Delivery-platform prices change with the collection context

01Write down the shopping scenario

Choose the store, channel, delivery or pickup mode, destination scope and observation window. Without a consistent scenario, two collections can return different prices for legitimate reasons. A change in the scenario should be recorded as such rather than interpreted as a market movement.

For brand-level analysis, define which stores represent the sample. A few locations do not establish national pricing. Keep regional coverage and unavailable locations visible so the analyst can judge how broadly a finding applies.

02Separate item economics from order economics

The listed item price is one component of an order. Required modifiers can change the purchasable item, while service fees, delivery, taxes and promotions can depend on the whole basket. Some components may be unavailable without steps outside the agreed collection scope.

Store each component with a status indicating whether it was observed, not applicable or unavailable. Do not use zero for an unknown fee. If you calculate a basket comparison, describe the standard basket and which components were included.

03Match stores and items cautiously

Use platform IDs and reviewed crosswalks where possible. Store names can be similar, and addresses can differ in formatting or suite details. Match location identity before comparing menus. At item level, a combo and a standalone product are not equivalent merely because their names share a phrase.

Preserve modifier groups and sizes during matching. A medium drink included in a meal is not always comparable with a separately priced drink. Ambiguous matches should remain in a review queue instead of creating an apparently precise price gap.

04Make time part of the record

Menus and availability can vary by daypart and store operating state. Keep observation time and the relevant local context. If a source exposes a breakfast menu in one run and an all-day menu in another, compare the appropriate subsets rather than treating every difference as an item removal.

For recurring analysis, align collection windows where practical and disclose gaps. A failed location check should not erase the previous menu or imply that the store closed. Keep successful observations and collection status as separate histories.

05Deliver a dataset that can explain a difference

When a comparison flags a change, the receiving team should be able to inspect the store, item configuration, channel and time behind it. Include stable IDs, original labels and the normalisation rules. The value of the feed is not just discovering different numbers; it is explaining whether those numbers describe equivalent customer choices.