As you've heard me #testify, in 80% of my Merchandise Forensics projects, there is a problem where customers are spending less, and they are spending less because of merchandising issues.
Increasingly, folks want to know who is spending less. That answer "depends".
In this case, look at what happened in months 12 - 30 (i.e. 1.5 to 3.0 years ago). Look at the bottom of the twelve-month buyer file ... between the 0th and 20th percentiles. These customers were spending 30% less than they were spending four years ago. Notice that their spending declines started first ... they were the harbinger that problems were brewing. The problems didn't get to the top of the customer file until several months later ... the problems were not as dire ... and the recovery was just a bit stronger than at the bottom of the file.
You are probably already running a comparable analysis, and you probably know who is spending more/less over time. It's more dire when best customers spend less ... but sometimes, when marginal customers spend less, it is because a handful of key products that marginal customers like are discontinued.
Your business is not succeeding/failing because you are not on the cutting edge of programmatic ad buys. Your ad buys are not working because of missteps in your merchandising strategy. Use your comp segment framework to understand who is spending less, and why.