“If Attribution Worked, Nobody Would Fight About It” – with Matthew Sciannella

What It Covers

Matthew Sciannella, VP of Innovation at Refine Labs, joins Revenue Vitals to break down why B2B attribution models fail, how brand spend should be measured against contribution margin rather than sourced revenue, and why strong product marketing — not channel execution — is the common thread across every high-performing demand generation program.

Key Questions Answered

• •Attribution as a flawed operating system: Multi-touch attribution only tracks digital touchpoints it can detect, and ad platforms have no incentive to support accurate cross-channel measurement — they want users staying on-platform. Sciannella argues that if attribution actually worked, nobody would fight about it. Teams should treat attribution as one data point within a broader measurement ecosystem, not a diagnostic tool for identifying performance problems.
• •Brand spend measurement via contribution margin: To make brand investment financially defensible, strip source and attribution from the conversation entirely and measure brand expenditure against contribution margin across time periods. Build a model — Sciannella uses an Excel sheet with geometric decay formulas — showing that if brand is working, marketing ROI should improve period over period. This gives finance a concrete framework to evaluate brand without constraining it to lead-gen metrics.
• •Product marketing as the top demand gen differentiator: Across dozens of client engagements, the single most consistent factor in strong demand generation is product marketing quality. Companies that maintain a tight customer and buyer feedback loop, clear positioning, and segment-specific messaging consistently outperform those that don't. Positioning is company-wide; messaging must be business-unit specific — and both erode as companies add horizontal product lines without staffing accordingly.
• •Win rate visibility before adding pipeline: Many marketing leaders do not know their win rates, sales cycle length, or ACV. Sciannella's diagnostic process consistently finds that low win rates — not insufficient pipeline volume — are the core problem. Improving win rate yields far greater revenue leverage than increasing pipeline volume fed into an underperforming sales process. Marketing should have full visibility into unit economics: cost per MQL, SQL, and opportunity at minimum.
• •Incrementality testing over attribution tooling: Rather than relying on multi-touch attribution platforms, Sciannella pushes toward designed experiments — geo holdout tests, split testing against target account lists, and cohort-based pipeline acceleration tests — to prove marketing lift. Teams should log experiments systematically, tracking inputs and outcomes by client, to build a catalog of what actually drives incremental revenue rather than what attribution software credits.
• •AI adoption requires clean data infrastructure first: Companies attempting to layer AI onto existing systems with poor data hygiene, broken UTMs, or misaligned CRM objects are producing worse outputs, not better ones. Sciannella observes that the clients with clean, well-structured data are the rare exceptions. Organizations should resolve technical debt before AI implementation — otherwise automation scales dysfunction. Compliance constraints at enterprise clients also frequently prohibit LLM use on engagement data entirely.