Get in touch
Log in
Measurement only matters if it changes what you do next.
That sounds obvious, but there's a common failure mode: teams invest in dashboards, models and reporting, then stop short of using those results to reallocate spend. The consequence is straightforward. As Tim Kreienkamp, VP of Measurement and Data Science at Funnel, puts it: "Acting on the insights is what actually creates the value." Without action, there is no return on investment, because measurement has real costs whether in software fees or internal team effort.
If you don't act on the insights, the ROI on your measurement program is always going to be negative. That's not a moral argument. It's math.
So the real question isn't whether you have measurement insights. It's whether you can convert them into budget decisions confidently, and quickly enough to matter.
Three capabilities make that possible: campaign-level optimization, scenario simulation and AI-assisted incrementality testing. This post covers what each one does, why it matters and how they work together.
Why acting on measurement is the whole job
A measurement solution is an investment. Even a well-designed one that accurately explains performance still costs money to run. Without follow-through, the ROI equation never closes. Measurement fees and internal labor are fixed costs. If decisions don't change after you run the model, there's no incremental gain to show for it.
What "acting" looks like in practice is usually a budget decision: scaling down a channel that's underperforming, reducing spend on campaigns with poor efficiency, moving budget toward activity with a favorable cost per acquisition, or rebalancing allocation to reflect diminishing returns in saturated channels.
The actions don't need to be dramatic. Sometimes the right move is a small correction on a specific campaign. But without a system that makes those corrections routine, continuous and tracked, even the best measurement insights sit unused.
Campaign-level optimization: from channel estimates to real budget levers
Most organizations can model performance at the channel level. The problem is that budget decisions often need to happen at a more granular level: specific campaigns, ad sets or placements within a channel. If your model can only tell you "spend more on paid social" but not "spend more on this paid social campaign," you're still left guessing.
Marketing mix modeling has historically been strong at channel-level understanding. The challenge has been extending that rigor down to campaign-level effects. Without campaign-level saturation curves and coefficients, teams remain, as Tim puts it, "relatively stuck on a channel level for confident budgeting."
Funnel's campaign-level modeling addresses this by connecting campaign-level estimates to channel-level patterns through a technique called partial pooling, drawn from hierarchical Bayesian modeling. Campaign effects are estimated individually but anchored to what the model already knows about the broader channel. That anchor acts as a guardrail, preventing the model from fitting noise and generating unreliable recommendations. The model also incorporates informative priors: constraints derived from platform attribution data and incrementality testing results that give campaign-level estimates the stability needed to be actionable, even when data for a specific campaign is relatively sparse.
The result is marginal CPA and marginal return on ad spend calculated at the campaign level, not just the channel level. That granularity is what makes optimization operational.
In Funnel's media planner, this translates directly into execution. You set an optimization strategy (maximize conversions, minimize spend under a CPA threshold, maximize revenue), choose a timeframe and run the optimization. The key new capability is campaign-level budget constraints: rules for specific campaigns, not just channels or sources. Budget caps, minimums or both, set precisely where the model has identified the most actionable signals.
Tim makes the point plainly: decisions without follow-ups are only daydreams. So the system doesn't just produce a plan. It tracks execution nightly, showing whether spend for each campaign is trending above, on or below plan. That visibility lets teams course-correct before overspending accumulates or potential revenue is lost.
Scenario simulation: stress-test before you spend
Even a strong measurement model carries uncertainty. Scenario simulation is how you manage that uncertainty before committing real budget.
Funnel's What If module (currently in beta) lets you test questions inside the model before acting on them: what happens to conversions and CPA if we increase spend on this channel by 50%? What if prices go up 25%? The module starts with model fit statistics so you can assess how much confidence to place in the outputs, then returns predicted outcomes you can compare against actual observed performance over the same period.
Two examples from Tim’s walkthrough show what this looks like in practice. When a 25% price increase was simulated, conversions dropped roughly 4.1% and CPA worsened by about 4.3%. The direction matches business intuition; the model quantifies the magnitude. When spend was increased in a heavily saturated TV channel, conversions rose only about 1% while blended CPA deteriorated meaningfully. Both scenarios are useful precisely because they reveal where marginal returns are weak before any money is spent.
Scenario simulation also works through Funnel AI, the in-product AI available to all Data Hub customers. Instead of configuring scenarios manually in the interface, you can ask directly: "What would happen if I increased this channel's budget in January?" The system runs the same underlying simulation and returns the results in context.
The practical benefit: you reduce the risk of spending more where returns are already diminishing, without discovering that through live spend.
AI-assisted incrementality testing: design experiments that produce signal
Measurement insights become most credible when backed by causal evidence. That's the role incrementality testing plays. But a poorly designed test can produce weak signal even when you spend to run it: wrong regions, wrong duration, insufficient expected effect size.
Funnel's AI-assisted incrementality workflow guides the test design so teams don't have to navigate that complexity alone. The goal of a GeoLift test, where you select treatment and control regions and measure the incremental effect of an intervention, is to find a configuration that produces detectable signal at a reasonable budget and duration. The underlying algorithm tests many combinations to minimize both. Tim summarizes the objective: a test that is "relatively low budget and relatively short time," without sacrificing the statistical power needed to detect real lift.
Instead of filling out a dense input form and interpreting the output alone, teams can start from a conversation. The AI checks data health, requests any missing assumptions, then generates multiple simulated configurations based on your channel, expected lift size and timeline preferences. You review the options, select a configuration and run the test.
The benefit isn't that AI replaces the judgment involved in incrementality design. It's that it surfaces the relevant variables, speeds up the path to a usable configuration and reduces the risk of running a test that can't answer the question it was designed to answer.
From insights to decisions
Campaign-level optimization tells you where to allocate budget. Scenario simulation tells you what will likely happen if you do. AI-assisted incrementality testing helps you prove it and build causal credibility for the next decision.
Together, they close the gap between knowing and acting. Measurement only creates ROI when the insight reaches a decision-maker with enough specificity and confidence to move budget. That requires more than a model. It requires a system designed to translate outputs into operational plans, track whether those plans are being executed and support the experimental work that builds conviction over time.
The goal isn't better reporting. It's better budget decisions, made faster, with more confidence and with less risk.
Watch the full webinar on demand
Tim Kreinkamp walks through each of these capabilities live, including a dashboard demo of campaign-level marginal CPA, a scenario simulation showing how a price increase affects conversions and a real-time incrementality test design with AI guidance.
Ready to make measurement pay off?
Funnel Measure gives you the granularity and confidence to act on your data, not just report on it.