AI made analysis faster to produce. Choosing the right question still required judgment. One portfolio company created more than a hundred dashboards within six months of adopting Claude. None were aligned, people chose whichever view they preferred, and the business made poor decisions from polished reports. Teams that understood the business could use AI to move faster. Teams without that understanding could scale their mistakes. Judgment did not get cheaper.

One founder described spending without accountability as betting with fake casino money. His approach: walk the team through what is happening in the business and how the numbers were derived, while keeping major allocation decisions with the people accountable for the outcome. Access to information and authority to commit budget need separate rules. The team should understand the evidence well enough to challenge a recommendation. The decision owner should be able to explain the investment and carry its consequences.

Retail media reports showing returns of 35 or 45 drew a pointed response from the room: returns that large should be unmistakable in the business. In one Amazon account, platform-reported return, incrementality, and ROI did not track together. An agency reallocating budget from the platform report alone could therefore improve reported ROAS without establishing whether the move generated additional profit. Before a number earns the next dollar, it needs to reconcile with the economics of the business.

A category average can conceal products with very different economics. One CPG engagement modeled roughly a hundred Amazon products individually, covering about 80% of Amazon revenue. A 10% price change could produce a 40% difference in demand, with responses varying across products. Reallocation scenarios were evaluated against both revenue and margin. That distinction belongs at the start of a measurement program. The Marketing Accounting Framework (MAF) organizes measurement around the products, customers, and markets that create value, so the model answers the financial question the business needs to resolve.

The useful question goes beyond whether CTV works. It asks whether moving a specific budget from search into CTV improves the business outcome. One program modeled shifting 30% of search spending into CTV, then designed tests around the proposed allocation. Smaller tests supported the move, but leadership wanted more evidence. The team split the country in half, comparing the existing media mix with the proposed mix. The result came in below the model’s projection but supported continued reallocation. The test gave leadership evidence for a specific budget decision.

Last-click reporting undercredits the media that creates demand and rewards the channels that capture it. One CMO described letting his CFO cut upper-funnel investment and redirect spending to search and Meta. Brand share fell, and the business became harder to grow until upper-funnel investment resumed. That experience, reinforced by repeated incrementality tests, helped build finance’s confidence. CTV budgets have since grown roughly fourfold. Total upper-funnel spending still represents about 5% to 6% of the marketing budget.

One brand built its program without withholding media from operating locations. It compared different spending levels across markets, then used natural experiments and comparisons over time as the program expanded. An initial CTV test reported roughly 14% incremental lift in total bookings and an estimated incremental acquisition cost of about $570, compared with a platform CPA of $400. The operator considered that result positive against a first-transaction breakeven requirement, citing a $2,000 product with low marginal cost. The team then repeated the test to validate the result before expanding further.

Sales concentration alone does not tell you where another dollar will produce growth. Virginia Food Group used Geospatial analysis to identify opportunities outside its usual markets, then ran a six-week CTV test. Alongside the early test findings, the team reported roughly 52% overall August growth for Virginia Diner, compared with 12% to 15% across sibling brands. Those business results provided context beyond the campaign readout. The allocation lesson: assess markets for their remaining opportunity, rather than directing more budget wherever sales already concentrate.

At one SaaS business, survey responses citing answer engines rose from 5% to 25% over 18 months, while analytics reported 4%. The speaker described answer engines displacing organic discovery as business performance weakened and reporting assigned more credit to direct, paid, and other channels. That changed the diagnosis. Before treating weaker performance as a media problem, the team needed to understand how buyers were finding the business. Customer-reported discovery and analytics offered different views of a shift that allocation decisions needed to reflect.

The closing panel challenged the instinct to sell a recommendation with more certainty than the evidence supports. Explain how confident you feel, what supports that judgment, and the spending range you want finance to consider. Involve finance before the final budget request, with agreement on what a successful test would justify next. Then match the rigor to the consequence of getting the decision wrong. A reversible budget test and an irreversible commitment warrant different levels of scrutiny. Spend the most effort on decisions the business cannot easily undo.