👋 Hi, it’s Rick Koleta. Welcome to GTM Vault - a breakdown of how high-growth companies design, test, and scale revenue architecture. Join 26,000+ operators building GTM systems that compound.
The ad shipped. The creative looked great. The campaign went live. Performance still stalled. Not because targeting failed, not because the budget was wrong. The team scaled ad volume faster than its ability to learn. The feedback loop between the ad account and the people making the ads never closed.
George Howes built Magicbrief to close that loop. Before Magicbrief, George spent years inside top creative agencies watching the same failure repeat at every brand. Teams launched ads, collected results, and moved on. Learning never compounded. Magicbrief was the first attempt at a creative intelligence platform that ingested ad performance and pushed it back to the teams making the ads. It worked. Canva acquired the company. George now leads Canva Grow, the product Canva is building to make the loop operate at enterprise scale.
Canva Grow ingests ad performance from every platform, ties each result back to the specific creative asset that produced the outcome, and pushes those learnings into the next round of creative before the next round of creative is made. The architectural premise is simple. Creative and media are not separate disciplines. They are the same system measured at different time scales. Every company that still treats them as two is losing compounding efficiency at every handoff.
In GTM 45, George breaks down why most performance marketing failures are not media problems but feedback loop failures, why creative teams and data teams sit at opposite ends of a broken handoff, and what changes when the learning loop runs in days instead of quarters. He explains why AI is a multiplier for creative feedback and not a replacement for it, why click-through rate lies more often than any other metric in the stack, and why small teams with all context in one place consistently outperform large teams with fragmented tooling.
This is not a conversation about better ad tools.
It is a conversation about why creative and media are the same system, and what happens when you design the feedback loop between them as infrastructure instead of a monthly report.
Inside this episode
This episode maps the structural gap between how most performance marketing teams run today and what the feedback layer has to become when ad volume scales faster than learning, when AI compresses creative production to minutes, and when the buyer never sees the test matrix you ran.
George starts with the pattern he kept seeing from agencies. The creative team and the data team never sat together. Creatives made the ads. Analysts read the ad manager. The handoff between those two functions was weekly meetings, slide decks, and aging context. The people responsible for the next round of creative were almost never the people looking at the dashboards that showed what the last round produced. The learning never landed. Agencies delivered work that was, by design, detached from the context that would have made it better.
We go deep on why creative became the bottleneck, not spend or distribution. A decade ago, the constraint was production speed. Tools were expensive, crews were slow, iteration cycles were measured in weeks. That constraint is gone. Anyone can ship a variant in hours. The new constraint is knowing what to make. Volume is cheap, insight is scarce, and the teams that figure out what the account is telling them about their next creative outperform the teams that simply ship more.
We cover Canva Grow’s architecture in detail. The platform ingests creative, the product being sold, prior design history, brand guidelines, and live ad performance from every platform the company runs on. It puts that context in front of the teams making the ads, written in plain language. The contrast with the incumbent stack is architectural. Most performance stacks live in three different products with three different owners: the ad platform, the analytics platform, and the creative tool. Each has context the other needs. None of them share. Canva Grow’s play is to compress that into one surface where the context is always present and always current.
We go into the AI layer. George is direct on this. AI is a multiplier for creative teams that already have feedback loops running, and a noise amplifier for teams that do not. The failure mode is always the same: generate ten times more variants without adding a corresponding ten times to your learning capacity, and the signal degrades. The ads that work get buried. The ads that do not work get re-created in the next batch. Agentic layers only help when the loop underneath them is already closed. Without the loop, the agent is shipping faster against the wrong assumptions.
We cover why click-through rate lies. George’s framing is sharp. CTR can spike from a single hooky video where the creator pulled the viewer in without setting the right expectation. The click happens. The landing page converts at zero. The campaign metrics look good mid-funnel. The business loses money anyway. Downstream signals (add to cart, lead submitted, booking confirmed) are what separate a good ad from a curious ad. Every optimization that runs on CTR alone is running against the wrong proxy.
We go into the compounding thesis. One great static ad in a working feedback loop becomes a video ad on Facebook, a pre-roll on YouTube, an email hero image, and a paid social carousel. The asset is not consumed when it ships. It is compounded. Small teams inside Canva Grow move faster than enterprise teams with ten times the headcount because the context is in one place, the learnings transfer automatically, and the same asset gets worked across five surfaces before the enterprise team has finished its monthly retrospective on the first.
Discussed in this episode
0:00 Intro: why ads stall before spend does
2:20 What actually breaks underneath when creative "stops working"
4:30 Why creative became the bottleneck, not spend or distribution
6:15 How the bottleneck caps growth long before CAC spikes
9:30 Canva Grow: the context layer between creative and data
11:00 AI as multiplier, not savior
15:50 What good feedback looks like written out for the team
19:00 SMB versus enterprise: retargeting and accessibility
21:20 Performance metrics that lie
26:10 How one static ad compounds into video, YouTube, and email
31:40 Why small teams ship and learn faster
35:00 What enterprise can learn from SMB velocity
41:25 Rapid fire: the metric that lies most often in performance marketing
42:55 Closing: learning velocity is the new spend
Key takeaways
Creative and media are the same system, measured at different time scales The failure mode in most performance orgs is treating creative and media as two disciplines with a handoff between them. They are the same system. Creative makes the asset, media runs it, and the feedback from the run is what should shape the next asset. When the loop is broken, every round of creative is built against assumptions that were already out of date in the account. The fix is not a better brief. It is an architecture where the context that lives in the ad manager is visible to the people making the ads, in plain language, before the brief gets written.
Volume is cheap. Insight is scarce. A decade ago, the constraint on performance marketing was production speed. Tools were expensive, crews were slow, iteration was measured in weeks. That constraint is gone. Anyone can ship fifty variants this week. The new constraint is knowing which variant to invest into before running the test. Teams that figure out what the account is telling them outperform teams that simply ship more. Ad volume scaling faster than learning capacity is the silent tax at every growth stage. You only see it in CAC, months after the damage was done.
AI is a multiplier on a working loop, and a noise amplifier without one The teams winning with AI in performance marketing already had feedback loops running. AI compressed their iteration speed from days to hours. The teams losing with AI tried to generate their way out of a broken loop. Ten times more variants without ten times more learning capacity degrades signal faster than it produces output. Agentic layers amplify whatever is underneath. If the loop is closed, they compound. If the loop is open, they make the mess bigger at scale.
Click-through rate is the most dangerous single metric in the stack CTR can spike on a hooky video that pulls attention without setting the right expectation for the product. The click happens. The landing page converts at zero. The campaign metrics look good mid-funnel. The business loses money. Any optimization layer that runs on CTR alone is running against a proxy that does not correlate with revenue. The operational move is to pair CTR with at least one downstream signal (add to cart, lead submitted, booking confirmed) and drop any creative where CTR spikes without the downstream signal following.
One great asset, five surfaces. The compounding thesis. Most teams treat a winning ad as the end of the job. A great static in a working feedback loop is the start of it. The same asset becomes a video on Facebook, a pre-roll on YouTube, a hero image in email, a carousel in paid social, a hook in outbound. The compounding is not automation, it is deliberate propagation. Small teams with context in one place outrun enterprise teams with ten times the headcount because the propagation happens in days rather than quarters, and because every asset in the library inherits the learnings from every asset that came before it.
Small teams with context beat large teams with fragmentation The most counterintuitive signal in the episode. Teams of three or four with all the context in one place consistently outperform enterprise teams with dedicated specialists for every role. The reason is not talent. It is that in a fragmented stack, context leaks at every handoff. A creative director with no view into the ad manager is making decisions against last quarter’s reality. An analyst with no view into the creative brief is measuring against the wrong goal. Enterprise teams can match small teams on speed and learning only when they rebuild the context layer underneath.
Frameworks from the episode
The closed feedback loop as GTM infrastructure The architectural premise of Canva Grow. Ingest creative, the product being sold, prior design history, brand guidelines, and live ad performance from every platform. Make that context present in one surface. Put the surface in front of the people who make the next round of creative. The output is a loop where every new asset is informed by every old asset and every piece of live performance data from the account. The loop is not a report. It is the operating system for creative production. The specific failure mode it prevents is the one every enterprise team runs into: the people making tomorrow’s ads are not looking at yesterday’s data.
Creative diversity as an optimization variable Most performance marketing orgs optimize the single best-performing creative and ship tighter variants of it. The optimal move is the opposite. Run a wider diversity of angles, narratives, and hook structures, and let the ad account distribute spend to whichever performs. The algorithms underneath Meta, Google, and TikTok now optimize across angles, not within them. Shipping a narrow variant set of a single winning ad shrinks the surface the algorithm has to work with. A wider portfolio of structurally different ads gives the platform more axes to optimize and raises the ceiling of the campaign overall.
The asset compounding model One great static becomes a video ad on Facebook, a pre-roll on YouTube, an email hero, a paid social carousel, a landing page asset, and a sales deck slide. The asset is not consumed when it ships. It is the starting point for five other assets. Teams that treat the asset as terminal are optimizing the input. Teams that treat it as compounding are optimizing the output. The move is to make the creative team accountable not for ads produced but for surfaces the produced asset eventually runs on. The number shifts from one to five without shipping five times the work.
What to do this week
Audit the distance between your ad account and the people making your creative. If the creative team is more than one handoff away from the live performance data, the feedback loop is broken by design. The fix is not a better meeting cadence. It is a single surface where the performance data is visible to the creative team in plain language, before they write the next brief.
List the metrics your performance team optimizes against. For each one, identify whether it correlates to revenue or only to attention. If the primary input is CTR, impressions, or reach, the team is optimizing for a signal that can spike without producing pipeline. Pair every upper-funnel metric with a downstream signal, and drop any creative where the upper metric spikes without the downstream one following.
Count the number of tools between the creative brief and the live ad. If it is more than three, every handoff between them is a place context leaks. The architectural question is not which tool to buy next. It is which tools to collapse so the context lives in one place.
Before you add another variant to the campaign, ask what you learned from the last variant you shipped. If the answer is “it is still running” or “we have not pulled the numbers yet,” you are scaling ad volume faster than your ability to learn. The fix is not to ship less. It is to close the loop underneath the volume you are already shipping.
Why this matters
The performance marketing era rewarded volume. More ads, more tests, more spend. The playbook was simple because the constraint was simple. Get production capacity up, get channels diversified, get spend out the door. That constraint is gone. Production is cheap. Channels are saturated. Spend is table stakes. What separates the teams that grow from the teams that stall is not how much creative they ship. It is how fast they learn from the creative they already shipped.
Most companies have not absorbed this. They still operate with creative and media as two functions, with a handoff between them, with the creative team looking at the ad manager once a quarter if at all. The motion is not broken yet. It is losing compounding efficiency at every round compared to teams that have closed the loop.
The structural argument George makes is that AI accelerates whatever is underneath it. Teams that already had a closed loop are compounding faster than ever. Teams that did not are drowning in variants nobody has time to review. The agentic layer is not neutral. It amplifies the existing architecture, for better or worse. Building the layer on top of a broken loop does not fix the loop. It just produces more broken output, faster.
Revenue does not fail because teams stop creating. It fails when the feedback loop never closes, when creative and media sit in two different meetings with two different metrics, and when click-through rate is the only number anyone pulls from the campaign. The companies that win the next cycle are the ones that installed a learning loop underneath their creative production before the volume caught up with them.
This is GTM Vault.
If this episode changed how you think about the relationship between creative and performance, forward it to one operator still treating the ad account and the creative brief as two separate meetings.
Connect
Follow George Howes // Canva Grow
Follow Rick Koleta // GTM Vault
Thanks for listening. See you in the next episode.
P.S. Annual paid subscribers get a Private GTM Blueprint Session. One working session to identify your primary GTM constraint and design the 90-day architecture to resolve it.
