How to Build a Trend Radar for Emerging Tech in 2026

Most technologies that matter don't announce themselves. Before the keynote, before the crowded booth and the analyst note, there is a patent family filed by three people in a converted office, a paper read to forty attendees at a specialist conference, and a job ad for someone who knows a material you have never heard of.

A trend radar is the habit of catching those early traces and turning them into something you can act on. It doesn't need a platform or a data team. It needs a defined question, a handful of sources you check on a schedule, and a scoring method honest enough to keep your own enthusiasm in check. Here is how to build one that outlives January.

Pick Two Questions, Not Twenty

Radars fail when they point at "technology" in general. Narrow the beam. Useful framing questions sound like: what replaces graphite anodes in grid storage within three years? Which manufacturing step in our industry is about to get cheap enough to move? What would make our best-selling product irrelevant?

Two or three questions is plenty, plus a horizon — usually 18 to 36 months out. Anything closer is a product decision. Anything further is a hobby. Write both on one page and keep it visible, because every source you add later has to earn its place against those questions.

The Four Signal Layers

Each layer answers a different question, and they disagree in useful ways.

Patents

Patents show where companies spend money on problems they don't want to discuss yet. Follow assignees rather than keywords: pick five or six labs and startups in your field and watch what they file. Espacenet, PATENTSCOPE and Google Patents all support saved searches, and classification codes catch filings that never use your vocabulary. Mind the delay — applications usually publish about 18 months after filing, which makes patents a confirmation signal rather than a heads-up.

Papers and preprints

Preprints move fast and vary wildly in quality. Read abstracts and figures, not conclusions. The strongest tell is repetition: when three unrelated groups attack the same sub-problem in three different ways, the field has found a wall worth climbing. Narrow conferences often beat journals for early signals — a talk at a specialist meeting can precede the published paper by a year.

Launches and hiring

Product launches are late signals, but honest ones. Job postings are better. A company hiring four engineers with thin-film encapsulation experience is publishing its roadmap by accident. So is a changelog that quietly mentions a new hardware partner.

Money

Rounds, grants and government programme awards show conviction. Follow public filings, grant award databases and specialist trade press. One round is interest, not proof. Three unrelated backers funding three similar companies inside six months is a pattern.

A starter set worth bookmarking:

• Two patent search tools with saved alerts, filtered by classification code.
• One preprint server plus one conference proceedings archive in your field.
• A public grant or innovation-agency award database, checked monthly.
• Job boards for three companies you consider early movers.

Set a Cadence You Won't Abandon

Frequency matters less than consistency. A rhythm that survives busy months:

1. Weekly, 20 to 30 minutes. Skim saved searches and your curated feeds. Copy anything interesting into a log with one line on why. Do not analyse yet.
2. Monthly, about an hour. Read the month's entries together. Look for repeats, cluster them, score them, delete the noise.
3. Quarterly, half a day. Rewrite your framing questions, retire sources that stopped producing, and pick one topic for a proper deep dive — read the papers, or talk to someone working on it.
4. Yearly. Rebuild the radar from scratch. If your questions haven't changed, you're collecting, not learning.

Score Signals So You Can Compare Them

Write scores down; memory is generous with things you already like. Rate each signal from 1 to 5 on four axes:

• Specificity. Does it name a material, process or component, or is it a slogan?
• Momentum. Is this the second or third time you've seen it, from different sources?
• Cost curve. Is there a plausible path to cheaper, faster or smaller?
• Pull. Can you name a buyer, a regulation or a bottleneck that creates real demand?

Anything scoring low on specificity or pull stays on the watchlist. Don't delete it; just don't act on it. Rescore everything each quarter so you can see direction, not just position.

Keep the Log Boring

The log is the whole system. One page or one spreadsheet with a few columns: date, signal, source, what would have to be true in 18 months, who would buy it, score. The "what would have to be true" line is what saves you from yourself, because it forces a statement you can later check and find wrong.

A radar is not a prediction machine. It is a record of what you noticed, when you noticed it, and how sure you were at the time.

Where Radars Go Wrong

• Press-release drift. Reading only what companies want read.
• Single-source confidence. One impressive paper is an anecdote.
• Keyword tunnel vision. New work rarely borrows your vocabulary.
• Survivorship. You see the funded companies and miss the quiet ones solving the same problem.
• Collecting without deciding. A radar that never changes a decision is a reading list.

The fix for most of these is human. Suppliers, technicians, recruiters and hallway conversations tell you what a filing never will. Two good conversations outweigh twenty saved links.

Start This Week

Write your two questions. Set three saved searches and one alert. Open a log and put five entries in it, even if they feel thin. Then protect the monthly hour the way you'd protect a meeting with your most important client.

Radar work is unglamorous, and it compounds. Six months of small, repeated noticing will teach you more about where your field is heading than any report you could buy — and you will be able to say exactly when you first saw it coming.

Photo: SpaceX-Imagery / Pixabay