How we find, qualify and weigh foresight voices
Most organisations see the future they expect. Change that matters usually starts at the edges, where it is faint and easy to dismiss. Outside voices can act like the many small lenses of an insect's compound eye: each one sees a different part of the periphery. This page explains how we choose those voices, and how we test what they tell us.
Why outside voices
Studies of corporate foresight suggest that only a minority of organisations can reliably spot and act on weak signals, and that internal scanning tends to stay inside one familiar field. Paul Schoemaker and George Day describe the answer as peripheral vision: sensing nodes at the edge of the organisation, close to where change appears.
Journalists, researchers, practitioners and community leaders can play that role. But following everyone quickly drowns an organisation in noise. So we apply explicit criteria, look for voices who disagree with each other, and track what they publish in the open.
Ten weak signals do not outweigh one strong contradiction. We look for evidence that could prove a signal wrong, not only evidence that fits.
Nine kinds of signal
A useful foresight voice does more than give opinions. Over time, their work covers several of these nine dimensions, from early hopes and problems to concrete next steps. Each card shows how many voices in our directory regularly share it.
Aspirations
What people hope for or fear, before it becomes market demand or policy: the "pull" of the future.
voicesProblems
Frictions and bottlenecks that incumbents have not noticed yet, framed by the cost of doing nothing.
voicesOpportunities
"Pockets of the future in the present": places where something new already works, and why the window is open now.
voicesTrends
Sustained changes in direction over years, kept apart from short-lived fads, with the conditions that could bend them.
voicesWeak signals
Early, faint evidence from the edges. A good one often makes experts laugh, deny it, wonder, or admit they never heard of it.
voicesNeeds for adaptation
Where old assumptions, rules and business models get in the way: the "weight of the past" that has to change.
voicesSolutions
New methods, designs or ways of working that could solve a known or emerging problem.
voicesInvestment opportunities
Where capital could help, judged on real evidence of use rather than stories alone.
voicesNext steps
Concrete actions that link a long-term vision to what can start now: pilots, partnerships, or things to stop doing.
voicesFour kinds of foresight voice
A balanced set of voices includes all four. Each sees a different part of the periphery, and each needs a different kind of check.
Lead-user practitioners
People who face tomorrow's needs today, and build their own workarounds. Eric von Hippel's lead-user research shows they often innovate before any market exists.
in the directory →Investigative journalists and researchers
People who check claims against primary evidence and publish their sources, so others can check them too.
in the directory →Thought leaders and analysts
People who connect findings across fields into trends, scenarios and frameworks, and who update them when the evidence changes.
in the directory →Community leaders and living-lab catalysts
People who bring citizens, researchers, business and government together to try new approaches, often in one place first.
in the directory →What we look for, and what rules a voice out
Five tests, each based on research about where good foresight comes from and how people judge whether information can be trusted.
| Test | What we look for | Red flags | Based on |
|---|---|---|---|
| Early sight of change | Meets problems months or years before the mainstream, and builds workarounds, code or new practices. | Talks only about problems that already have widely adopted solutions. | Lead-user theory (von Hippel) |
| Bridging fields | Translates between research, business, policy and communities, and spots where they meet. | Stays inside one field and repeats its consensus. | Boundary spanning and knowledge brokerage |
| Openness to evidence | Separates facts from assumptions, updates views, and seeks out people who disagree. | Locks onto one story early and filters out anything that does not fit. | Need for closure (Kruglanski); futurist traits (Hiltunen) |
| Credible source | Real expertise or first-hand proximity; discloses sponsors, stakes and beliefs. | Borrowed authority; hidden commercial or political interests. | Epistemic vigilance (Sperber, Mercier) |
| Sound, open distribution | Publishes regularly in the open, with links to primary data, leaving a trail others can audit. | Screenshots without context, sensational headlines, circular citations. | Extended epistemic vigilance (Bielik, Krell) |
The live data on each voice card (how often they publish, their latest piece) is there to support the last test. It is not a ranking of quality.
How we find voices
Rare voices are not found by keyword searches or follower counts, which mostly surface the already famous. We combine five methods.
Ask who is further ahead (pyramiding)
We ask each voice who they learn from: someone who feels the problem more sharply or has already built a workaround. Following these referrals step by step reaches rare experts with far less effort than broad surveys; research by von Hippel, Franke and Prügl, and later Stockstrom and colleagues, found it takes a fraction of the effort of screening. Five to twenty conversations per topic are often enough.
We start from visible people such as bloggers and forum moderators, stop a chain when two referrals in a row add nothing, and follow referrals to organisations, repositories or other fields too.
Map who follows whom
People with shared specialist interests follow similar accounts. Starting from a small set of trusted seed accounts per frontier, audience overlap reveals clusters of related voices, and the "bridges" between clusters, who often see convergence first.
Balance the portfolio
We mix voices with little audience overlap, to widen what we can see, with a few bridges between them, so that a signal seen by several independent voices can be checked. Voices who disagree are included on purpose.
Track open activity
Public feeds, repositories, papers and job moves show where practitioners are actively building. Data helps counter the bias of personal networks, which tend to find people like ourselves.
Be worth finding
The best voices guard their time. We publish our method and our signals openly, and invite voices to scanning roundtables, weak-signal galleries and living-lab sessions in the Summanity and Humanity Solutions family.
From a scanning hit to a decision
Once voices are sharing, the risk shifts from too few signals to too much noise. Every contribution travels through four steps.
Capture and triage
Each hit gets a short title and a description of at most 150 words that keeps the raw finding (with its source) apart from the interpretation. We tag it by theme and check three things: is it novel, plausible and timely? Then we score it 1 (confirms what is known), 2 (adjusts a known driver) or 3 (could transform the system).
Based on Policy Horizons Canada and framework foresight.
Read beneath the headline
A signal has three parts: the thing observed, the issue it points to, and someone's interpretation (Hiltunen's "future sign"). We look at the interpretation in layers: the headline, the systems behind it, the worldview that keeps those systems in place, and the myths and metaphors underneath.
Based on Causal Layered Analysis (Inayatullah) and the Three Horizons framework (Sharpe).
Score weak and strong signals separately
Weak signals would always lose against mature trends, so they get their own matrix: how fast research and experimentation move, against how broad the potential is. Exploration, market consolidation, technology consolidation, or "transition ready". Stronger signals are scored on maturity and momentum and placed in one of five action zones:
01 RecogniseMonitor, no investment yet02 ObserveDesk research on uses03 TrialPilot in real conditions04 PlanRoadmap and budget05 ImplementIntegrate at scaleBased on the ENISA Technology and Innovation Radar. Our attention radar uses a simple, public version of the momentum idea.
Try to prove it wrong
Before anything is acted on, we list competing explanations: a real shift, a temporary hype cycle, or a promotional campaign. Evidence that fits every explanation is set aside, because it tells us nothing. We look for contradictions, check how much the conclusion depends on one report, and agree signposts that would change our mind.
Based on Analysis of Competing Hypotheses (Heuer), with Subjective Logic (Jøsang).
The same signal means different things in different places
As William Gibson observed, the future is already here but unevenly distributed. A weak signal in one region can be an everyday trend in another, and a solution that works in one place may need a different form elsewhere. So we record where every signal comes from and who sees it:
- Signals and solutions record the place they emerge and its world region, so you can filter the directory by region.
- Voices record where they are based, which regions their signals are about, and the languages they publish in.
- Live posts are tagged with the countries and regions they name; when a post names no place, we use the region its voice covers, and show that differently.
- Attention is compared across twelve Wikipedia languages, as a rough map of where each frontier is gaining ground.
When you weigh a signal, ask: where is this true, for whom, and where might it arrive next?
Eleven world regions
We use ten regions plus "Global": Sub-Saharan Africa; Middle East & North Africa; Europe; Central Asia & Caucasus; South Asia; East Asia; Southeast Asia; Oceania & Pacific; Latin America & Caribbean; North America.
What the live data can and cannot tell you
- Publishing cadence shows that a voice is active and leaves a public trail. It says nothing about whether they are right.
- Audience size shows reach. Research on social networks finds that influence on what spreads depends more on position and interaction than on follower counts.
- Wikipedia attention shows which issues the public is looking up. By the time a topic gets steady traffic, it is usually a trend, not a weak signal.
All live data is fetched and cached by our server from public sources. Visitors' browsers do not contact these services.
Voices by the region they cover
We keep adding voices from under-represented regions and languages, especially Central Asia, the Pacific and non-English sources. Nominate a voice.
Sources behind this method
- Schoemaker, P. and Day, G. How to Make Sense of Weak Signals. MIT Sloan Management Review.
- Hiltunen, E. The future sign and its three dimensions. Futures.
- von Hippel, E. Lead Users: A Source of Novel Product Concepts. Management Science. See also Lead user.
- von Hippel, E., Franke, N. and Prügl, R. Pyramiding: Efficient search for rare subjects. Research Policy.
- Stockstrom, C. et al. Identifying valuable users as informants for innovation processes. Research Policy.
- Hyysalo, S. et al. Intermediate search elements and method combination in lead-user searches.
- Inayatullah, S. Causal Layered Analysis 2.0. See also Causal layered analysis.
- Three Horizons (Bill Sharpe) and the H3Uni community.
- Policy Horizons Canada. Foresight training manual, module 3: Scanning.
- ANTICIPATE. Signals, trends and megatrends: the difference.
- The ENISA Technology and Innovation Radar methodology, explained.
- Pope, S. and Jøsang, A. Analysis of Competing Hypotheses using Subjective Logic. See also Analysis of competing hypotheses.
- Bielik, T., Krell, M. et al. Developing and evaluating the extended epistemic vigilance framework.
- Kruglanski, A. on the need for closure: Closure (psychology).
- BackTrend: Evaluating scientific weak-signal prediction. arXiv, 2026.
Put the method to work
Follow the voices, watch the live radar, and share the signals and solutions you notice.