What is Attribution Modelling ?
Assign credit for revenue to every channel in the buyer journey so budget follows evidence, not assumption.
Attribution modelling assigns credit for commercial outcomes to each touchpoint in the buyer path from first contact to close. The method corrects a structural investment problem: marketing channels produce value in sequence, and the channel that closes a deal operates on awareness and intent accumulated across months of prior interaction. Without a credit framework, the final touchpoint captures all revenue attribution and every preceding channel captures none, creating an allocation bias that compounds against demand-generating channels across successive budget cycles.
Rule-based models apply a fixed formula to the conversion path. Last-touch assigns full credit to the final interaction before close; first-touch to the entry point; linear distributes credit equally across all touchpoints; time-decay weights interactions more heavily as they approach conversion; position-based models assign forty percent each to the first and last touchpoints and spread twenty percent across the middle. Data-driven alternatives replace fixed rules with statistical estimation: Markov chain calculates each channel’s removal effect on conversion probability, while Shapley value computes marginal contributions across every journey permutation. Both require roughly two hundred attributable monthly conversions before estimates stabilise.
Attribution modelling measures only touchpoints that leave a traceable digital record. Offline interactions, executive relationships, and in-person events remain invisible to every attribution system regardless of model sophistication. The correlation problem limits accuracy further: the touchpoint most reliably preceding conversion is often one buyers seek after forming an intention, not one that caused it. In B2B environments these limitations intensify because purchasing authority is distributed across committees averaging six to ten stakeholders, and most systems model individual contact journeys rather than account-level influence patterns.
Why It Matters
The commercial consequence of misattribution is not a measurement problem — it is a capital allocation problem. Organisations running on last-touch routinely concentrate budget in the channels that harvest demand while defunding the channels that generate it. Correcting the model corrects the budget, and correcting the budget compounds pipeline productivity in every quarter thereafter.
Organisations without attribution models allocate budget through historical inertia and last-click data, both of which systematically undervalue awareness and consideration channels. When attribution modelling reveals that sixty percent of the pipeline closed through paid search was first generated by organic content and events, the investment case for that content changes entirely. The discipline does not create incremental demand — it reveals where demand was already being generated and ensures the investment that produced it continues to receive resources. This correction compresses customer acquisition costs per channel.
Revenue attribution makes the invisible visible. Content assets, event appearances, analyst reports, and brand campaigns produce commercial influence that last-touch models classify as contributing nothing to revenue. Multi-touch credit distribution restores those contributions to the measurement record, giving commercial leadership an accurate view of which channels appear in paths that close and at what stage of the journey. This matters most where marketing-influenced pipeline is a board-level metric — and that case is always made with attribution data showing which activities appeared in the paths of accounts that closed.
B2B sales cycles spanning three to eighteen months generate large volumes of pre-conversion touchpoint data that attribution modelling transforms into predictive intelligence. When Markov chain or Shapley models are applied to historical journey data, patterns emerge: certain channel sequences show higher conversion rates, certain content types appear disproportionately in paths that close large accounts, and time-between-touchpoints correlates with deal velocity. Commercial teams that harvest this intelligence design nurture sequences informed by what the record shows actually works, not by vendor benchmarks or instinct.
Model selection is a measurement philosophy decision with commercial consequences, not a technical one. Last-touch optimises channel investment for closing behaviour and progressively defunds acquisition; first-touch optimises for awareness at the cost of ignoring conversion mechanics; linear assumes all touchpoints contribute equally, which is false but avoids systematically distorting the record in one direction. The appropriate model depends on sales cycle length, the volume of attributable conversions available for data-driven calibration, and the specific commercial question the organisation needs to answer.
How It Works
Attribution modelling begins with journey data — each buyer’s or account’s ordered touchpoint sequence from first contact to conversion — assembled from CRM records, marketing automation, and web analytics. Data quality determines model reliability: cross-device gaps, offline blindspots, and incomplete tracking introduce measurement error that no model can correct retroactively.
Rule-based implementation defines the conversion event, identifies touchpoints in each path, applies the credit formula, and aggregates by channel. The output is a credit score per channel set against spend to calculate revenue per dollar invested across the full mix.
Data-driven implementation requires a minimum conversion dataset — roughly two hundred attributable outcomes — before models produce reliable estimates. Markov chain calculates each channel’s removal effect on conversion probability; Shapley value enumerates all possible touchpoint coalitions and computes marginal contributions. Both produce a weighted credit distribution that more accurately reflects conversion behaviour than any rule-based alternative, at the cost of greater data volume and computational requirements.
Advisory Insight
The most common attribution failure mode is treating last-touch output as evidence of channel performance. Last-touch records only which channel the buyer visited before converting — not which channels built the conviction that made that visit worthwhile. Organisations that run on last-touch for more than two years develop a systematic bias against awareness investment: the model never credits those channels, and budget committees never see a case for funding them. The correction is not a new model — it is recognising that last-touch produces a predictable distortion, then auditing every channel investment made in the last three years through that lens. What last-touch has defunded typically represents more commercial opportunity than any attribution technology change.
Common Misconceptions
MYTH
Last-click attribution reveals which marketing channels are truly responsible for winning new customers and generating closed revenue.
REALITY
Last-click records which channel the buyer visited before converting, not which channels built the conviction that made that visit worthwhile. In B2B, the converting touchpoint finalises a decision formed across months of prior engagement — none of which last-click credits.
MYTH
Data-driven attribution models work reliably regardless of how many monthly conversions a business actually generates.
REALITY
Data-driven models require roughly two hundred attributable conversions monthly before estimates stabilise. Below that threshold they amplify measurement noise rather than reveal true channel contribution, and a calibrated linear model outperforms them on most B2B datasets.
MYTH
Implementing multi-touch attribution requires overhauling or replacing the existing marketing technology stack entirely.
REALITY
Multi-touch attribution runs at the analysis layer on existing CRM and analytics data, requiring no platform changes. The investment is in data unification and model configuration — many teams run Markov chain models in spreadsheets before committing to dedicated tools.
MYTH
Attribution modelling solves the problem of measuring offline marketing activity and the influence of executive relationships.
REALITY
Attribution modelling captures only digital touchpoints. Offline events, executive conversations, and relationship development remain invisible to every model. Organisations with substantial offline activity need a complementary measurement layer alongside it.
Sources & Further Reading
Evolving Attribution Models in Digital Advertising: Balancing Clicks, Views and Customer Journeys
Unveiling Digital Dynamics: Do Digital Media Investments Impact Organic Branded Searches?
Causal Inference for Multi-Touch Attribution in Digital Marketing
Multi-Touch Attribution and Media Mix Modelling for Marketing ROI Optimisation
related
Growth MarketingDecks are easy. Decisions are not.
Bring us the real question. We’ll come back with how we’d approach it. Not a brochure. A starting point.
Take the Next Step