Vesta Software — Marketing-mix that earned a 14% budget shift
A bespoke marketing-mix model proved that brand investment was 2.4x more efficient than paid search at Vesta's stage. The reallocation produced 31% more SQLs at flat spend.
Vesta Software's marketing budget had drifted into 73% paid search and paid social over three years. The CFO had stopped questioning the mix because the attribution dashboards always showed strong last-click ROI from search. Pipeline, however, had flat-lined for six quarters. The CMO suspected brand was carrying the demand engine but couldn't prove it inside the existing measurement framework. The brief: build a measurement system credible enough to move budget — not just describe it.
Why last-click attribution quietly bankrupts B2B brands
B2B buying cycles span 9–14 months and 11+ stakeholders. Last-click attribution credits the final touchpoint with 100% of the value and discounts everything that built awareness, consideration and short-list inclusion to zero. The systematic result: brand and upper-funnel investment look unprofitable on the dashboard, get cut, and pipeline collapses 6–9 months later with no obvious culprit.
Marketing-mix modelling (MMM), once a CPG technique, has become the only credible alternative for B2B teams above ~$50M revenue. Modern MMM can be built in 6–8 weeks on 18–24 months of weekly spend and outcome data, runs continuously, and produces channel-level efficiency curves that survive a CFO review.
The hardest part is not the maths. It's the organisational courage to act on the answer. Most MMM exercises produce results that contradict the last-click dashboard the team has been optimising against for years. Teams that don't have leadership cover to move budget waste the model.
Average efficiency of brand vs paid search at mid-funnel SaaS stage
WARC B2B Effectiveness 2024
Typical lag between brand cuts and pipeline collapse in B2B SaaS
LinkedIn B2B Institute / IPA
Of B2B marketers report their attribution under-credits brand
Forrester 2025
"Last-click attribution had been quietly draining Vesta's pipeline for two years. The dashboard looked healthy. The pipeline was not."
How we actually ran it.
Data assembly
Build the spine the model will run on.
Pulled 24 months of weekly spend by channel (search, social, display, content syndication, events, sponsorships, brand campaigns, organic), plus outcomes (MQLs, SQLs, opportunities, closed-won). Joined to macro covariates: category search volume, competitor share-of-voice, seasonality, salesforce headcount.
A single tidy dataset, 104 weeks × 38 columns, with documented lineage.
Model specification
Build a Bayesian MMM that respects B2B reality.
Used PyMC with adstock and saturation curves per channel, hierarchical priors borrowed from the WARC B2B benchmark, and a 9–14 month decay on brand spend. Validated against a held-out 12-week window and against a historical paid-media blackout the team had inadvertently run in Q2.
Posterior efficiency curves per channel, with 80% credible intervals. Brand efficiency: 2.4x paid search at current spend levels.
Scenario planning and CFO walk-through
Translate posterior distributions into a budget recommendation a CFO will sign.
Built three scenarios: hold (current mix), shift-10 (10% from search to brand), shift-20. Modelled expected SQL impact with credible intervals over 6, 9, 12 months. Walked the CFO through assumptions, sensitivities and known unknowns.
Recommendation: 14% shift from paid search to brand + content syndication. Approved.
Activation and monitoring
Move the budget and watch the model in production.
Phased reallocation over 8 weeks to avoid step-change noise. Weekly model refresh with new spend and outcome data. Drift alerts if any channel's posterior moved more than one credible interval.
By week 16, SQL volume +31% at flat total spend; CAC down 19%.
Operating cadence
Make MMM part of the planning rhythm, not a one-off exercise.
Monthly model refresh, quarterly scenario re-plan, annual full re-specification. Standing seat for MMM in the CFO's quarterly business review.
Marketing planning shifted from last-click reports to mix scenarios. The dashboard war ended.
The chronology of the work.
- Day 0
Brief: prove or disprove that brand is being under-funded.
- Day 21
Data spine assembled. Two months of missing event-spend data reconstructed from finance records.
- Day 42
First model run. Brand efficiency 2.4x paid search at current spend.
- Day 56
CFO walk-through. 14% reallocation approved on a 90-day test basis.
- Day 84
First SQL lift visible. +12% over baseline.
- Day 112
SQL volume +31%, CAC −19%. Reallocation made permanent.
- Day 180
MMM running monthly. Vesta retired three legacy attribution dashboards.
The three places most programmes die.
Building MMM without CFO sponsorship
Bring finance into the first workshop. A model the CFO has not co-designed is a model the CFO will not act on.
Ignoring adstock and decay
B2B brand pays back over 9–14 months. A model without decay will systematically under-credit it and rebuild the bias you are trying to remove.
One-and-done MMM
Markets shift. Channels saturate. Treat MMM as a continuous monthly process, not an annual report — the second-year model is where the real ROI compounds.
A bespoke Bayesian MMM proved brand was 2.4x more efficient than paid search at Vesta's stage, earned a 14% budget reallocation, and produced a 31% lift in SQL volume at flat total spend within 16 weeks. The CFO retired the last-click ROI dashboard and made MMM the company's primary marketing planning tool.
SQL volume at flat total spend
CAC reduction in 16 weeks
Budget reallocated from search to brand
We'd been optimising the wrong number for two years. The MMM didn't just change the budget — it changed how marketing and finance talk to each other.
Lessons we'll carry into the next programme.
Last-click attribution is the single most expensive habit in B2B marketing. Replacing it pays for itself inside a quarter.
MMM is a CFO product as much as a marketing product. Co-design or lose the budget fight.
Adstock and saturation curves are non-negotiable for B2B brand investment.
Continuous refresh is where MMM compounds — annual exercises produce reports, not decisions.
Always validate against a real-world blackout if you have one. Synthetic validation flatters the model.
You cannot move budget on a dashboard your CFO doesn't believe. Build the measurement system that earns the right to be acted on — then act.
Want a programme like this one?
Tell us where you want pipeline to come from next quarter — we'll show you how the next 90 days could look.