T6

Last-touch attribution is the default in most CRM and marketing automation systems because it is the only attribution model that requires no additional assumptions. Changing to multi-touch or time-decay models requires contested assumptions about the shape of the attribution weight function. Marketing mix modeling (MMM) — which explicitly models distributed lags — is expensive, requires 2-3 years of data, and produces confidence intervals that executives distrust. The practical result is that firms that know last-touch is wrong continue using it because the alternatives are harder to defend in a board meeting.

The Nerlove-Arrow adstock model (1962, independently validated in media mix research) shows that advertising creates a stock of mental availability B_r that depreciates slowly: B_r(t) = delta * B_r(t-1) + alpha * Spend(t), where delta ≈ 0.7-0.9 for brand-level investment in B2B. This means a brand campaign run in Q1 contributes to pipeline closed in Q3 via a distributed lag — the effect is real but arrives outside any single attribution window. Last-touch attribution assigns 100% of credit to the final touchpoint (often a demo request or a direct search on the company name) while ignoring all upstream brand touches that built the mental availability that drove the search. Over time, this systematically under-measures brand ROI and over-measures demand-capture ROI. The institutional error: companies cut brand investment when pipeline is tight (because it shows zero last-touch ROI), causing B_r to depreciate, which reduces qualified demand 2-4 quarters later — at which point they are forced to increase brand spend at higher CPM due to reduced share of voice.

• Les Binet & Peter Field (IPA Databank, 1,400 case studies): brand investment pays back over 6-24 month horizons; activation/demand-capture pays back in 0-3 months — the two effects require different measurement windows
• Google/Ipsos B2B brand study: companies with strong brand preference generated 5x more branded searches than weak-brand competitors, and branded search converts at 3-5x the rate of generic search — the demand-capture channel is partly a brand effect
• LinkedIn B2B Institute 'Long and Short' analysis: B2B campaigns that balanced brand (60%) and activation (40%) achieved 2x the long-run ROI of activation-only campaigns over 3-year measurement windows
• Salesforce, HubSpot, Drift brand-building periods: each company invested 3-5 years in thought leadership before seeing measurable closed-won impact — consistent with the Nerlove-Arrow adstock lag structure

• Multi-touch and MMM models require stable attribution assumptions over the measurement period; brand-building periods that coincide with product launches or category events confound the measurement
• For early-stage companies with < $5M ARR, brand investment may genuinely be premature — the addressable audience is too small for brand-level frequency effects
• Some brand spend (sponsorships, awareness ads) has no measurable GTM mechanism even at long horizons; the Nerlove-Arrow model requires spend that reaches the target buyer persona repeatedly

Implement a two-horizon attribution system: (1) last-touch or first-touch for demand-capture channel optimization (2-week to 3-month window), and (2) marketing mix modeling or media contribution analysis for brand investment decisions (12-24 month window). Never cut brand investment based solely on last-touch attribution — it is structurally incapable of measuring brand's contribution. In practice: protect a 'brand' budget line item of 20-40% of marketing spend from quarter-to-quarter reallocation, and evaluate it on a 4-quarter rolling basis rather than a single quarter.