Validate demand before you build

The problem. The idea filter’s first gate is proven demand, and the honest way to check it is to see whether anyone’s searching for the thing at all. The usual way to find out is a slow afternoon in Google Trends: type a term, set the range, screenshot the chart, add the next term, repeat, then eyeball five images and guess. It doesn’t scale past one idea, and the guessing is the part that fails you — a flat line is easy to talk yourself into reading as growth. Trends has no official API, so the manual browser dance became the default. It doesn’t have to be.

The play. Have Claude write a short script that pulls the same data headlessly and hands you a table instead of screenshots: interest over time for each term, the rising related queries (where the breakout energy is), and a one-line read on whether the category is growing, flat, or declining. Point it at three things — your idea, the category it sits in, and the closest competitor — and read the shape in one pass. This is the evidence that feeds the idea filter’s Gate 1. Run it before you spend a week building on a hunch.

Setup (5 min). The tool is a thin wrapper around pytrends, an open library that reads Google Trends. You don’t have to write it from scratch or know Python — ask Claude to build it and run it for you:

Write a small Python script using pytrends that takes a comma-separated list
of search terms, a geo, and a timeframe, and prints: interest-over-time per
term (mean, peak, recent, and a rising/flat/declining verdict), the top rising
related queries, and a one-line red-flag check (is the whole category
declining?). Then run it for these terms: [your idea, the category, a competitor].

Claude installs the dependency, writes the script, runs it, and reads the result back to you. If you’d rather run it yourself: pip install pytrends pandas, then the command below.

The command:

python trends.py --terms "your idea, the category, a competitor" --geo US --timeframe "today 5-y" --related

Up to five terms per run — the same limit as a Google Trends comparison. Use --geo "" for worldwide, and a shorter --timeframe "today 12-m" to catch a recent surge.

What you get. A markdown table you can paste straight into your notes: each term with its mean, its peak, its recent level, and a trend verdict (Rising, Flat, Declining) computed from the first third of the range against the last. Under it, the rising related queries — often the most useful part, because a breakout query is a demand signal before it’s a headline. And a red-flag line: if every term is in multi-year decline, that’s a reason to stop, not a detail to skip.

What to watch for.

• Relative, not absolute. Google Trends reports interest on a 0–100 scale within your comparison, not real search volume. A term can be “rising” and still be tiny. For actual monthly volumes, a keyword tool (Ahrefs, Semrush) is the next step up.
• The library is unofficial. pytrends reads Trends’ internals, so Google can rate-limit it (wait and retry) or break it when they change the page. If it stops working, trendspy is a drop-in replacement.
• Search is blind to a category that doesn’t exist yet. If you’re building something genuinely new, nobody searches for it by name — a flat line here isn’t a no, it’s the wrong instrument. Validate that kind of idea with conversations and a paid test, not Trends.
• For an app idea, check the App Store too. Apple’s search-hints endpoint shows what people type into the App Store, which is a closer signal than web search for something you’ll ship as an app.

Make it repeatable. This isn’t a one-time check; it’s Gate 1 of the idea filter, mechanized. Keep the command around and point it at every idea before it earns a week of your time. When the demand read is honest and the trend isn’t dying, take the result into the idea filter, clear the rest of its gates, and — on a Keep — start the roadmap.