Guide · 8 min read
How to find trending topics before they peak
Every trend list shows you what’s already loud. Finding a topic before it peaks is a different job: you’re looking for acceleration, not volume — and you have days, not weeks. Here’s the workflow we run every day, with the measured numbers behind it.
Published Jul 23, 2026 · by the Lolik team
Why trending lists are structurally late
A volume-ranked feed — YouTube Trending, Twitter/X trends, most “what’s hot” dashboards — surfaces a topic at its loudest. By definition, that’s at or after the attention peak: maximum competition, minimum window. You’re not early; you’re the audience.
The useful question is never “what’s big right now?” but “what’s growing faster this week than it was last week — and hasn’t peaked yet?” Those are two different measurements, and only the second one leaves you time to act.
The signal that matters: acceleration below the prior peak
We label every alert in our system with a phase, computed from the signal’s own history — never assigned by hand:
- Early — attention is accelerating, but the signal is still below its previous peak. There’s measurable room left.
- Momentum — the signal is at its all-time high. It’s confirmed, it’s loud, and in our measurements this phase fired too late to catch a single bigger move.
That distinction came from our own failure: our first backtest (July 3, 2026, public in the methodology) showed our original alerts lit up exactly at the peak. We rebuilt the rule around acceleration-below-prior-peak, then re-measured. The result:
Measured, in public
In our own backtest, early-phase alerts kept growing over the next 5 days in 38.8% of cases vs a 5.3% baseline across every tracked signal-day — 7.3× more likely to keep growing than a random tracked trend (n=165, Jul 2–23, 2026). An independent window with zero overlap (Jul 13–23, 2026) reproduced the lift at 8.6× (n=75).
Point-in-time figures, re-measured weekly — full methodology and every call, wins and misses, in the public log.
Read that honestly: 38.8% is not a guarantee — most tracked signal-days (5.3%) go nowhere, and an early alert is a 7.3× bet, not a sure thing. The edge is real and measured; it is not magic.
The 4-step workflow
- 1. Watch week-over-week change, not size. In Google Trends, compare the last 7 days against the previous 7 — a topic at 40/100 interest that doubled beats a topic at 90/100 that’s flat. Our free “Is it still early?” checker runs exactly this test on any topic you type.
- 2. Demand cross-platform confirmation. A real trend leaks across surfaces: search interest rising AND new YouTube uploads getting outsized views AND news coverage starting. One platform alone is often an algorithm artifact. This cross-check is the core of our scoring.
- 3. Check the supply side. Attention growing is half the signal; the other half is how much content already serves it. High demand + thin supply is the arbitrage. Search the topic on YouTube, sort by upload date, and count how many quality videos shipped in the last 72 hours.
- 4. Act inside the window. Median lead in our data: ~2 days from early alert to peak. That means same-week production, not “add it to the content calendar.” If you can’t ship within days, pick evergreen angles on the trend instead of the trend itself.
What this method cannot do
- It cannot tell you which rising topic becomes a monster. Bigger-move stats in our backtest are unpublished because the sample is too small — and we’d rather show you a smaller, true number.
- It cannot rescue slow production. The edge is measured in days; a two-week pipeline eats the entire window.
- It cannot replace judgment about fit. A rising topic your audience doesn’t care about is a rising topic for someone else.
Possibilities, not guarantees — results depend on what you do with them.
Common questions
How early is ‘before the peak’ in practice?
In our measured data, early-phase alerts fired a median of about 2 days before the signal’s peak velocity. That is the realistic window: days, not weeks. Anyone promising weeks of advance notice on short-form trends is describing luck, not a method.
Can any tool predict which topic will explode?
Not honestly, and our own data says so: on bigger moves (1.5× growth or more) our sample is still too small to publish, so we don’t claim it. What is measurable today is direction — whether a topic keeps growing at all — and that’s the number we publish with its sample size.
Do I need paid tools to do this?
No. Google Trends, YouTube autocomplete and manual cross-checking cover the basics for free — and our own free tools run the same live checks without signup. Paid tools buy you time and coverage, not a different physics.
Run the early-vs-peaked check on your next topic
The free tools run on the same live data as the paid radar — no signup, no card. Start there; upgrade only if the timing edge earns it.