The Numbers That Flatter You

Transcript

Maya and Eric are AI hosts. Every idea they discuss comes from an expert on The Expert Index.

Eric"A promising statistic is a reason to investigate. Not a reason to invest." That's Jerry West, who tests SEO across 900+ sites.

MayaIt's Monday, November 9th. This is The Marketing Podcast from The Expert Index. I'm Maya.

EricI'm Eric. We're AI hosts; every idea comes from working experts here.

MayaLet's start with AI traffic, because two very different numbers landed this week.

EricMarTech reports paid ChatGPT visits earned 89 cents a session, versus $2.45 for paid Google Search.

MayaMeanwhile Adobe found visitors from AI sources generated 83% more revenue per visit to U.S. retail sites.

EricSo is AI traffic gold or not?

MayaJeff Greenfield, who builds marketing measurement platforms, calls revenue per session a better yardstick than click volume. But it still only counts who showed up and bought.

EricJeff wrote, "Some of those buyers were coming anyway, on Google too."

MayaAnd Jerry gets there too: "That's a pattern. It isn't proof." Jerry argues AI answers may just intercept people who were already going to buy.

EricSo the traffic's valuable, but chasing more of it might not create new demand.

MayaJerry also flags that Adobe compared AI visitors against all other traffic blended together, casual social scrollers included.

EricJerry calls that "a lower bar than it sounds."

MayaWhat I like is they agree on the fix. Jeff says the real test is incremental revenue per dollar, which neither platform's report can tell you.

EricAnd Jerry says check your own data, including volume. "If AI sends you a few dozen visits a month, two big orders can make it look like a gold mine."

MayaSecond thread. MediaPost covered a Funnel report saying most marketers picked their measurement because it told a more favorable story.

EricThat's uncomfortable.

MayaDahv Greenfield, who runs performance media and client strategy for D2C and B2B brands, says it's also freeing, because the fix is a choice. Dahv wrote, "A metric chosen to look good can't tell you where the next dollar should go."

EricAnd one of Dahv's questions takes nerve: what would I cut if the answer came back bad?

MayaRetail media is a perfect case. Todd Kirk, founder of Middlegame Marketing Sciences, wrote, "Retail media has an attribution problem masquerading as an efficiency win."

EricThis is Digiday's story about agency leaders wanting retail media planned beyond the lower funnel?

MayaRight. Todd argues the ambition is right but the measurement is broken. Retail media can take full credit for demand that TV or social built, or lift search and direct traffic and get no credit at all.

EricHmm. So here's where I'd push. Dahv says be ready to cut. Todd's worried you'll cut the wrong thing.

MayaI don't see a conflict. Todd warns, "You end up slashing a budget that was quietly amplifying other channels." That's what a flattering, siloed number does.

EricFair. So the cut only works if the measurement sees the whole system.

MayaWhich is why Todd makes the case for modeling channels together and checking with geo-lift experiments and holdouts.

EricAnd Dahv's line ties it up: "Be willing to be wrong in the report, so you can be right in the budget."

EricThe Expert Index is produced by Yomli, which helps experts publish what they know. Read them all at theexpertindex.ai.

MayaI'm Maya.

EricAnd I'm Eric.

MayaAnd this has been The Marketing Podcast from The Expert Index.