Do I need an API key, and why does it keep failing?

No key is needed, but that is also the weakness: the tool uses pytrends, an unofficial scraper of Google Trends, which Google rate-limits aggressively. Single keyword mode fetches a Google NID cookie first to reduce blocks and both modes retry (1 to 10 times, default 3, with backoff). If you still get no data, wait a while before retrying; there is no paid tier to fall back on.

Why is the most recent week missing from the actuals?

Google Trends flags the latest incomplete period as partial, and the tool drops every row where isPartial is true before training. The model therefore only ever sees complete weeks, which stops a half-finished week dragging the forecast down.

What time granularity does the forecast use?

Weekly. The NeuralProphet model is fitted with weekly frequency (freq='W') on interest-over-time data pulled at pytrends' default window of roughly the past five years, and it forecasts 1 to 104 weeks ahead (default 52). Values are Google Trends' relative 0 to 100 interest index, not absolute search volumes.

How long does batch mode take and what happens when a keyword fails?

Keywords are processed one at a time with a configurable delay between requests (1 to 30 seconds, default 5), plus model training per keyword, so large lists take a while; duplicates and blank rows are removed first. Keywords that fail or return no data are collected into an issues list shown at the end, and their worksheet just says 'No data available'. Each worksheet is named after its keyword with special characters stripped and truncated to Excel's 31 character limit.

Back to Tools

Legacy Tool - Unreliable API

This tool relies on an unofficial API (pytrends for Google Trends) which is frequently rate-limited and unreliable. The tool may fail or return incomplete data. Use with caution.

Google Trends Forecasting

Use cases

Planning content calendars around seasonal peaks Timing product launches and campaigns Identifying emerging vs declining topics Batch forecasting for keyword research

Pulls Google Trends interest-over-time data via pytrends and fits a NeuralProphet model (weekly frequency, daily seasonality enabled) to forecast future search interest.

Single keyword mode renders an actual vs predicted matplotlib chart with CSV download.

Batch mode deduplicates uploaded keywords and writes one Excel worksheet per keyword with an embedded forecast chart via xlsxwriter.

Streamlit App

Platform

Browser-based (no installation required)

Input

Single keyword, or CSV/TXT keyword file (encoding auto-detected via chardet)

Forecast weeks, language/region, and retry settings in the sidebar

Batch mode: delay between requests (1-30 seconds, default 5)

Output

Single mode: CSV with date, actual, and predicted columns. Batch mode: Excel workbook with one worksheet and embedded scatter chart per keyword.

Launch App View Source

Features

Single keyword or batch CSV processing
NeuralProphet weekly model with daily seasonality
Forecast window 1-104 weeks (default 52)
Configurable API retries (1-10, default 3)
18 Google Trends language/region options
Batch mode: one Excel worksheet per keyword with embedded chart

How to use

1 Choose Single Keyword or Batch Upload mode
2 Enter a keyword or upload your CSV and select the keyword column
3 Set forecast weeks, region, and retries in the sidebar
4 Click Submit to fetch trends data and train the model
5 Review the actual vs predicted chart
6 Download the forecast (CSV for single, Excel for batch)

Frequently asked questions

Do I need an API key, and why does it keep failing?: No key is needed, but that is also the weakness: the tool uses pytrends, an unofficial scraper of Google Trends, which Google rate-limits aggressively. Single keyword mode fetches a Google NID cookie first to reduce blocks and both modes retry (1 to 10 times, default 3, with backoff). If you still get no data, wait a while before retrying; there is no paid tier to fall back on.
Why is the most recent week missing from the actuals?: Google Trends flags the latest incomplete period as partial, and the tool drops every row where isPartial is true before training. The model therefore only ever sees complete weeks, which stops a half-finished week dragging the forecast down.
What time granularity does the forecast use?: Weekly. The NeuralProphet model is fitted with weekly frequency (freq='W') on interest-over-time data pulled at pytrends' default window of roughly the past five years, and it forecasts 1 to 104 weeks ahead (default 52). Values are Google Trends' relative 0 to 100 interest index, not absolute search volumes.
How long does batch mode take and what happens when a keyword fails?: Keywords are processed one at a time with a configurable delay between requests (1 to 30 seconds, default 5), plus model training per keyword, so large lists take a while; duplicates and blank rows are removed first. Keywords that fail or return no data are collected into an issues list shown at the end, and their worksheet just says 'No data available'. Each worksheet is named after its keyword with special characters stripped and truncated to Excel's 31 character limit.

Want me to run this for you?

I run this tool as a managed service, or build something custom around your data. You get the insights without touching the code.

Tell Me What You Need

BCG Matrix GeneratorScriptReporting

Generate BCG matrix visualisations from GA landing page reports with cascading folder analysis.

Resolution Screenshot ToolAppReporting

Capture screenshots at GA visitor resolutions using Pyppeteer async browser automation.

Core Update AnalyserAppReporting

Analyse Google core update impact by URL folder using Ahrefs data.

Share of Voice CalculatorAppReporting

Calculate competitive share of voice from Ahrefs traffic data with Excel chart export.

Delta Audit ToolAppReporting

Detect significant traffic changes in GSC data to identify Google update impacts.

SERP Snippet PreviewBrowserContent

Preview how your page will appear in Google search results.

Need something built for your business?

This tool started as bespoke client work. I build custom scripts, data pipelines, and full apps for SEO and product data problems that off-the-shelf tools don't solve.