Good Enough Isn't podcast

A 90% Failure Rate Is the Whole Point: Akash Pathak on Experimentation and Scaling What Works

0:00
45:30
Rewind 15 seconds
Fast Forward 15 seconds

Send us Fan Mail

Akash Pathak’s fix for stalled AI pilots is one added word: test and learn and scale. He spent more than 20 years leading marketing and digital teams at McDonald’s, Humana, and Trunk Club. Today he advises enterprise and mid-market companies on AI, teaches AI and marketing at Northwestern University, and writes the Growth Amplifier newsletter. He also co-hosted Level’s dinner in Chicago on getting AI past the pilot stage, and this conversation carries that theme further.

Brad Stephenson starts with the programs behind the résumé. Pathak explains why most innovation comes out of a failure rate above 90 percent, and what a leader has to do so a team keeps experimenting anyway. He walks through the McDonald’s mobile loyalty and CRM launch, where performance marketing was new to the organization. He also covers the Trunk Club churn data that showed a 50 to 60 percent gap between men and women and pushed the brand’s focus toward women. A McDonald’s coffee punch card taught him something about retention: the moment a customer redeems the free coffee is what brings them back.

The back half turns to AI. Pathak and Brad compare notes on synthetic audiences, which are useful for poking holes in an early idea before it reaches a real audience. Pathak then lays out the gap he sees between what leaders expect from AI and what teams get, citing BCG and HBR figures on CEO and employee satisfaction, and explains why executives should get hands-on with the tools every week. He covers how he prepared data for a predictive model at Humana, and why stories about results, failures included, are what move a pilot into a process. He closes with why most brands look vanilla and a book recommendation from Jim Kwik.

What You’ll Learn

  • Why most innovation comes out of a 90-plus percent failure rate, and how leaders can keep a team experimenting through it
  • What it took inside McDonald’s to launch a mobile loyalty and CRM program when the company was seven to eight years behind Starbucks
  • Why a brand at 95 percent awareness has a different problem than one in the 20s, and how Trunk Club’s churn data redirected its focus
  • What a coffee punch card revealed about loyalty: redeeming the free cup, not earning it, is the moment that brings customers back
  • How to use synthetic audiences to vet early ideas and find gaps, and why Pathak would never green-light a slogan or logo on their say-so
  • Why executives who get hands-on with AI each week make better decisions, and why associate-driven AI projects succeeded in a 55-project study Pathak cites
  • How to get data ready for a predictive model: pull every customer attribute, find and clean the data, then run correlations against written hypotheses
  • How to move a pilot into a process by telling the story with the data, reporting failures honestly, and recruiting champions so the work outlives its builder

Resources Mentioned

Connect With Akash

Know a marketing leader whose AI pilot worked and then went nowhere? Send them this one.

About the Show

Good Enough Isn't is a podcast from Level Agency, a digital marketing agency, about the hard truths behind business growth and the people who refuse to settle for "good enough."

Listen and Subscribe

Connect With the Show

How to Support the Show

  • Subscribe so you never miss an episode.
  • Share it with someone who would get value from it.
  • Rate and review if it brings you value. It helps more listeners find us.

More episodes from "Good Enough Isn't"