buy horizon scanning tools only when the spreadsheet visibly breaks, usually past a few hundred items or five contributors. Buying a platform first is the classic mistake: it gives you more volume without more judgement, and judgement is the part that's scarce.

10From signal to risk: where horizon scanning fits

Horizon scanning is the first part of a longer path. A signal becomes a cluster, the cluster passes the four tests, and it becomes an emerging risk once someone owns it.

1. Signal
2. Cluster
3. Emerging risk: passes the four tests, then traced, sized and owned
4. Known risk that can be measured and priced

Horizon scanning owns the first two stages. The Horizon Loop takes a cluster through the third. It traces how the change would reach you, sizes it with a stated confidence level, and names an owner. From there, normal risk management takes over once the risk is measurable and priced.

The international guidance on emerging risk points the same way. Published summaries of ISO/TS 31050 describe a repeating cycle of framing, collecting and analysing, interpreting, and turning the result into intelligence for decisions [11]. Horizon scanning is how that cycle is carried out in practice.

Themes are also good raw material for scenario planning. Each driver you've written down is a question about which way things could go, and a handful of them is enough to build a small set of scenarios.

Scanning vs watching

The two are often confused, and you need both.

Scanning is open. You don't know exactly what you're looking for, so you cast wide and see what turns up.

Watching is targeted. You've already decided what matters, picked specific indicators, and you check them on a schedule.

Scanning finds the issue; watching tells you whether it's arriving. A scanning programme with no watchlist never turns into action, and a watchlist with no scanning only tracks what you already thought of.

11Why horizon scanning programmes fail

The failures are predictable, and most come from treating scanning as reading rather than as a process with an end point.

• No boundary. You drown in interesting material and make no decisions.
• Collecting what everyone knows. The file fills up with news, and nothing in it is early.
• Treating volume as success. "We logged 400 signals this quarter" isn't an achievement. It's a warning sign.
• No owner at the end. Things get spotted, discussed and forgotten.
• Never retiring anything. The file becomes unreadable and people stop opening it.
• Buying the tool first. You get more input with the same judgement, and the signal-to-noise ratio gets worse.
• One person's hobby. When they leave, it stops. It needs to be part of someone's role, not their enthusiasm.
• Confusing signals with trends. In our experience, the most common failure of all. If everything in your file is a trend, you've built a news digest.

12Six tests of a working horizon scanning programme

Give a new programme a year, then check it against these six questions:

1. Did we spot at least one thing before it was widely discussed?
2. Did our risk register change because of something we found?
3. Does at least one theme have a named owner outside the scanning team?
4. Can we show the source and date for every claim we've made?
5. Have we retired roughly as much as we've added?
6. Could someone who wasn't involved read a scan record and understand it?

If none of these is true after a year, the programme isn't working, and no amount of extra tooling will fix it.

For what a good entry looks like once a theme reaches the risk register, see the register entry in the Horizon Loop.

13Conclusion

Horizon scanning isn't about predicting what happens next. It's about noticing what has already happened, while it's still small, and deciding whether it matters to you.

The method fits in a few lines. Collect things that actually happened. Write down what each one would mean if it spread. Group the ones that share a cause. Hand those groups to someone who can act. Retire the rest.

Everything else is detail, and the detail mostly exists to stop you drifting back into reading the news.

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

Further reading

UFOQ

• The Horizon Loop: UFOQ's framework for emerging risk

Standard

• ISO/TS 31050:2023, Risk management — Guidelines for managing an emerging risk to enhance resilience.

Free sources worth subscribing to

• Swiss Re SONAR
• CRO Forum Emerging Risk Initiative
• Lloyd's insights
• World Economic Forum Global Risks Report
• Policy Horizons Canada
• OECD Strategic Foresight

Background

• Hiltunen, E. (2008), "The future sign and its three dimensions", Futures. Separates the sign itself, the change underneath it, and how the receiver interprets it. A signal can be weak in any of the three.
• Hines, A. & Bishop, P. (2013), "Framework foresight: Exploring futures the Houston way", Futures 51.