Research note
The Feature Rate Is a Retail-Technology Signal
How weekly promotion data can inform pricing workflow design without pretending that a grocery circular is an independent-shop forecast.
Pricing signal
Participation and intensity can diverge
The reported weekly movement separates market breadth from promotional activity.| Measure | Value | Context |
|---|---|---|
| Feature rate | 14.0% | down |
| Activity index | 9.3% | up |
The Apparent Contradiction in Beef Feature Data
A weekly grocery report can present two numbers that seem to pull in opposite directions. According to USDA AMS Weekly Grocery Store Beef Feature Activity, the beef feature rate fell 14.0% while the activity index rose 9.3% in the same reporting period. For a pricing workflow designer, this is not a data error. It is a signal worth understanding before building any automation around it.
Defining the Two Signals
The report's explanatory notes define Feature Rate as the amount of sampled stores advertising any reported item during the current week, expressed as a percentage of the total sample (USDA AMS). It measures participation in the sample. It does not state why a store advertised an item or how deep a discount was.
The same notes define Activity Index as the total number of stores for each advertised item (USDA AMS). It is an absolute frequency measure based on store-item advertising. It is not volume-weighted. It does not measure sales volume, discount depth, competitor size, or store traffic.
In the current report, USDA AMS gives an activity index of 104,065 and a feature rate of 82.0% across 25,521 outlets. The reported weekly changes moved in opposite directions. The definitions explain how that is possible without turning either measure into a sales or pricing result.
Mapping Each Signal to a Pricing Workflow
These two signals serve different functions in a retail-technology stack.
The feature rate can serve as a sample-participation field. A workflow can record whether participation rose or fell. It cannot assign the movement to margin pressure, supply constraints, category strategy, or another cause without separate evidence. It also cannot tell a merchant whether to hold or change a price.
The activity index can serve as a store-item frequency field. A workflow can record the reported level and direction. It cannot identify high-volume competitors, campaign aggression, trade-area exposure, realized demand, or a defensive response.
Used together, the fields describe sample participation and absolute store-item frequency. They do not classify a week as shallow or deep because neither field measures discount depth.
Comparison with Consumer Price Series
The AMS fields describe advertised features in its sample. A separate Federal Reserve Economic Data series reports ground chuck at 6.850 dollars per pound in July 2026. The two records have different measures and periods. The supplied evidence does not connect them at a store level.
A feature-rate decline does not state that consumer prices fell. An activity-index increase does not state that realized sales rose. A system should keep the AMS fields and the FRED observation separate unless another verified record supplies a valid join.
System Design Guardrails
Several constraints should be built into any system that ingests these signals.
First, scope the data to the report. USDA AMS describes prices advertised by major grocery retailers and gathered through a weekly survey. The current table lists 25,521 outlets. The source does not claim to be a census of all retail beef sales.
Second, retain each period rather than rewriting one weekly movement as a long-term trend. The current evidence establishes the reported week only. A later trend claim needs additional verified periods.
Third, keep an automated action outside the data-ingestion step. The two AMS measures do not authorize a price change. Cost, inventory, realized sales, trade area, product specification, and merchant review would need their own evidence paths.
What Not to Automate
The most important guardrail is categorical. Feature data from a grocery panel is not a forecast for an independent retailer's demand. Automating a price change at an independent shop because the USDA AMS feature rate moved in a particular direction treats a market-level signal as a store-level instruction. These are different things. The signal informs workflow design and human review. It does not replace the merchant's judgment about their own cost structure, customer base, and competitive position.