Five ways to attribute B2B social selling outcomes, and what each can prove

B2B social selling attribution methods answer different questions. A last-touch rule allocates credit. A controlled comparison may estimate an additional effect. Choosing between them requires the same criteria: unit, eligible touch, outcome, window, data burden, assumptions and causal strength.

This non-ranked desk comparison was completed on 5 September 2026. Inclusion required a current first-party service record or UK government evaluation source that exposed credit or causal logic. Supplier case studies, performance benchmarks and undeclared identity matching were excluded. No commercial conflict exists, and no account was tested.

What to take away

Declared-source attribution

Ask the buyer or opportunity owner which interactions informed the next step, then assign credit under a written coding rule. The unit is an account or opportunity; the window is the stated recall period; the evidence is a dated response plus the underlying activity record.

This method captures conversations that tracking misses, but recall and internal incentives can bias answers. Use a fixed question, allow "unknown" and preserve contrary evidence.

Response coverage is coded answers divided by eligible people asked in that recall period; exclude requests withdrawn before asking, but not silence. It supports a contribution account, not a causal effect. HM Treasury's 2026 Magenta Book describes theory-based approaches as requiring a tested causal chain, multiple evidence sources and attention to alternatives.

Last recorded touch

Give all credit to the last eligible recorded interaction before the defined outcome. The unit is a contact or account path; eligibility may be a tagged visit, logged message or meeting; the window ends at qualification or sale. Data burden is relatively low, but the result is highly sensitive to missing logs and the chosen end event.

Google Analytics' global attribution documentation defines last-click allocation within its product and explains which interactions receive credit. That rule does not automatically cover an offline B2B conversation. Publish the internal touch hierarchy. Missing-path rate is eligible paths lacking a required touch divided by all paths in the attribution run; exclude outcomes whose window remains open and state the run period.

Fractional multi-touch model

Split credit across eligible recorded interactions by a fixed or data-driven rule. Use one opportunity and its interaction path as the unit, state the lookback period and ensure allocated shares sum to the whole outcome. It can represent a longer path, but changing the model does not change the observed sale.

HubSpot's first-party attribution definitions distinguish contact, deal and revenue reports, identify tracked interactions and disclose product sampling for very large activity sets. Access depends on the stated subscription. Before comparing results, export the model, eligible assets, association rules and late-data treatment.

Account-influence rule

Mark an opportunity influenced when an eligible social activity is linked to a relevant person or owner within the chosen window. The unit is the opportunity; the numerator is opportunities meeting the influence rule; the denominator is all eligible opportunities closing in that period. Exclude open opportunities and report unmatched CRM records separately.

LinkedIn's global Sales Navigator ROI FAQ describes aggregate activities and influenced or sourced revenue using CRM sync. It is a supplier attribution view, not independent proof that the platform caused the outcome. Record opt-outs, association logic and product eligibility.

Controlled effect estimate

Compare a prespecified outcome between assigned intervention and credible comparison groups. For each group, outcome rate is assigned units with the outcome divided by all eligible assigned units in the evaluation period; apply only exclusions fixed before allocation. The effect is the difference between those rates. This asks the causal question, but needs adequate sample, group separation and specialist design.

The UK government's quality-in-impact-evaluation guidance explains that monitoring a change does not establish that an intervention caused it and centres the counterfactual. Private firms may adapt this logic, but qualified statistical review is essential.

Choose the simplest method that answers the decision. Report attribution and controlled effect separately; never merge them into an unexplained return figure.

Before you act

Common questions

What is the difference between attribution and a controlled effect estimate?

Attribution allocates credit under a rule, such as last-touch or fractional multi-touch. A controlled effect estimate compares a prespecified outcome between assigned intervention and credible comparison groups to ask the causal question. The article states these answer different questions and should be reported separately.

Why can declared-source attribution be biased?

It relies on asking the buyer or opportunity owner which interactions informed the next step. Recall and internal incentives can bias the answer. The article advises using a fixed question, allowing unknown and preserving contrary evidence to reduce bias.

What does a fractional multi-touch model prove?

It splits credit across eligible recorded interactions by a fixed or data-driven rule. It can represent a longer path, but changing the model does not change the observed sale. It supports a contribution account, not a causal effect.