Custom Attribution Models: A Practical B2B Guide
Library > Custom Attribution Models
Written by Alexandra Flygare
Published on 27 October 2026
A custom attribution model lets you define how credit is divided among the interactions in a buying journey. You might give particular types of activity a larger share, change which sessions qualify, or use rules that reflect a specific reporting question.
The useful question is whether that customization helps you make a better decision. A model built around your preferences can make a favorite campaign look stronger without adding any evidence about its effect.
When does a custom attribution model make sense?
Start with a question a standard model cannot answer well enough. For example, you may want to understand how channel rankings change when a defined set of evaluation activities receives more weight.
Separate the reason for customization from the result you hope to see. Different people can help you test the assumptions:
Marketing: explain which campaign or investment decision the comparison should inform.
Revenue operations: check the event definitions and the relationships between contacts, accounts, and opportunities.
Sales: add context about what happened during the buying process and what buyers needed.
Finance: confirm the value being allocated and how it relates to costs and business outcomes.
Define the question before the weights
Suppose you are considering a larger investment in evaluation-stage webinars. A useful analysis would specify:
The decision: whether to expand the program, improve its content, or investigate its role further.
The outcome: opportunities created or deals won, measured separately using a consistent value definition.
The comparison: how the same journeys look under equal credit and a rule that emphasizes the selected webinar activity.
Build on the account journey
In B2B, the relevant activity can come from several people at the same company. A single opportunity might include:
A marketer discovering a practical article through organic search.
An operations specialist attending a webinar about integrations.
A department leader reviewing a business case.
A procurement contact checking requirements before the deal is signed.
Connect that activity to the relevant buying process before applying weights. Keep separate opportunities at the same company distinguishable, particularly when a renewal and a new purchase overlap.
Dreamdata's account-based data model connects activity, contacts, companies, and outcomes across connected sources. That provides the foundation for applying an attribution model; the credit rules are a separate choice.
What belongs in the measured journey?
The report needs clear answers to five questions:
Which account and opportunity does this journey represent?
Which contacts and identifiable activities can be linked to it?
Which recorded interactions qualify, and how are repeated events grouped?
Which outcome ends the journey, and what does its value represent?
Which time limits or exclusions apply?
Missing activity stays missing when you change the weights. If an event is attached to the wrong company or a campaign source is mislabeled, correct the input before interpreting the new allocation.
A custom attribution example
Consider the example of a $120,000 opportunity with six eligible sessions. Your proposed model assigns 60% equally to two webinar sessions and 40% equally to four other sessions. The groups are distinct and cover the whole journey:
Organic article visit: $12,000, or 10%.
First webinar session: $36,000, or 30%.
Paid social visit: $12,000, or 10%.
Second webinar session: $36,000, or 30%.
Case study visit: $12,000, or 10%.
Demo session: $12,000, or 10%.
The calculation is straightforward: $120,000 × 60% ÷ 2 = $36,000 per webinar session; $120,000 × 40% ÷ 4 = $12,000 per other session. The allocated total remains $120,000.
These are illustrative weights, not a recommended default. They express the team's assumption about how credit should be shared; they do not establish that webinars caused 60% of the opportunity.
Compare the same journey under linear attribution
Linear attribution gives each of the six sessions $20,000. That produces a useful comparison:
The two webinar sessions receive $40,000 in total under linear, compared with $72,000 under the custom rule.
The other four sessions receive $80,000 in total under linear, compared with $48,000 under the custom rule.
The $32,000 shift comes entirely from the weighting choice. Investigate webinar attendance, account fit, buyer questions, and later deal progress before using the higher figure to justify more spend.
What can you customize in Dreamdata?
Dreamdata supports custom rule-based attribution models using session groups and percentage weights. Groups can use conditions such as channel, source, event, and conversion type.
Group membership determines which sessions qualify for a share.
The group weight determines the portion of credit assigned to that group.
The sharing rule determines how that portion is divided, including equal sharing or channel-based sharing.
A session can belong to more than one group and receive credit from each.
A fallback model handles unassigned credit, including a group's share when that group has no matching sessions.
Those details matter when reading a result. Check group overlap and fallback behavior in actual journeys so you can explain why a session received its share.
Attribution exclusions also affect which sessions receive credit. In Dreamdata, they do not erase the activity from the customer journey. When every session is excluded, credit is shared equally among those sessions.
Choose weights for a stated reason
Before adopting a custom rule, ask whether its assumptions are useful and testable:
Can you explain why this activity receives more credit without pointing to the higher total it produces?
Would the same rule apply to a comparable program owned by another team?
Does a modest change in weights reverse the channel ranking?
Can you explain the result when the preferred activity is absent from a journey?
Evaluate the model before changing the budget
Check both the calculation and the decision it informs. A model can add up correctly while offering little new insight into where the next marketing dollar should go.
Reconcile the numbers: for fully allocated journeys, session shares should sum to 100% and values should match the selected outcome total.
Review individual customer journeys: inspect typical and unusual cases with people who know the accounts.
Compare allocations: keep outcomes, opportunity groups, dates, and inclusion rules consistent when testing different weights.
Record changes: distinguish a new model version from a real change in campaign performance.
Include lost and stalled opportunities in the wider campaign review, alongside won deals. Understanding only successful journeys can leave you unaware of how often the same activity appears in unsuccessful buying processes.
Turn the finding into a decision
If a webinar attracts relevant buyers but leaves recurring evaluation questions unanswered, improve the material. If you propose increasing investment, define the outcome you expect and use a suitable comparison or experiment to assess the effect. A larger allocated share alone is not enough.
FAQs
Should you replace standard models with a custom model?
Only when the custom rules serve a clear purpose. Keep a standard model as a comparison so you can see how much of the result depends on your assumptions. A custom model adds maintenance and explanation work, so it should earn its place.
How is custom rule-based attribution different from data-driven attribution?
Custom rules assign credit according to choices you specify. Data-driven attribution uses patterns in the measured journeys to determine allocations. Dreamdata's data-driven model uses a Markov approach and tends toward linear allocation when data is limited. Neither approach makes missing journey data complete.
Use your own GTM data to compare B2B attribution approaches before committing to one for planning. The model should help you explain the evidence behind a decision, including how sensitive that decision is to the allocation rules.