Jumbo Data Scraping

Attribution is not determinable, and we say so

Where a retailer publicly commits to matching or beating competitor prices, a price movement has at least two possible causes:

• An own pricing decision, reflecting its own strategy.
• A mechanical response to a competitor's move, arriving with whatever lag the process has.

A listing shows the new price. It does not show which. So we set price_change_attribution to a constant not_determined rather than letting an analyst assume the first.

Why that constant matters

Without it, a competitive analysis reads a series of price cuts as a strategic campaign. Some of them are a guarantee executing, which is a different thing — it says something about the competitor rather than about this retailer.

This is the same position our Colruyt page takes, and it generalises: any retailer with a published matching policy needs this field, and most datasets do not carry it.

Which means the competitor set and the schedule are the engagement

If some movements are responses, then the sequence is the analysis — who moved first, and how long the response took.

That has two design consequences:

• The competitor set must be collected too. Observing this retailer alone shows you responses with no stimulus.
• They must be on the same schedule. Retailers observed hours apart produce apparent leads and lags that are entirely timing artefacts.

So a Jumbo engagement is usually scoped as a synchronised panel rather than a single retailer, with panel_schedule_id shared and observation_interval_minutes recorded.

And a higher refresh than we would normally recommend

Our cadence page argues that frequency should follow observed price movement, and that we frequently recommend lower than a client asks. This is the opposite case.

Where response lag is the question, a daily observation can only resolve lags to the nearest day — which for a same-day matching process is no resolution at all. Sub-daily is genuinely warranted here.

Dutch conventions, and what we do not produce

Statiegeld

Container deposits apply and are refundable. deposit_amount and deposit_type are separate fields, never folded into the price — as on our Albert Heijn page.

Unit pricing

Displayed unit price captured, ours computed alongside, disagreement flagged.

Own label

Several ranges across tiers, flagged with tier from range naming, unmatched across retailers.

What we do not produce

• An attribution for a price change. The constant says not determined, and we do not model one.
• Whether a guarantee was invoked on a specific product. Not published.
• Margin impact of the guarantee. Requires cost data nobody publishes.
• Loyalty account data, sales, customer or employee data.

The second one is worth naming. It would be easy to infer a match from a price aligning with a competitor's. Alignment is not evidence of a match — two retailers can arrive at the same price independently, and frequently do on branded lines.