Same price,
smaller bag.
A Gemini agent that checks documented size cuts and shows the price per ounce you now pay, with sources.
Run the agent's playbooks yourself
The real agent has Gemini plan Elasticsearch queries. This explorer runs the same playbooks in your browser over the same documents, so no API key is needed. Each result lists the tool calls the agent makes.
Demo data: 34 products curated from public shrinkflation reports (FullCarts, Mouseprint, ShrinkWatch, Capital One Shopping research). Sizes and dates follow the reports; prices are representative, not live quotes. Reddit-only reports are marked unverified. Methodology.
Price per unit is the sticker price divided by the size. At $4.99, 9.75 oz works out to $0.51 an ounce. At 9.25 oz the same $4.99 is $0.54 an ounce. That change in price per ounce is the hidden increase.
This is one example from 29 documented size cuts in the demo data, not a typical result. The average across those cuts is in the figures below.
What the data shows
- ~33%of grocery items affected by shrinkflation (Capital One Shopping research, 2025)
- +12.2%average hidden price-per-unit increase across the 29 documented size cuts in this dataset
- +25%worst case here: Haribo Goldbears, 5 oz → 4 oz at the same $1.99
Six investigations, one agent
With the full agent you ask in plain English and Gemini picks the playbook. Pick one to run it above.
Product lookup
Has my Doritos bag gotten smaller?
Size history, price-per-unit trend, verdict.
Run it → 02Brand investigation
Show me all shrinkflation from Frito-Lay
Products ranked by shrinkage, via an ES|QL aggregation.
Run it → 03Category analysis
Which category has the worst shrinkflation?
Average shrinkage per category and the worst three products.
Run it → 04Shopping list audit
Check these: Doritos, Charmin, Tide, Oreos
Each item checked, plus the total hidden cost.
Run it → 05Trend dashboard
What are the worst trends this year?
The ten biggest cuts and a count per year.
Run it → 06Data discovery
What data do you have?
Lists the indices and reads their mappings. Needs the full agent, so see how it works.
How it works →One Gemini agent, five read-only MCP tools
Built with Gemini Flash on Google ADK, connected to Elasticsearch through the official Elastic MCP server. A product question takes three tool calls: search for the product, pull its history with ES|QL, then look up the documented size change.
USER ──► GEMINI FLASH (ADK agent)
│ plans steps, picks tools
▼
ELASTIC MCP SERVER
├─ search find the product
├─ esql price-per-unit analytics
├─ list_indices discover data
├─ get_mappings understand schema
└─ get_shards shard status
│
▼
ELASTICSEARCH (products · product_history · shrinkflation_alerts · categories)
Running the full agent needs an Elasticsearch cluster and a Gemini key; see the quickstart.