Transparent by design
Price Pulse methodology
Price Pulse is designed to measure how the typical tracked forecourt moved without allowing frequently updating retailers to dominate the result.
The short version
- 1Find each station's final recorded price for the day.
- 2Carry that price forward until a later price-change event replaces it.
- 3Average one closing value per station so every station has equal weight.
Why raw event averages are misleading
The price-history table records changes rather than a complete daily observation. A station that changes its price ten times creates ten rows; a station whose price stays unchanged creates none. Averaging all rows would give the first station ten votes and the second no vote. Price Pulse instead reconstructs one end-of-day price for each station and fuel.
Weekly comparisons
A weekly change compares only stations that have a reconstructable price at both endpoints. The headline average still describes all qualifying stations at the weekly close, while the displayed comparison count shows how many stations contributed to the like-for-like change.
Coverage and validation
- A station and fuel enters the historical series from its first genuine price-history event.
- Prices below 50p or above 300p per litre are excluded as implausible.
- Retailer and postcode-area comparisons require at least five like-for-like stations.
- A dated permanent closure removes a station from that date onward when the source provides one.
- Current live cards are a separate cohort: active stations updated within the previous seven days.
What the average represents
This is an equal-station average, not a sales-volume-weighted average. A small rural forecourt therefore counts once, just like a busy supermarket site. The measure answers “what did the typical tracked station charge?” rather than “what did the typical litre sold cost?”
Known limitations
- The current schema does not contain a complete dated history of station activation and deactivation.
- Brand, postcode-area and motorway labels use each station's current classification because dated dimension history is not yet stored.
- A change event proves the price changed; it is not a feed-observation heartbeat. “Station count” therefore means reconstructable changed-price coverage, not a same-day manual check.
- Fuel withdrawal is not currently represented by a null price-history event, so a last price may carry forward until another trustworthy status signal is available.
- Industry bodies may publish volume-weighted figures and can therefore report a different average.
Reproducibility and corrections
Each published edition stores a frozen data payload and methodology version. Later station updates do not rewrite the figures in an older edition. CSV and JSON downloads expose the figures and station counts used in the report.
To question a figure or request a correction, email [email protected] with the edition date and metric.