This project explores whether real-time search behavior can surface early signs of consumer demand pressure before it appears in company-reported metrics.
Using Google Trends data, I tracked behavioral signals around DoorDash — cancellations, price sensitivity, and switching intent — and compared them against earnings call narratives to identify potential gaps between what consumers are experiencing and what companies report.
- Tracks 6 behavioral keywords capturing:
- cancellation intent
- price sensitivity
- switching behavior
- Overlays DoorDash earnings call dates as reference points
- Compares search signal trends against earnings call language
- Identifies divergence between real-time consumer behavior and reported metrics
- "Cancel DoorDash" — remained elevated at 60–80 throughout 2025
- "DoorDash not worth it" — spiked to 100 in April 2026
- "Switch to Uber Eats" — peaked at 100 in March 2026
- Q4 2025 Earnings (Feb 18, 2026) reported:
- record subscribers
- all-time high MAUs
👉 Despite strong reported performance, search signals related to dissatisfaction, cancellations, and switching were rising across the same period.
Interpretation:
Despite reporting record growth, DoorDash shows a visible divergence between reported performance and consumer search behavior — suggesting that aggregated metrics and real-time sentiment may move on different timelines.
Search signals appear to precede or coincide with earnings narratives, suggesting potential value as an early indicator of emerging consumer sentiment shifts.
Consumer sentiment does not always decline when performance metrics do — it often shifts earlier, in subtle ways.
Search behavior may provide one such early-stage signal of changing consumer intent.
- Python — pytrends, pandas
- Streamlit — interactive dashboard with sidebar filters
- Plotly — time series visualization with earnings overlays
- Data Source — Google Trends (US, 12 months, weekly)
pip install -r requirements.txt
streamlit run dashboard/app.py- Google Trends measures relative search interest (0–100), not absolute volume
- A single spike may reflect external events (e.g., viral content), not sustained behavior
- Search data is used as a proxy for consumer intent, not a direct measure of actions
- Correlation is not causation — rising search interest does not prove demand decline
- Earnings calls reflect aggregated, lagging metrics, while search behavior is real-time
Search behavior is a real-time, unfiltered signal — unlike curated earnings narratives. While not a direct measure of demand, it can capture early-stage shifts in consumer intent that may not yet be reflected in aggregated metrics.
The observed divergence highlights an opportunity: Combining behavioral signals with traditional reporting could improve how companies detect and respond to demand changes earlier.
- Incorporate additional data sources (e.g., Reddit, app reviews) for stronger signal validation
- Apply NLP-based classification to refine behavioral intent detection
- Quantify lead-lag relationships between search signals and earnings narratives
- Expand analysis to other consumer platforms (e.g., streaming, BNPL)