Attribution Modeling Types: Rules, Examples, and Tradeoffs
Library > Attribution Modeling Types
Written by Maimouna Corr Fonsbøl
Published on 23 October 2026
What changes when you change an attribution model?
An attribution model determines how credit for a selected outcome is allocated across eligible recorded interactions. The same deal can produce different channel totals under different models, even though its value and journey stay unchanged.
The main types are single-touch models, rule-based multi-touch models, and data-driven models. Understanding their allocation rules helps you interpret a report before using it to support an investment decision.
For B2B marketers, first connect the relevant activity to the account and opportunity. A model cannot compensate for missing contacts, incorrectly linked deals, or interactions outside the recorded journey.
Three questions help you choose a useful view:
Recorded introduction: which eligible interaction appears first in the journeys you are analyzing?
Recorded endpoint: which eligible interaction immediately precedes the selected outcome?
Shared credit: how does value distribute when you include interactions across the measured journey?
Choose a primary model you can explain, then compare another view using the same accounts, outcome, dates, and exclusions. This shows how much the conclusion depends on the allocation method.
Single-touch models
Single-touch models place all credit on one eligible interaction. They give you a focused view of one part of the journey, while leaving the other interactions out of the allocation.
First-touch attribution
First-touch assigns 100% of credit to the first eligible recorded interaction. If a $100,000 deal begins with a paid social visit in your data, that interaction receives all $100,000.
Useful for examining which channels first appear in the recorded journeys of accounts that become opportunities or customers.
Assumption: the start of the measured journey is the point you want to emphasize.
Tradeoff: later evaluation and sales activity receive no credit. An earlier untracked recommendation may also have preceded the first recorded touch.
Last-touch attribution
Last-touch assigns 100% of credit to the last eligible interaction before the outcome. If that $100,000 deal's final eligible interaction is an email click, email receives the full amount.
Useful for examining activity immediately before a selected milestone.
Assumption: the endpoint of the measured journey is the point you want to emphasize.
Tradeoff: earlier interactions receive no credit, even if they involved other members of the buying group.
These models can help you investigate where journeys begin and end. Before moving budget based on either view, examine the omitted activity, campaign costs, and account quality.
Rule-based multi-touch models
Rule-based multi-touch models share credit using a defined allocation method. You can inspect a journey and calculate the result from its eligible interactions.
The rules represent assumptions about how to distribute value. They do not measure the extra business an interaction caused.
The examples below illustrate those rules. Model names and milestone definitions can vary by tool, so check the implementation behind your report.
Linear attribution
Linear divides credit equally among eligible touchpoints. With four eligible interactions and a $100,000 deal, each interaction receives $25,000.
Useful for an explainable shared-credit view and comparison with more heavily weighted models.
Tradeoff: repeated routine interactions can accumulate substantial credit because each eligible touch receives the same share.
Time-decay attribution
Time-decay gives recent interactions more weight. For an illustrative seven-day half-life, touches 14, seven, and zero days before the outcome have relative weights of one-quarter, one-half, and one. After normalization, a $70,000 outcome allocates $10,000, $20,000, and $40,000 respectively.
Useful for examining how the result changes when recency is emphasized.
Tradeoff: the chosen decay rate determines how quickly earlier activity loses weight. Recency does not, by itself, establish buying intent or causal impact.
U-shaped attribution
U-shaped emphasizes the first and last eligible interactions. Dreamdata's U-shaped model gives each 40%, with 20% shared by the middle. In a four-touch, $100,000 journey, the allocation is $40,000, $10,000, $10,000, and $40,000.
Useful for comparing the recorded introduction and endpoint with supporting activity between them.
Check which outcome ends the journey. The final eligible touch is not automatically the lead-creation event in every analysis.
Tradeoff: middle interactions share a relatively small allocation even when buyers used them for substantial evaluation.
W-shaped attribution
W-shaped emphasizes three positions or milestones. Dreamdata's model allocates 30% each to the first touch, last touch, and touch containing the conversion, with 10% shared among the middle touches.
Useful for examining the result when an additional point receives substantial weight.
Check the tool's exact milestone definitions and how they relate to your selected stage and eligible activity.
Tradeoff: incorrect event dates or milestone relationships can place extra credit on the wrong interaction.
Full-path attribution
Full-path commonly describes a rule-based approach that weights several milestones through a later outcome, such as a won deal. Some implementations emphasize first touch, lead creation, opportunity creation, and close, with remaining credit shared elsewhere.
Useful for examining allocations across multiple milestones when those milestones are relevant to the reporting question.
Check the actual weights and eligible events. The label does not guarantee identical behavior across products.
Tradeoff: more milestone rules add assumptions and data requirements. Complexity alone does not make the result more accurate.
Dreamdata's standard models include first-touch, last-touch, linear, U-shaped, W-shaped, and data-driven. Time-decay and rule-based full-path are included here as general model types, rather than additional Dreamdata defaults.
Choose complexity only when the extra assumptions help answer the question. A clear linear comparison can be more useful than a detailed milestone model whose allocation no one can explain.
Data-driven attribution
Data-driven models calculate weights from patterns across journeys rather than relying only on a fixed allocation rule. Compare their results with a rule-based view to investigate meaningful differences.
Dreamdata's data-driven attribution model uses a Markov approach to analyze paths leading to a selected stage. It can be applied with limited data; when there is little evidence for different weights, its allocation tends toward linear attribution.
The result depends on the recorded journeys and selected outcome. You cannot explain every weight by inspecting one account alone, and a modeled pattern is not an experiment showing incremental lift.
If the model gives a campaign substantial credit, inspect the relevant accounts and whether tracking or repeated events could help explain the pattern. Use the result to form an investment hypothesis, then test the change when a causal estimate matters.
Account-based attribution describes the data relationships
Account-based attribution brings relevant activity from people at the same company into an account-level view connected to business outcomes. It is compatible with several credit models.
For example, a champion's webinar attendance and an analyst's integration research may relate to the same opportunity. Connecting those records lets you include both in the measured journey instead of examining each contact in isolation.
Dreamdata's account-based data model connects contacts, activity, and companies across sources. That’s the foundation used to automate collection and attribution; the selected credit model then determines the allocation.
Inspect individual customer journeys when a channel ranking changes unexpectedly. Check the account and opportunity relationships, eligible interactions, and selected outcome before concluding that the campaign's performance changed.
FAQs
Which model should I try first when moving beyond last-touch?
Linear is a useful starting comparison because its equal weighting is easy to calculate and explain. Keep the outcome and data scope the same as the last-touch view, then inspect where the allocations differ. There is no required progression toward a more complex model; choose the assumptions that fit your question.
How should untracked influences fit into attribution?
Use customer interviews, self-reported discovery information, and sales notes to add context. Keep those reports distinct from observed event records, and avoid inventing a timestamp or model weight for an uncertain interaction. Use the combined understanding to interpret results and choose what to investigate.