Attribution for a press release is an assignment rule, not proof of cause

Press release distribution attribution methods assign an observed event under a rule. Causal evaluation asks whether the announcement changed an outcome relative to what would otherwise have happened. Financial reconciliation establishes what the buyer's accounting records recognise. These are separate questions, and none can be substituted for another.

This desk comparison assumes the same eligible outcome population, observation period, timezone, route labels, deduplication and correction rules for every method. It covers one authorised announcement issued through a UK newswire or direct-to-media route. No buyer data, product or supplier feature was tested, no conflicts are known and there is no universal winner.

What to take away

Common comparison case

The buyer asks whether an evidenced enquiry was associated with the announcement and whether an incremental effect can be estimated. Each delivery lane remains a separate exposure record.

Define the enquiry event, eligible population, numerator, denominator for a rate, matching rules, window, timezone, duplicates, missing identifiers, privacy route, source/query versions, owner and uncertainty. A changed population or window breaks the comparison.

Press-release attribution differs from general campaign measurement because the unit of exposure is an embargoed announcement distributed to a named media list. Journalist pick-up, newswire reporting dashboards and coverage or AVE reports each apply their own matching rule. Before comparing methods, the buyer must record whether the event being attributed is a distribution receipt, an editorial pick-up, a published article or a downstream enquiry.

HM Treasury's Magenta Book distinguishes process evaluation from assessment of attributable impact and discusses theory-based, experimental and quasi-experimental approaches. It is government guidance rather than evidence that any method will work for this commercial case.

Like-for-like comparison

The manual baseline preserves operational observations and assigns no credit when dependable linkage or a defensible comparison is absent.

Keep the evidence layers visible

A supplier rule must retain the supplier's exact label, definition, version and export date. Newswire services such as PR Newswire, Business Wire and PA Media each publish their own distribution-receipt definitions; media-monitoring platforms such as Cision and Meltwater apply separate clipping and coverage rules. The buyer may reconcile a supplier export with an endpoint or newsroom event, but must not silently rename technical acceptance as publication, readership or response. Independent editorial work cannot be treated as a guaranteed distribution deliverable.

A financial record needs finance-owned definitions for transaction status, currency, VAT, refunds, period and cost allocation. A tagged enquiry is not a recognised sale, and a later sale is not automatically incremental return.

ONS presents coherence and comparability among its official-statistics quality dimensions. Applied as a challenge, those concepts require the buyer to explain whether the same event, geography, period and method are being compared. They do not validate a private estimate.

Decide without overclaiming

Use the least ambitious method that answers the buyer's actual decision. A rule may allocate cases for workflow planning. Matching may help reconcile systems. A properly reviewed evaluation design may address incrementality. Report each output under its own label and uncertainty.

Stop causal wording if the counterfactual is missing, exposure leaks across groups, outcomes are defined after inspection, data loss differs materially between conditions, or the comparison assumptions fail. Stop the attribution view if identities, windows, duplicate rules or source versions cannot be reproduced. Privacy, PECR, security and finance reviewers have separate vetoes.

No approach permits advertising-value equivalence or monetisation of estimated impressions. No result should be generalised to the United Kingdom without a defined population and valid design. This publication-hold draft requires qualified statistical and causal review against the buyer's actual records before use.

Before you act

Common questions

What is the difference between attribution and causal evaluation?

Attribution assigns an observed event under a rule. Causal evaluation asks whether the announcement changed an outcome relative to what would otherwise have happened. Financial reconciliation establishes what the buyer's accounting records recognise. The article states these are separate questions, and none can be substituted for another.

What does the Magenta Book say about impact assessment?

The article says HM Treasury's Magenta Book distinguishes process evaluation from assessment of attributable impact. It also discusses theory-based, experimental and quasi-experimental approaches. The article notes it is government guidance rather than evidence that any method will work for this commercial case.

When should causal wording be stopped?

Stop causal wording if the counterfactual is missing, exposure leaks across groups, outcomes are defined after inspection, data loss differs materially between conditions, or the comparison assumptions fail. Stop the attribution view if identities, windows, duplicate rules or source versions cannot be reproduced.