Bolt Market Data Scraping

Availability that follows commuter peaks

A dedicated-fleet grocery operator's delivery availability moves with grocery demand. Bolt's courier capacity is drawn from the same pool serving passenger rides, so it moves with both.

• Morning and evening commuter peaks pull capacity toward rides, which can lengthen grocery promise times even when store stock is unchanged.
• Weather that raises ride demand has the same effect.
• A promise time lengthening is therefore not necessarily a fulfilment problem — it may be a capacity allocation decision.

That has a direct consequence for how the data must be collected: a single daily observation is close to useless. We sample across dayparts, record daypart and the local-offset timestamp, and deliver promise time and stock as independent fields so the two can be separated.

This is the same reasoning behind our Grab work, where a shared fleet across ride-hailing and delivery produces the same interaction.

Owned inventory, so the fee stack carries the variation

Bolt Market holds its own stock, so there is one price per item per store rather than competing offers to resolve. As on Gopuff and DashMart, that moves the commercial variation elsewhere.

• Delivery fee, service fee and small-order fee move independently of item price and of each other.
• Minimum order value shapes basket behaviour more than a few cents on an item at this basket size.
• Any subscription or membership typically waives fees rather than discounting items, which is a structurally different mechanic.

We deliver each fee as its own field and never fold any into the item price. A single effective price here would hide the mechanic that actually varies.

A footprint spanning very different markets

Bolt operates across the Baltics, Eastern Europe and a number of African cities. Those markets differ in assortment, price level, currency stability and how established delivery is.

• country and city on every record, with markets recorded per batch rather than assumed.
• Local currency primary, with fx_rate and fx_observed_at stamped at the observation, since several currencies in the footprint are volatile.
• Local-language names retained exactly, across Estonian, Latvian, Lithuanian, Polish, Ukrainian, Romanian and several African market languages.

A regional average across this footprint would describe none of it, so cross-market comparison is a deliberate join rather than something we do for you.