Automation removed the cost that once told marketers when to stop chasing a lead

The two BCG findings describe one population. If four in five respondents liked personalized experiences and two-thirds had recently run into inaccurate or invasive ones ^[4], then at least about 47% of the sample were in both groups: 80 plus 66.7, minus 100 ^[1].

According to the column, the machinery grew because nothing in the budget limited it. A site visit adds someone to a retargeting audience, a download starts an email sequence, and a click raises a lead score ^[2]. Once those costs are gone, the decision to stop has to be written down as a rule. "That's why I think one of the most underrated marketing skills today is knowing when to stop," the contributor wrote ^[10].

In a board deck, the fix is a frequency cap. Google Ads already lets advertisers cap how often people see display and video ads ^[7]. That answer is incomplete because a cap lives inside one tool. According to the column, most companies run separate email, CRM, advertising, SMS, social, sales-automation and customer-service systems, each with its own rules and its own view of the customer ^[9]. A cap in the ad platform does not count the emails. The customer counts everything and, the contributor wrote, thinks: "Why won't this company leave me alone?" ^[13].

The trade-off is real, and the column names it. Repetition is how someone comes to remember a brand, but past some point another impression achieves nothing ^[14]. A 2024 paper in Decision Support Systems modeled ad fatigue as an inverted U, with returns rising up to a point and then turning flat or negative ^[8]. The column says there is no universal threshold. Five impressions can be too many in one case and nowhere near enough in another ^[15]. A single company-wide number would cut off prospects who needed a sixth touch and still over-contact people who were done after two.

Both of the column's sample rules cross systems. If someone buys a product, stop advertising it to them. If someone has ignored the last 15 emails, reconsider the 16th ^[12]. Each rule needs a fact recorded in one tool to halt a send from another ^[9]. In my view the stop rule belongs to whoever owns the customer record, and that person needs authority to halt sends in channels they do not run.

The sequencing splits across two quarters. This quarter's decision is whether to let AI draft messages and trigger the next step on its own ^[3]. Next quarter brings the volume that decision produces, and the disengagement the BCG and Site Impact respondents described comes after the sends ^[4][6]. The Site Impact figure is what respondents said about their own behavior, and neither study measures whether a stop rule wins lapsed customers back.