Shoppers Ranked the Personalization They Find Intrusive. The List Is a Product Sheet.

· CX Pulse

Shoppers ranked the five personalization tactics they find intrusive. Every one is a line on a personalization platform's product sheet.

ThoughtSpot published a consumer survey last Thursday, run by YouGov across 4,833 adults in the US and UK. The headline number is that 91% of shoppers find at least one modern retail personalization method intrusive, and 74% say retailers already collect too much of their personal data. Both numbers got covered. The list underneath them didn't.

Asked which tactics cross the line, shoppers ranked them: tracking activity across other websites or apps (57.6%), sending product recommendations or ads too frequently (56.7%), using real-time location (48.8%), referencing sensitive or personal topics (48.7%), and predicting needs before any interest has been shown (48.7%).

Put that next to the capability section of any personalization platform's product sheet. Cross-site tracking is the data layer. Frequency is the engine running as configured. Location is the mobile module. Predicting a need before the customer has shown interest is the pitch, close to word for word. Shoppers were handed the feature list and asked how they felt about it, and roughly half of them objected to every line.

The program measures lift. Nobody measures the other column.

A personalization program reports on lift: click-through, conversion, revenue per session against a control group. Those numbers are real and usually positive, which is why the programs keep growing. What the survey describes is the cost side, and it never lands in the same report. It shows up as unsubscribe rates in the email platform, consent rejections on the cookie banner, tracking opt-outs in the app, and a support contact that reads "how did you know I was looking at that." Each of those has an owner. None of them is attributed back to the recommendation engine that caused it.

Gartner has a number for how the two columns move together. Content built on ten data dimensions is 3.6 times more likely to be perceived as creepy than content built on one. Ten dimensions is the unified customer profile, the thing the platform is sold on. Every signal added makes the model more confident and, by this measure, the output more unsettling. Only the first of those effects is on a dashboard.

At scale, the fix is three product requirements most stacks haven't shipped

Asked what would earn their trust, shoppers picked transparency about what data is used (34.0%), opt-out controls (31.4%), and an explanation for the recommendation (30.6%). Those are product requirements, and in a large organization they cross four teams: marketing owns the program, IT owns the data model, privacy owns the consent record, and CX owns the queue where the complaints arrive.

The third requirement is the hard one. Explaining a recommendation means the system can say which signal produced it. A model fed ten dimensions generally can't, which is why the "why am I seeing this" line is missing from most stacks even though a third of customers just asked for it. The second is easier and mostly a governance question: a frequency cap is a configuration value, and today it's usually set by the channel team on send volume, not by anyone measuring irritation. The first is a copy change, and it's the cheapest trust the survey offers.

The 74% who think you already collect too much are also the population that sees your consent screen. That screen is a customer experience touchpoint, and the lift model never sees the people who declined on it.

The smaller operation has the opposite problem

A smaller business usually personalizes on one dimension, what the customer bought from you and when, and that dimension isn't on the intrusive list. The pressure is to buy the bundle and catch up. The survey reads as an argument against that. Gartner analyst Brad Jashinsky put the alternative plainly to CX Dive: customers find it less creepy the more directly you ask them for the information. Three in five, per Gartner, want something in return for it, which a smaller operation can offer without a platform.

What to do this week

Put three numbers on the same page as the personalization lift figure: the unsubscribe rate since the program went live, the consent rejection rate, and a count of support contacts that mention being watched or ask for it to stop. If those three have never sat next to the lift number, the program has only ever been measured on the side that argues for it.

Then read your own capability list against the survey's. Shoppers filled in the column the dashboard never had, and they ranked it.

Sources: ThoughtSpot's report and CX Dive's coverage, which carries the Gartner figures.