MARKETING ATTRIBUTION MODELS: WHAT THEY SHOW AND WHAT THEY MISS

First-touch gives credit to the earliest known interaction and is useful for understanding discovery. Last-touch credits the interaction nearest conversion and can help inspect immediate demand. A linear model shares credit across recorded touches; time-decay weights recent ones more. Position-based models emphasize the first and final touches. None is a universal truth about causation.

Consider a roofing buyer who receives a neighborhood mailer, later searches for the company, calls, and accepts an estimate after follow-up. Last-touch may call it search. First-touch may call it direct mail. If the mailer lacks a trackable response or the call is not matched to the job, both answers may be incomplete.

Attribution reports are often precise-looking because they show percentages. That precision is misleading when campaign names vary, forms omit source, calls are unassigned, or sales records do not connect back to the lead. Before buying another dashboard, standardize UTMs, source fields, lead stages, and the join from inquiry to opportunity and sale.

Keep unknown and mixed sources visible. Do not force every sale into a channel just to make a chart add up. Direct traffic, word of mouth, long consideration periods, and offline influence will leave uncertainty.

Compare qualified opportunities, estimates, sold jobs, revenue, and margin where available. Review cohorts and trend changes, not only a single conversion event. For larger budget decisions, a controlled pause, geographic comparison, or lift test can add evidence that click-path attribution cannot provide.

A useful operating review asks what decision the model changes. If two models disagree, identify the assumption causing the difference and decide what data or experiment would resolve it. The goal is better spending decisions, not a perfect allocation pie chart.

Map the real lead journey. Define a campaign naming convention and source taxonomy. Capture calls and forms consistently. Choose CRM stages that reflect actual sales work. Then create a simple report by channel that shows spend, qualified opportunities, and closed business, with an explicit unknown bucket.

Only after that foundation should you compare software or design a custom model. The next guide covers the buying criteria.