How to choose an attribution model for B2B marketing

Choose a model you can explain in a budget meeting

A paid social campaign can look valuable under first-touch attribution and less valuable under last-touch. That does not automatically make either report wrong. The models answer different questions about the same recorded journey.

Your job is to understand the difference before moving budget.

Use the model to assess contribution alongside cost, account quality, and campaign maturity. Attribution allocates credit; experiments help test whether an investment adds outcomes you would otherwise miss.

Step 1: Start with the marketing decision

For a discovery campaign, you may want to know which channels first engage suitable accounts. For a nurture program, you may want to understand activity across the path to a qualified opportunity.

Choose the question before choosing the model.

Agree on the comparison

• Decision: name the investment you are considering, such as renewing an event or increasing a campaign's budget.

• Outcome: select a clearly defined stage, such as qualified opportunity or closed-won, and specify whether you are comparing counts or value.

• Review: choose a useful cadence and name the person responsible for turning the findings into a recommendation.

• Journey period: check your own time from engagement to the selected outcome. Understand which early activities the reporting setup includes.

Use earlier outcomes to monitor developing campaigns, then revisit revenue as the account group matures. Keep the stage visible so a report on qualified opportunities is not mistaken for a report on won deals.

For example: “We compare attributed qualified pipeline for last quarter's event campaigns to decide which events to repeat.” That statement makes the campaign group, outcome, and decision explicit.

Step 2: Check the outcome and account journeys

Dreamdata's account-based data model connects contacts, companies, and activity with pipeline and revenue. It provides the foundation for automated collection and attribution. The stage definition and eligible journey determine the records and value available for the model to allocate.

Check the inputs

1. Sources: connect the CRM and the relevant campaign, event, and website activity for the decision you want to make.

2. Account matching: inspect known contacts and companies, including subsidiaries or accounts with several opportunities.

3. Naming hygiene: check your top sources and campaigns for UTM/campaign naming gaps that create uncategorized sessions.

4. Stage definitions: in Data Hub, check the source records, filters, outcome date, and value used for the stage you are measuring.

5. Journey scope: confirm which people and activities are associated with the selected stage, particularly when a company has multiple deals.

6. Known accounts: review a small sample and confirm the expected recorded activity appears on the correct dates.

Check early activity and deal values

• Include relevant recorded activity before pipeline entry. A webinar attendee and a later demo requester may be different people at the same company.

• Confirm the stage value and currency against the CRM. A correct attribution rule cannot repair an incorrect deal amount.

Use Customer Journeys to inspect familiar accounts before comparing model results. You should be able to explain which activity is included and which gaps need attention.

Step 3: Match the model to the question

Pick a primary view for recurring planning and use other models to investigate specific questions. A model is useful when its assumptions fit the decision and your team can understand the resulting allocation.

Illustrative example: a $100,000 deal has four eligible touches in order: paid social, a webinar, organic search, and an email. First-touch assigns $100,000 to paid social; last-touch assigns $100,000 to email; linear assigns $25,000 to each.

Before cutting the webinar, look at its role across other journeys and the quality of the accounts it reaches.

Compare the options

• Linear: gives each eligible touch equal credit. Use it as a clear baseline when you want to see contribution across the journey without choosing heavier milestone weights.

• U-shaped: emphasizes the first and last eligible touches while sharing some credit with the middle. Use it when those endpoints are central to your question.

• W-shaped: emphasizes selected milestones and includes supporting interactions. Check what counts as a conversion milestone in your setup.

• Data-driven: uses patterns across journeys to allocate credit. Dreamdata's model is Markov-based; with limited data it tends toward linear allocation. You do not need to invent a minimum-volume threshold before trying it.

Keep comparisons useful

• Use first-touch to inspect where recorded engagement begins and last-touch to examine activity before the outcome. Neither shows the whole journey on its own.

• If direct traffic receives substantial credit, inspect the underlying journeys and tracking first. Compare an exclusion only when it fits the reporting question, and explain how it changes eligibility.

• Show the model name, stage, dates, and filters alongside each report so the team compares the same scope.

Dreamdata's attribution models let you compare these views. Investigate the channels whose allocation changes enough to affect a budget decision, rather than trying to make all models produce the same answer.

Step 4: Document exclusions and any custom rules

Begin with a standard model. Add custom rules when you have a specific reason, such as a defined milestone that the default weighting does not represent. Keep report channel groupings separate from the rules that allocate credit.

For example, grouping webinar and conference activity under “Events” can simplify a budget report. Assigning Events a fixed share of every deal's credit is a different decision and needs a clear rationale.

Make the rules understandable

1. Start with a predefined model and inspect its behavior on your selected outcome.

2. If you need a custom model, define attribution groups by the relevant event, source, channel, or other supported conditions. Check customization availability for your account.

3. Document group weights, how credit is shared within each group, and the fallback used when a group is empty or some credit remains unallocated.

4. Review exclusions for test traffic or other activity that should not qualify. Check the fallback behavior on journeys containing only excluded sessions.

5. Inspect sample journeys to verify the resulting credit. Where groups overlap, check how the session's total allocation is calculated.

Keep a short record of the model

• Included: channels/sources that can receive credit and how they’re grouped.

• Excluded: what you remove and why.

• Stage scope: which stage model the attribution is tied to (revenue vs pipeline vs SQL).

• Ownership: who maintains the rules, why they were chosen, and when changes will be reviewed.

Use Dreamdata's custom attribution documentation for the detailed setup. Your written explanation should still be simple enough for sales and Finance to understand without opening the configuration.