Gavin Trippe

Someone must own follow-through: an agency cannot control budgets it does not manage.

In one post, Trippe describes a vendor receiving day-by-day spend recommendations but not executing them. After the quarter missed its target, the brand questioned the model; Trippe’s first proposed check was whether the dollars had gone out as planned.

Read original

They recommend checking broader evidence, such as an MMM or incrementality results, before reallocating spend.

A Shopify client’s fast-converting product looked stronger in Meta than a higher-value product with a longer conversion time. Trippe says an MMM showed the slower product was driving more revenue, so the team restored its budget despite weaker in-platform attribution.

Read original

Trippe argues for channel-level high and low estimates, with a specific explanation when results fall outside the range. Longer-term forecasts should reflect uncertainty in budgets, promotions, and launches.

In a staged CMO–CFO exchange, a 2.4 blended ROAS does not answer what a $10 million monthly budget might return by channel. In another dialogue, Trippe suggests showing a board a forecast both with and without an unsettled clearance sale rather than quietly carrying its revenue forward.

Read original · Original 2

The plan should account for conversion lag: build reach with slower channels early, then shift toward remarketing as the sale approaches.

For Black Friday and Cyber Monday, Trippe contrasts Meta campaigns that can affect revenue within zero to seven days with TV, audio, and podcasting effects that may take 60 to 80 days. The post recommends starting longer-lag prospecting earlier and using remarketing closer to the sale.

Read original

Trippe argues that someone must connect test results and models to a channel-level recommendation, implement the change, and check whether spend followed it. They make a related distinction about AI: running an MMM is easier than validating it enough to trust with major budget decisions.

Trippe describes changing agencies after neither could turn MMM and incrementality results into channel decisions.

• Their proposed test is to walk through a recent lift test, compare it with the MMM, review the resulting budget recommendation and campaign change, and check the following week.
• In a separate post, they caution against moving $5 million on an unvalidated AI-built model.

Read original · Original 2

Start here because the order of checks matters. Trippe rules out spend adherence before anyone touches attribution.

When a Shopify brand misses its revenue forecast, Gavin Trippe first compares every media partner’s actual spend against plan, a check they run weekly for clients.

• In their telling, someone is almost always off plan, usually the partner in the least contact with the brand.
• It is the vendor skipping calls and strategy reviews, yet still getting a free pass.
• Trippe describes clients sending another vendor a day-by-day spend recommendation broken out by tactic, only for it to go unexecuted.
• The quarter closes, revenue misses, and the conversation turns to what was wrong with the model.
• Their argument: a forecast describes the plan, so if the dollars never went out that way, you cannot compare the result to the forecast.
• Someone inside the brand has to own keeping partners to plan; an agency only controls the budget it manages.

For brands planning a campaign

A brief for this work should ask for a spend-versus-plan audit across the full media mix and name the internal owner responsible for adherence, rather than starting with attribution or model changes.

A short dialogue: the CMO reports a 2.4 blended ROAS and says marketing is working; the CFO asks which channels drive it and what the $10M being spent next month returns by channel.

• Trippe says the CFO isn't asking for much.
• If they are giving you $10M for the month, they want to know where every dollar goes and what to expect back per channel, with a high and low bound on each: Meta between this and this, Google between this and this, CTV less certain but here's the range.
• Hit the ranges and you are their best friend.
• Miss them and that is fine too, if you explain, for example that Meta missed the low end because the audience saturated in week three and frequency spiked.
• Come back with "it didn't work" and you lose the room and the budget.
• Trippe notes having spent years brand side in exactly this meeting, and points to a podcast episode where this was discussed.

For brands planning a campaign

Frame deliverables as channel-level ranges with stated assumptions, plus a written explanation when a range is missed, rather than a single blended number.

A Shopify client sells two products: one with a higher average order value and a longer time to conversion, and one cheaper product that converts quickly.

• In Meta, the fast one looked like the clear winner and the high-AOV one looked terrible, so spend shifted to the fast seller and came down on the slower one.
• Meta cannot see past seven days, so when one product converts quickly and another takes longer, the fast one can look more efficient in-platform even when the slower one is doing more for the business.
• The damage is delayed: everything looks fine for a while, then sales slip with no obvious cause.
• An MMM showed the product that looked terrible was driving more revenue—and was the one spend had just been pulled from.
• Trippe flipped the budget back, knowing attribution would look worse, and the decline turned into a positive comparable.

For brands planning a campaign

Briefs should flag products and channels with unequal conversion lag and ask for an MMM or incrementality cross-check before cutting spend on slow converters, even when the in-platform dashboard disagrees.

Brands buy every multi-touch attribution tool, then every lift test tool, then an MMM.

• The agency agrees to use them.
• Still, someone internally has to translate the readouts into budget changes and make sure the buying team follows through.
• Trippe describes this from the brand side: they changed agencies because the first couldn’t turn MMM and incrementality results into channel-level budget decisions.
• The second had the same problem: the software was available, but nobody owned the measurement work behind the media plan.
• The proposed test: an agency claiming to handle both measurement and media should walk through one recent lift test, from the readout to the MMM comparison, budget recommendation and campaign changes, then show how it checked the following week whether spend followed that recommendation.

For brands planning a campaign

Use the walkthrough as a scoping question for a campaign that needs an end-to-end example from a recent test, plus a follow-up showing that spend matched the recommendation.

Don’t wait until Black Friday to ramp media spend and expect incremental revenue that same weekend.

• A brand’s biggest revenue days don’t mean every channel should carry its biggest spend on those dates.
• That works for remarketing campaigns that convert inside the sales window, but not for prospecting TV, where people reached need weeks to decide.
• Meta conversion campaigns affect revenue within zero to seven days; TV, audio and podcasting carry effects over 60 to 80 days.
• The product and campaign set the timing, so spend has to happen early enough for purchases to land during the sale.
• Trippe recommends starting with the long-lag channels now, tapering prospecting as the sale approaches, and shifting toward remarketing the audience already reached, with spend sized around the audience built and the frequency needed.
• Beyond that point, Google will gladly take more budget on the day and report a 10-to-1, but total revenue won’t scale with it because most of that demand was built earlier.

For brands planning a campaign

Seasonal briefs should specify channel lag and phase the plan into build, taper and remarket stages rather than concentrating spend on the sale date.

A CFO needs a 2027 board forecast; a CMO needs a media budget and blended ROI target before producing one.

• The CMO asks whether this year’s results can be the starting point, then notes that they depend on what will be repeated: if this year’s revenue included an inventory-clearance sale that won’t run again, carrying that revenue forward would overstate next year’s plan.
• If the board meeting falls before plans are agreed, the CMO suggests using a proposed budget and calendar and showing a contested decision, such as that sale, both with and without it, so the board sees how much depends on it happening.
• On confidence, the CMO is cautious around six months out even with spend and calendar agreed, and gives wider monthly revenue ranges further out, especially when promotions or launch dates are unsettled.
• The exchange closes with a straightforward swap of inputs: finance provides a budget scenario and ROI target, marketing proposes promotions and launches, and revenue expectations are updated as plans are agreed.

For brands planning a campaign

Ask for forecast outputs tied to named assumptions, promotions, launches and a budget scenario, with ranges rather than a single number and one-off items separated out.

A debugging joke opens the post: an AI assistant calls something the smoking gun, folds when challenged, then calls a new answer the smoking gun, apparently forgetting it just changed position.

• Trippe applies the pattern to AI-powered MMMs and the frequent question of whether a brand can build its own MMM on a general model, asking who would take those numbers to a CFO.
• They say they are no AI skeptic and run AI coding tools daily; the concern is the confidence-then-capitulation pattern, not the tools themselves.
• Running an MMM was never the hard part; one can be assembled in a spreadsheet fairly quickly.
• The challenge is validating it and trusting it enough to move $5M on its recommendations.
• For now, AI belongs on top of your data: answering questions, running optimizations through an API, interpreting results and digging—not holding the wheel or the money.

For brands planning a campaign

For a campaign whose proposed approach includes AI tooling, require evidence of how the model was validated before any budget decision depends on it.

In their framing, the model forecasted the plan, so if the dollars never went out that way, the result cannot be compared with the forecast.

• Partners most often off plan get the least scrutiny, while vendors in weekly calls tend to stay close to budget.
• They also point out that translating a spend recommendation into execution requires someone inside the brand to own the follow-through; an agency can control only the budget it manages.

Compare actual and planned spend across the full media mix, identify which partners drifted, and examine the model only after spend matches plan.

For brands planning a campaign

Useful for a brand that suspects a measurement or forecasting problem but has not verified whether agencies and vendors executed the approved budget. The diagnostic sequence offers a concrete first step: check whether spend matched plan.

Examples: Source 1 · Source 2

Gavin Trippe argues that hitting those ranges earns trust, and that missing one with a specific explanation, such as audience saturation in week three, still preserves credibility.

• Coming back with "it didn't work" loses the room and the budget.
• A companion dialogue covers annual revenue forecasting, where promotions and launches affect what can be promised and confidence should narrow to about six months before widening into monthly ranges.

Present per-channel low and high bounds, tie each to budget scenarios and planned promotions or launches, and widen the ranges as the planning horizon extends.

For brands planning a campaign

Fits a brand that needs its marketing team to give finance defensible channel-level numbers rather than an aggregate return figure. The approach shapes how a measurement engagement would be scoped and reported.

Examples: Source 1 · Source 2

Meta's in-platform reporting made the fast product look like the clear winner, so the brand shifted spend toward it, while the slow product's longer lag meant the damage did not show up immediately.

• An MMM revealed that the product that looked poor in-platform was actually driving more revenue, and the team restored its budget knowing attribution would look worse.
• The warning: don't treat a platform dashboard as a fair comparison between products with different conversion timelines.

Compare platform attribution with MMM or incrementality results before cutting spend on a slow-converting product, and accept weaker in-platform numbers when aggregate evidence supports the slower item.

For brands planning a campaign

Relevant for catalog brands where several products convert on different timelines; the example shows the evidence Gavin Trippe would bring to a budget reallocation argument.

Examples: Source 1

Meta conversion campaigns pull revenue forward within zero to seven days, while TV, audio, and podcasting carry effects over 60 to 80 days.

• The recommendation: start long-lag channels early, taper prospecting as the sale approaches, then shift toward remarketing the audience already reached so the offer gives them a reason to convert.
• Size remarketing around the audience built and the frequency needed.
• The post also notes that Google will absorb extra budget on the day and report a strong return while total revenue does not scale with it.

For each channel, decide when spend must happen for purchases to land inside the sale window, then reallocate toward remarketing once long-lag reach has been built.

For brands planning a campaign

Directly applicable to seasonal planning for a Shopify brand that allocates budget by sale date; it frames timing as a measurable planning decision rather than a creative one.

Examples: Source 1

The argument is that running an MMM is not the hard part: spreadsheet versions exist, and AI can run one too.

• The hard part is validating the model and trusting it enough to move $5M based on its output.
• In their view, AI belongs on top of existing data, answering questions, running optimizations through an API, and interpreting results—not holding the wheel and the money.

Separate the mechanics of running a model from validating it and deciding whether to act on it; keep AI in an analytical layer above validated data.

For brands planning a campaign

Sets expectations for evaluating a measurement proposal: a brand seeking a defensible MMM should expect validation and trust-building work, not just model construction.

Examples: Source 1

They recount changing agencies twice on the brand side because nobody owned the measurement work behind the media plan. Their standard: an agency should be able to walk a client through one recent lift test, from the readout to the MMM comparison, budget recommendation, and campaign changes, then show the following week whether spend actually followed that recommendation.

Require a documented chain from a single lift test through the MMM comparison, budget decision, campaign change, and follow-up adherence check.

For brands planning a campaign

Gives a brand a concrete test for assessing a measurement and media partner, and addresses the ownership gap Gavin Trippe flags in forecast and budget discussions.

Examples: Source 1 · Source 2

The post links to the full episode of the Selling on Giants podcast.

View example

Paxton's sales-drop checklist checks tracking, store conversion, ad-account changes, and seasonality before blaming ads. Their home-goods case looks beyond blended ROAS to contribution margin and retargeting concentration, overlapping with Trippe's arguments about spend adherence and attribution.

Paxton's examples focus on account operations and client turnarounds; Trippe spends more time on forecast governance, channel-level finance communication, and validating marketing mix models before large budget decisions.

Read original · Original 2

Livermore argues that AI search is changing discovery and that authority must be built before traffic declines.

• She also explains why a promotion can convert well yet damage contribution margin when costs and inventory economics change.
• Her focus on the economics behind a seemingly good dashboard result parallels Trippe's warnings about blended ROAS and platform attribution.

Livermore focuses on discovery, trust, and promotional economics rather than media measurement ownership, spend adherence, or marketing mix model validation.

Read original · Original 2

Wittick's seeding example says an unboxing ad from a small creator beat a Joe Rogan clip by 8x, making the case for volume and organic creative over big names.

• Their branded-Shopping audit argues that new and returning customers must be separated before judging acquisition.
• Both creators urge readers to verify the underlying metrics before scaling or reallocating.

Wittick focuses on creative sourcing, product seeding, and Google customer classification; Trippe focuses on media measurement, forecasting, spend adherence, and marketing mix modeling.

Read original · Original 2

The posts show two relevant situations.

• First, a sudden sales drop that needs a fast, ordered diagnosis: check whether tracking is broken (for example, Shopify, GA4 and Meta disagreeing by more than 10%), then whether the store is converting, then what changed in the ad accounts, then seasonality and outside context—and isolate one culprit rather than four.
• Second, a Meta account where blended ROAS looks acceptable but the business is not improving: Paxton's published case is a home goods client spending $120K a month at 2.8x blended ROAS and under 12% contribution margin, with 65% of Meta budget retargeting the same 14-day window.
• The changes described were reducing retargeting to 25% and pushing the rest into Advantage+ prospecting, building a post-purchase sequence, raising the free-shipping threshold after checking average cart value, and cutting weak SKUs from the Shopping feed.
• The reported result was $197K in month four, $280K by month nine, and 23% contribution margin.
• That is the right match when the failing lever is in-account execution and per-order economics, and success can be measured in weeks by margin and revenue.

The work shown is operator-side account turnaround and unit-economics diagnosis, not finance-facing measurement.

• Across both posts, there is no media mix model validation, lift or incrementality testing, audit of actual-versus-planned spend across outside media partners, or per-channel forecast ranges with low and high bounds.
• Those are the areas where Gavin Trippe's posts are specific.
• A brand whose miss traces to one partner failing to execute the plan, or to an unvalidated model, will not find that work in Paxton's posts; the approach starts and stays closer to platform reports.

Examples: Source 1 · Source 2

The post argues that a promotion can convert beautifully and still be a poor growth decision once tariffs, higher landed costs, fulfillment, media spend, borrowing costs, and the cash needed to replace inventory are counted.

• A 20% discount priced against 2025 margins comes out of a much thinner 2026 contribution margin; the business paid more, and often earlier, to bring the inventory in, while each discounted unit recovers less cash toward the next purchase order.
• That is the more useful starting point for a brand whose revenue held up but cash and margin did not, and whose question is whether the promotion should run at all on current economics.
• A second post covers customers using AI search to compare and shortlist providers, which matters if the brand also needs to know whether buyers are finding it through AI answers rather than its own channels.

This is offer-level and P&L analysis, plus discovery visibility—not channel media measurement.

• The posts do not cover media mix model or incrementality validation, actual-versus-planned spend by partner, or channel-level forecast ranges.
• The work answers whether a promotion still makes economic sense at current margins and whether the business is visible where customers now search.
• It does not answer where next month's budget should go by channel or how to defend a spend plan against a finance team's per-channel expectations.

Examples: Source 1 · Source 2

An audit of a roughly $10.8M DTC brand found that about 95% of the branded Shopping traffic examined had no useful new-versus-returning customer classification.

• That makes the cost of acquiring a new customer unanswerable and lets repeat purchases count as acquisition wins, making the channel look healthier on blended ROAS than its acquisition economics warrant.
• The recommended order is to confirm purchase events come from a clean source, separate new and returning customers where possible, keep customer lists current so Google is not working from stale data, and judge prospecting against new-customer economics rather than blended platform ROAS.
• That is the right first move when a brand suspects its reported acquisition performance is inflated by existing customers.
• A second post covers flooding Q4 with creator content—product seeding to hundreds of micro creators with no brief, capturing untagged posts, locking in usage rights, and running the best posts as ads through the creator's handle—which is relevant if the bottleneck is creative supply for paid social rather than measurement.

The posts cover one platform's data hygiene and creator-led creative volume, not the whole assignment.

• They do not show spend-adherence audits across media partners, media mix model or lift-test validation, or finance-facing channel forecast ranges.
• Choosing Wittick for a forecast miss narrows the work to fixing customer classification in Google Shopping or producing creative at volume.
• The brand would need another owner to assess which partners executed the plan, whether a model can be trusted, and what each channel should return next quarter.

Examples: Source 1 · Source 2