Marketing mix modeling for finance: where 73% of new customers really came from

The challenge

For direct financial services brands, short-term ROI drives the Brand to Performance debate, and short-term ROI systematically under-prices Brand investment. In other words, the problem is one of measurement grain. A single-equation model collapses the funnel into one sales line and attributes conversions to whichever channel ran closest to the sale; as a result, the upstream brand-building effect flows into the baseline as unexplained organic growth, and the next quarter reads as a Performance win. This case study describes how a direct financial services brand corrected the gap by modeling four funnel KPIs as separate MMM equations, chained from Consideration through to New Customers, and what the chain revealed about where its media really worked.

A single equation collapsed the chain

The previous framework ran one MMM against New Customer conversions. Channel rankings were credible; however, the baseline absorbed every conversion attributable to Brand investment in the previous twelve to twenty-four months as organic growth, and the model had no way to say so. On a same-period lens Performance ROIs sat well above the Brand channels; consequently, the reallocation implied by that reading pointed in one direction every quarter: shift from Brand to Performance. The team suspected the implied direction was wrong; however, it could not refute it with the model they had.

Paid media and effective base at each funnel step

Share of each funnel KPI, same period

Paid media shareEffective base share

Consideration

69%

31%

Brand Search

72%

28%

New Customers

27%

73%

Four measurement gaps in the single-equation framework

• Long-term Brand carryover absorbed into the baseline. Brand channels build over the following twelve to twenty-four months, but same-period attribution captured none of that, so the carryover flowed into the baseline.
• Brand and Performance framed as a trade-off, not a chain. A single-equation model cannot say that Consideration drives Brand Search, Brand Search drives Applications, and Applications drive New Customers. Instead, each quarter pitched the two against each other as if substitutes.
• The effective base treated as exogenous. The brand’s own equity grew with its investment, but the model had nowhere to put that growth. Consequently, it became drift in the base rather than the lagged outcome of years of upstream media.
• Optimization against conversion ROI alone cannibalized Brand. An optimizer pointed at the bottom of the funnel shifts budget away from the channels that build the top. Moreover, the model had no mechanism to protect the upstream investment that fed conversion.

The rebuild therefore addressed all four by naming each funnel step as its own MMM equation. Table 1 sets out the three equations closest to the commercial outcome.

The solution

MASS Analytics rebuilt the framework on the structural basis a brand-led direct business requires: four parallel MMM equations, one per funnel step, chained so the output of each upstream equation feeds the next. Consideration. Brand Search. Applications. New Customers. Each equation decomposes its KPI into a paid-media contribution and an effective base; in turn, the base itself decomposes into structural equity, market trend, long-term media carryover, price position, competitor pressure and seasonality. Overall, the framework spans three years of weekly data on a Monday to Sunday grain across all four funnel stages.

Each funnel step was modeled on its own equation

First, the team modeled Consideration against a brand tracker index with hierarchical pooling, so structural equity and market trend sat separately from paid media. Brand Search entered as owned-term query volume, with the Consideration output fed in as a driver so the lift that paid media created upstream was credited at the next step rather than lost in the Brand Search baseline. Next, the Applications equation used completed intent forms: the mid-funnel signal that a prospect has entered the purchase journey without yet committing. Finally, the New Customers equation used confirmed conversions, with long-term adstock carrying the Brand effect twelve to twenty-four months downstream. A two-stage optimizer then anchored a Branding floor on Consideration and maximized New Customer revenue under that floor, and the same configuration supports scenario modeling, budget optimization and forecasting. The client owns the model, which runs continuously as an always-on MMM.

The data foundation made the chain modelable

The Consideration index came from a continuous brand tracker. Similarly, Brand Search was built from owned-term query volume on a defined keyword set, de-duplicated against navigational traffic. Meanwhile, Applications and New Customer counts came from the CRM, reconciled against the media-buying teams’ delivered impressions and spend per channel on the same week. In addition, external data covered the category search index, a competitor pressure proxy and macro indicators. Metric alignment at receipt was the decisive enabler: a four-equation chain fails if one week drifts out of alignment at one step.

Results and Impact

The rebuilt framework produced four findings that reshaped the next planning cycle. In particular, of the New Customers acquired over the period, 73% came from the effective base built by prior media rather than from the paid media that ran in the same period, the single largest reattribution the chain produced.

The chain revealed where the base sat

The Consideration equation attributed 69% of index movement to paid media; most of that media sat in channels run as Branding. Likewise, Brand Search attributed 72% of volume to paid media, with a strong lagged correlation to Consideration: Brand Search is not a standalone demand source, it is the measurable spillover of upstream Brand investment. At New Customers, however, the pattern flipped: paid media in the same period accounted for 27% of conversions; the remaining 73% sat in the effective base. Specifically, 26% of total New Customers were attributable to long-term Brand carryover, the mechanism the single-equation framework had absorbed into the baseline. As a result, Branding ROI with that carryover included came in at 4.4x.

$3.1M reallocated at flat media spend

The two-stage optimizer anchored a Branding floor of $13M on a twelve-month $34M media envelope by maximizing Consideration first, then maximizing New Customer revenue under that floor. In turn, the optimizer produced $3.1M of media-to-media reallocation away from channels close to saturation towards channels with response-curve headroom, at flat total spend. Consequently, portfolio ROI rose from 2.6x to 3.0x, projecting +15% revenue at the same budget. Throughout, the plan preserved DRTV, the single channel carrying the chain from upstream to downstream.