75% of Marketers Can't Say Which Promotions Pay. Martech Futurist | October 8, 2026
Who wrote the objective your marketing AI optimizes against?
When I ask marketing teams that question, most give me the vendor's name and cannot tell me who chose the target or which decisions the team runs against it.
Four writers and analysts have published on this question since Tuesday. JP Castlin asks in Marketing Week how much authority marketers should give a metric. Niamh Carroll reports research from Les Binet and Dom Boyd showing that consumer goods brands spend about twice as much on price promotions as on paid media, and that most marketers cannot say which promotions make money. Rachel Juley at Gartner finds that only 38 percent of CHROs and CIOs share an understanding of how AI changes work. Emily Weiss, also at Gartner, finds that shoppers who use AI for gift ideas consult more sources than shoppers who do not.
On Tuesday I wrote about logging the corrections your reviewers make to AI output. Reviewers correct output against a standard, and today I cover who sets that standard and how many decisions a team should run against it.
In my AI capability framework, I put goals above everything else. A person chooses the target a model scores against, and that person stays accountable for it.
List the decisions you run on each metric
Castlin published Are marketers optimising themselves out of marketing? on October 6. He traces thirty-three years of one-to-one marketing, from Don Peppers and Martha Rogers through programmatic advertising. He also recounts an adtech firm where business development staff with no design training produced ad mock-ups for prospective clients. The client would pick one, remove the creative cost from the budget, and present the difference as a saving. The advertising got worse, and the client reported a better ROI because it had cut the denominator.
Castlin argues that once a team treats optimization as its governing logic, it gives priority to whatever it can measure cleanly. With AI, a team can run targeting, bidding, placement, copy, imagery, and testing through one optimization loop. He asks: "what, exactly, are we optimising for?"
Near the end of the column he lists decisions in order of size. He accepts that an estimate of incremental purchases can sensibly change a bid. He then asks whether a team should use the same estimate to choose the audience, the creative, the media mix, the people who produce the work, and the size of the budget.
Most marketing teams have never written that list. Someone accepted a platform default during onboarding, and the team has run every decision the platform offers against that default since. I have sat in reviews where a team debated creative for an hour and nobody in the room could say who picked the conversion event.
I draw the same line in the framework I am building on Coherence at Velocity. Analysts and models describe what happened and what will likely happen next. A named person decides what the business should pursue and answers for that choice.
Commentary: You can use Castlin's list as a governance exercise. Write down the decisions your team runs on each metric, from smallest to largest, and mark the point where a named person takes over.
Put the promotion budget inside the objective
Carroll reported on October 7 that brands spend twice as much on price promotions as on paid media. Censuswide surveyed 250 marketing and insight leaders at UK consumer goods businesses on behalf of Binet and Boyd, who is chief strategy officer at Kantar. The two planned to present the study at the IPA Effectiveness Conference the same day.
Marketers now spend more on price promotion than on any other single component of marketing. Seventy-five percent of respondents cannot estimate what share of their promotions turn a profit. Eighty percent believe many promotional buyers would have purchased anyway. Eighty-one percent evaluate promotions over periods shorter than 13 weeks, and 68 percent would prefer to run fewer of them.
I read those figures as a problem with the objective. A team that counts volume over 12 weeks will keep discounting, because a discount raises 12-week volume almost every time. A team that gives the same target to an automated offer engine will discount faster.
Budget ownership adds to the problem. In most consumer businesses, sales or merchandising holds the promotion budget and marketing holds media. Each function optimizes its own line, and nobody works from an objective that includes both costs.
It reminds me of what Kim Storin, Chief Marketing Officer at Zoom, said when I interviewed her on The Agile Brand podcast: "If we think about performance marketing or demand generation by itself and in service of just itself, we will be throwing money at a problem but not solving it."
Binet and Boyd recommend a six-step program. Two of the steps are evaluating price, promotions, and advertising consistently through econometrics, and measuring short-term and long-term effects together. If you add promotion depth as a variable in your media mix model, both budget owners work from the same result.
Commentary: Three in four marketers in Carroll's report cannot say which promotions pay, and promotions are the largest line they influence. I would fix that before tuning any media objective.
Agree one indicator with IT, finance, and HR for each AI program
Juley presented at the Gartner HR Symposium/Xpo in London on October 7, and Gartner published the finding that only 38 percent of CHROs and CIOs share an understanding of how AI impacts the future of work.
She says CIOs carry growing accountability for AI outcomes while several executives now own the capabilities behind those outcomes. She also says organizations evaluate AI investments through a technology lens and leave workforce costs out of the business case. She recommends that CHROs formalize shared ownership of key enterprise capabilities and measure success through shared indicators. She also recommends embedding HR staff in AI delivery teams so they can address workforce questions during the build.
Juley wrote for CHROs, and marketing leaders will recognize their own AI programs in her description. Marketing owns the use case, IT owns the platform, finance owns the business case, and HR owns the roles, and each function keeps its own scorecard. A team can hit marketing's volume target and IT's uptime target in a quarter when the people reviewing AI output have more work than they can check, and none of the four functions measures the reviewers' workload.
It reminds me of what Michael Gants, CEO at Encore, said when I interviewed him on The Agile Brand podcast: "There's no one who's tracking this metric every day and obsessed with it. And so it naturally falls through the cracks."
Gants was talking about subscribers who decline a paywall, a group that no team owns. I see the same thing in AI programs. Teams neglect an outcome that has no owner and no metric.
In the framework I am building, intelligence, operations, and experience run across every team. I would give any capability that crosses teams one indicator that all of those teams report on.
Commentary: Juley's 38 percent covers two functions, and I expect a lower figure if you add marketing and finance. Agree the shared indicator before the next budget cycle.
Measure AI shopping as part of the whole purchase journey
Weiss published a Gartner survey on October 7 showing that 34 percent of US consumers are very or extremely willing to use generative AI tools for holiday gift recommendations. Gartner surveyed 1,004 US consumers in August 2026.
Consumers who use AI tools for holiday shopping inspiration consult an average of 4.48 sources, and consumers who do not use AI consult 3.24. Seventy-seven percent of consumers plan to shop in stores and 75 percent plan to shop on Amazon.
Weiss says: "Marketers should not treat AI shopping tools as a standalone channel."
If you treat AI shopping as its own channel, you will score it on its own, and the score will mislead you. A shopper who asks an assistant, checks a retailer site, and buys in a store leaves no last-click credit for the assistant. If you score AI discovery on direct conversions, you will underfund the product data that assistants read.
Under Principle 5 of the Agile Brand principles, you focus on the customer relationship ahead of individual channels, and I would measure AI discovery the same way. I would start with coverage. Pick your top products and check whether a shopper gets the same price, spec, and availability from an assistant, a search result, your site, and your marketplace listing.
Commentary: Take the 4.48 and 3.24 figures into holiday planning. Shoppers who use AI add a source to their research and keep using the others.
Run the numbers on a holiday objective
Here is a hypothetical. A specialty home goods retailer spends $6 million on paid media in the fourth quarter. The marketing team sets one objective in its automated buying system: seven-day attributed revenue at a return on ad spend of 5. The team lets the system set bids, select audiences, and choose between two creatives, a full-price ad and a 20 percent discount code.
The retailer closes the quarter with $30 million in attributed revenue and a ROAS of 5.0, so the team hits its number. Seventy percent of that revenue, $21 million, came through the discount code.
Those orders had a list value of $26.25 million, so the retailer gave up $5.25 million in discounts. Merchandising holds the promotion budget, and the marketing team did not count that cost in ROAS. The team then runs a geographic holdout and finds that 70 percent of code orders happen at full price when shoppers do not see the code. The retailer gave about $3.7 million in discounts to buyers who would have paid full price.
The CMO and the chief merchant rewrite the objective together. They set the new target as contribution after media cost and discount cost, and they read it against the holdout. They also write down who decides what. The team leaves bids and audiences to the system. A named marketer and a named merchant own offer depth, and they cap the code at 35 percent of impressions.
I ran the same quarter under the new rules. The retailer loses half of the orders that depended on the code, about $3.15 million in revenue. It collects full price on half of the orders that never needed the code, which adds $1.84 million. The retailer ends the quarter with $28.7 million in attributed revenue and a ROAS of 4.8.
I assumed product cost at 60 percent of list price. The retailer gives up $790,000 in margin on the lost orders and recovers $1.84 million in discounts. Contribution improves by $1.05 million in the quarter on the same media budget and the same system.
The marketing team would have missed a ROAS target of 5 with that result, so the CMO has to change the target before asking the team to change the offer.
Take these four steps before the holiday code freeze
Name the author of every objective. Open each system that optimizes on your behalf, including ad platforms, send-time tools, offer engines, and agent builders. Record the target, the measurement window, and the person who chose them. If your team accepted a vendor default, record that as well.
Write the decision list for each metric. Sort decisions into three groups: decisions you leave to the system, decisions where a named person decides with the metric as input, and decisions the metric plays no part in. I would put bids and send times in the first group, offer depth and creative mix in the second, and budget size and team structure in the third.
Put promotion cost and media cost in one model. Agree a single measure with whoever holds the promotion budget, extend the evaluation window past 13 weeks, and run a holdout on your largest recurring discount.
Set one shared indicator for each AI program. Agree it with IT, finance, and HR, and include workforce cost in the business case. Report AI-assisted discovery as part of your journey measure and track product data coverage as the input.
I have been on the acquisition side of three company transactions over the course of my career. In each integration, one of the hardest early conversations was about what each side counted as a win. Both teams had hit their numbers for years while measuring different things. We could not combine the teams until one person wrote a single definition and both sides agreed to it.
This week, open the system your team uses most, find the objective setting, and write down who chose it.
Featured this cycle:
Are marketers optimising themselves out of marketing? JP Castlin, Marketing Week, October 6, 2026
Brands spend twice as much on price promotions as paid media, research finds. Niamh Carroll, Marketing Week, October 7, 2026
Gartner Survey Finds Only 38% of CHROs and CIOs Share an Understanding of How AI Impacts the Future of Work. Rachel Juley, Gartner, October 7, 2026
Gartner Marketing Survey Finds 34% of U.S. Consumers Are Willing to Use GenAI for Holiday Gift Recommendations. Emily Weiss, Gartner, October 7, 2026