The Tools Change. The Questions Don't.
I started in performance marketing back when it was called direct response, and the hardest-working person in advertising was the guy slicing a tomato with a Ginsu knife.
We had a phone number for every station, publication and mail drop. Every call was matched back to the media that produced it. We knew, down to the airing, what a sale cost.
Or we thought we did.
Decades later, I'm still sitting in meetings where the agency reports one number, the ad platforms another, GA4 a third and the CRM a fourth. Marketing, sales, and finance each have their own as well. Everyone can defend their piece. Half the meeting goes to deciding whose number is real.
Then someone asks what the business should actually do next.
That was the hard question in the Ginsu era. It still is.
The matchback trap
Direct response's matchback was our greatest strength, and when mismanaged, our greatest weakness.
It credited the media that harvested the response: the airing that ran right before the phone rang. It said nothing about the media that cultivated it, the weeks of exposure that made someone ready to call. So budgets drifted toward whatever looked efficient this week. Cut what you can't see, and response slips. Response slips, so you cut again. That's the direct response death spiral.
It's like driving to your destination while staring in the rear-view mirror. You can see exactly where you've been. You find out what's ahead when you hit it.
We also made customers work harder to respond. Frequency is what makes an offer stick, and we split it across a dozen phone numbers nobody could remember, plus promo codes and special URLs, all for the sake of accurate attribution. The models I built later kept pointing the same way: smaller stations airing often beat bigger, occasional buys.
For a Ginsu or a ThighMaster, harvesting was the right model: prove it on TV or print, push it into retail, squeeze every ounce, move on to the next product. For a brand that needs customers for years, it's a slow way to go out of business.
Better plumbing. Same trap.
Phone numbers gave way to URLs and promo codes, then pixels and cookies. Digital rebuilt the matchback with better plumbing, and the same trap came with it. Branded search and retargeting reach people already close to buying, so they get the credit. The media that made those people interested looks expensive.
Then the signal dimmed. When Apple introduced App Tracking Transparency in 2021, most iPhone users chose not to be tracked. Adjust's data put the global opt-in rate at 34% in the second quarter of 2023, and just 24% in the U.S. Browser limits and privacy laws took more.
So the industry went back to modeling and testing. In a July 2025 survey of 196 U.S. marketers by EMARKETER and TransUnion, 47% said they plan to increase spending on marketing mix modeling in the next year, and 52% already run incrementality tests. Direct response worked those same questions decades ago, with matched-market tests and matchback and attribution models.
Television keeps making the same trip. Cable split the audience and opened up cheap, targetable airtime. Streaming split it again, and now it's bundling back together. Deloitte's fall 2026 Digital Media Trends data shows 65% of streaming households have at least one ad-supported service, and the average subscribing household spends $76 a month. Monthly bills, bundles, and commercials. Back where we started, slightly improved.
Every shift opens new ways to buy media. None of them explains why someone buys.
Sometimes the media isn't the problem
Spending media well matters enormously. It's usually the biggest line item for a performance company, and buying it well means knowing who is watching, when, at what negotiated CPM, and whether a spot is there to cultivate demand or harvest it. Attribution matters too. A rebuilt attribution model once kept a client with me for years.
But get the rest wrong, and even well-bought media looks terrible. I used to think all media is good media at the right price. That isn't entirely true.
An insurer marketing to federal employees and postal workers ran creative aimed at men, because most of the workforce was male. Deeper work showed the wives were making the healthcare decisions. We shifted the creative to speak to them, and results shot up. A media model alone would have blamed the media and cut it.
A national elective medical procedure client had markets full of leads and almost no customers. The difference was presence: without in-person meetings, people didn't convert. Adding a part-time consultation office doubled conversion, with the same national media already running. Markets with a surgical office converted at twice that rate.
And a surgery practice went quiet the moment a patient booked, afraid of talking them "out of the sale". Patients talked themselves out of it instead. A pipeline analysis found the drop. A simple follow-up on what to expect, what to bring, and what to ask, and conversions climbed.
Media, audience, creative, geography, follow-up. Only one of those is a media problem. The others still show up in your media results.
The meeting we keep having
Each of those answers took weeks. Buying-system data, CRM records, forecasts, distance to the nearest office, insurance coverage, and where patients could actually be treated, all stitched together by hand. When the goal changed, another week.
Meanwhile, the business still needed a decision.
What happens if we cut price 10%? Revenue may rise. Does profit? How much more volume would we need, and can operations handle it? Are we winning new customers, or are we discounting sales we'd have made anyway? ROAS and cost per acquisition can't answer that. Does adding a premium at no charge lift conversion enough to cover its cost and increase profit?
It isn't just my meetings. In a 2022 Gartner survey of 377 marketing analytics users, analytics influenced only 53% of marketing decisions. A third said decision makers cherry-pick data to fit a choice they've already made, and 24% said leaders go with their gut.
I've watched the result. The forecast drifts toward the number people want to see. Concerns get raised, then disappear into a presentation with no credible path from the current business to the promised result. The spreadsheet balanced. The business logic didn't.
The person taking that plan into the leadership meeting needs to know what drives the number, what has to happen for it to hold, and what to do if it doesn't.
Will AI fix it?
AI has been genuinely useful in my own work. It will shrink those weeks to hours, and that matters. Weeks spent assembling data are weeks not spent on the decision.
But every tool in this story promised to close the loop. Gartner predicts more than 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value, and inadequate risk controls.
AI makes answers faster. It doesn't make the questions better. Give an AI only the insurer's media data and ask which placements to cut, and you'll get a confident answer in minutes. The wives making the decision are nowhere in that data, so they're nowhere in the answer.
Someone still has to drive.
Why I built Elevate Signal
The statistician George Box said all models are wrong, but some are useful. Every company I've worked with needed a different useful one. Some needed a sharper message. Some needed to see what competitors were targeting and what customers actually preferred, then decide whether to evolve their position or reinvent it. Some had a leak in the funnel. Some had saturated their footprint and needed to shift from acquisition to retention as new customers got more expensive. And some needed to know where the next dollar should go.
So I built Elevate Signal, a strategic growth decisioning platform. It brings customers, brand, competitors, marketing, and finance into one growth plan: whom to pursue, what will make them choose you, where competitors aren't, where to invest, and what has to change first.
It starts with the business decision and works top down. Attribution tools credit individual placements from the bottom up. Signal looks at the strategic shifts that drive growth and profitability, including what must be true for a media investment to pay off. Where the evidence is thin, it ranks the tests worth running first, then feeds the results back into the plan and recalibrates the model. Leadership gets the decision, the evidence behind it, and the financial consequences. It won't make every dashboard match, but it shows where the numbers split and why, so marketing, sales, and finance can plan from the same picture.
It's built around the strategist, not the tool. The person defending the number should be able to see what drives it, and spend their time as a strategist, not a data wrangler. It's designed for growth-stage and midmarket companies making enterprise-sized decisions without an enterprise analytics team. When a company would rather have someone lead that work, that's what I do at Elevate Marketing.
I'm finishing Signal's final validation now.
The question that doesn't change
The tools will keep changing: phone numbers, URLs, pixels, cookies, models, AI. The question doesn't. Where should the next dollar go, and what has to change before it pays off?
If you've sat in that meeting, what decision did it hold up?
Erin Barr founded Elevate Marketing and created Elevate Signal. Erin helps growth-stage and midmarket companies turn marketing investment into profitable growth, with experience spanning healthcare, elective medical, insurance, senior markets, ecommerce, and consumer goods.
Comments