Colgate-Palmolive, the consumer products company, has attempted to use artificial intelligence (AI) to gain a clearer perspective on the payback from its retail promotions – but with limited results.
Rahul Tyagi, Colgate-Palmolive’s worldwide director/analytics center of excellence, addressed this topic at a recent New York conference, where he argued that since consumer packaged goods (CPG) manufacturers typically allocate between 30% and 40% of their budgets to such activity, finding anything near a definitive answer could generate significant value for the business.
“Even the slightest impact can have, basically, a big impact on the bottom line,” he stated. (For more details, read WARC’s report: Colgate-Palmolive finds the limits of artificial intelligence.)
That means building a detailed breakdown of data that can elucidate the impact of distinct promotional tactics at the retail level, including product displays, in-store features and price reductions.
“We need to basically get another level below where we say, ‘Okay we got this net incremental [growth] but to what do we attribute this net incremental lift?’” Tyagi explained.
But achieving that view requires overcoming a number of operational obstacles, ranging from data availability and completeness to its quality and verifiability.
Further, any AI system would need to deliver results that are trustworthy and comparable over time – no small undertaking in the ever-changing CPG marketplace, where competitor activity, seasonality and what other retailers are doing make reproducibility an extremely difficult aim to achieve.
Tyagi’s team came to the conclusion that artificial intelligence, at best, was going to be a partial solution for Colgate-Palmolive to track the impact of retail promotions.
Moreover, the shiny new tool often could not match the tried-and-true research methods already more familiar to the company. “We very quickly realized that [measuring retail-promotion ROI] was not a problem amenable to AI,” he said.
“We actually had to go back to the drawing board and, ultimately, what we have now is a combination which I would say is close to 80% traditional techniques, and some elements are still AI. But 80% of the solution uses traditional techniques.”
Rahul Tyagi, Colgate-Palmolive’s worldwide director/analytics center of excellence, addressed this topic at a recent New York conference, where he argued that since consumer packaged goods (CPG) manufacturers typically allocate between 30% and 40% of their budgets to such activity, finding anything near a definitive answer could generate significant value for the business.
“Even the slightest impact can have, basically, a big impact on the bottom line,” he stated. (For more details, read WARC’s report: Colgate-Palmolive finds the limits of artificial intelligence.)
That means building a detailed breakdown of data that can elucidate the impact of distinct promotional tactics at the retail level, including product displays, in-store features and price reductions.
“We need to basically get another level below where we say, ‘Okay we got this net incremental [growth] but to what do we attribute this net incremental lift?’” Tyagi explained.
But achieving that view requires overcoming a number of operational obstacles, ranging from data availability and completeness to its quality and verifiability.
Further, any AI system would need to deliver results that are trustworthy and comparable over time – no small undertaking in the ever-changing CPG marketplace, where competitor activity, seasonality and what other retailers are doing make reproducibility an extremely difficult aim to achieve.
Tyagi’s team came to the conclusion that artificial intelligence, at best, was going to be a partial solution for Colgate-Palmolive to track the impact of retail promotions.
Moreover, the shiny new tool often could not match the tried-and-true research methods already more familiar to the company. “We very quickly realized that [measuring retail-promotion ROI] was not a problem amenable to AI,” he said.
“We actually had to go back to the drawing board and, ultimately, what we have now is a combination which I would say is close to 80% traditional techniques, and some elements are still AI. But 80% of the solution uses traditional techniques.”
Sourced from WARC
