CROSS-CHANNEL ATTRIBUTION MODELLING AND ITS IMPACT ON DIGITAL MARKETING PERFORMANCE

Abstract

In an increasingly fragmented digital media landscape, marketing managers face a critical challenge in identifying which digital channels truly contribute to customer conversions. Attribution modelling provides a framework for assigning conversion credit across multiple customer touchpoints; however, different attribution models often produce significantly different interpretations of marketing effectiveness. These differences directly influence marketing budget allocation decisions and campaign optimization strategies. The present study conducts a comparative analysis of five major attribution models — First-Click, Last-Click, Linear, Time Decay, and Markov Chain Attribution — using a simulated omnichannel e-commerce dataset representing 5,000 customer journeys, 18,450 touchpoints, and 1,050 conversions across five digital marketing channels: Social Media, Paid Search, Display Advertising, Email Marketing, and Organic Search. The primary objective of the study is to examine how attribution model selection affects channel credit distribution and marketing budget allocation. Python-based analytical techniques and statistical evaluation methods were used to calculate attribution credit under each model and compare their effectiveness. The study further evaluated model accuracy using Root Mean Square Error (RMSE), considering the Markov Chain model as a quasi-ground-truth benchmark. The findings reveal substantial divergence in channel valuation across attribution approaches. Under Last-Click Attribution, Paid Search received 42% conversion credit compared to only 31% under the Markov model, indicating systematic overvaluation of lower-funnel channels. Conversely, Social Media received 38% credit under First-Click Attribution but only 24% under Markov Attribution, demonstrating the tendency of first-touch models to overestimate awareness-stage channels. Time-Decay Attribution produced results closest to the Markov model, achieving the lowest RMSE value of 4.9, compared to 13.5 for Last-Click Attribution. The study also demonstrates significant variation in budget allocation outcomes. When a hypothetical monthly marketing budget of ₹10,00,000 was distributed based on attribution outputs, Last-Click Attribution allocated ₹4,20,000 to Paid Search, whereas Markov Attribution recommended a more balanced allocation of ₹3,10,000. Similarly, First-Click Attribution disproportionately favored Social Media and Display Advertising. These findings indicate that attribution model selection can materially alter marketing investment decisions and potentially distort organizational understanding of channel performance.Customer journey analysis further revealed that modern digital consumers interact with an average of 4.2 touchpoints before conversion, confirming the inadequacy of simplistic single-touch attribution approaches in omnichannel environments. Sensitivity analysis showed that rule-based models become increasingly unstable as customer journey complexity rises, while data-driven approaches remain comparatively robust. The study concludes that data-driven attribution models, particularly Markov Chain Attribution, provide more realistic and strategically balanced estimates of channel contribution than traditional rule-based methods. The findings hold significant managerial relevance for digital marketers and Indian Direct-to-Consumer (D2C) brands seeking more accurate marketing performance measurement and efficient budget allocation in increasingly complex digital ecosystems.