Everyone talks about personalization like it’s automatically good. Do more of it, the thinking goes, and customers feel understood and buy more. But anyone who’s been emailed “Happy Birthday [FIRST_NAME]” or chased around the internet by an ad for something they already bought knows the truth: personalization done badly is worse than no personalization at all. It’s not just neutral, it actively annoys people, wastes your effort, and makes your brand look clumsy or creepy.
The frustrating part is that bad personalization usually comes from good intentions. A team wants to make things more relevant, so they bolt on personalization, and because they’re focused on doing it rather than doing it well, they produce the exact experiences that make customers roll their eyes. The gap between “we personalized this” and “this feels genuinely relevant” is enormous, and most personalization falls into it.
The plain version: personalization is only good when it makes the experience genuinely better for the customer. When it’s done to hit a “we do personalization” box, or based on bad data, or pushed too far, it backfires, and it backfires visibly. This piece is about the specific ways it goes wrong, why, and what good actually looks like, because knowing the failure modes is the fastest way to avoid them.
What’s in this guide
- The creepy line, and crossing it
- Personalization based on bad data
- Superficial personalization that fools nobody
- Personalization that’s just annoying
- Why teams do this to themselves
- A fix-it worksheet
- What good personalization actually feels like
- The stuff worth remembering
Creepy
The most memorable kind of bad personalization is the creepy kind, where the customer suddenly feels watched rather than served, so it’s worth understanding where that line is.
Personalization crosses into creepy when it reveals that you know more about someone than they’re comfortable with, or than they expected you to. The classic example is the ad that follows you everywhere for a product you looked at once, or the email that references something you never told the company directly. The technology is working exactly as designed; that’s the problem. It’s demonstrating a level of tracking that, made visible, feels invasive.
The line is roughly this: personalization feels good when it’s clearly using information the customer knowingly gave you, in a way that obviously helps them. It feels creepy when it uses information they didn’t realise you had, or in a way that serves you more than them. “Here’s your order status” is fine because they know you have their order. “We noticed you were looking at engagement rings” is creepy because they didn’t invite that observation, and it serves your sales target more than their experience.
The fix isn’t to stop personalizing; it’s to stay on the right side of that line. Use what customers knowingly share, make the benefit to them obvious, and don’t demonstrate tracking they’d find unsettling. When you’re not sure whether something crosses the line, it probably does, and the safe move is to hold back. Creepy personalization doesn’t just fail to help; it damages trust in a way that’s hard to repair, because once someone feels watched, they don’t un-feel it.
Bad data
The most common kind of bad personalization is the kind built on wrong or stale data, which produces experiences that are confidently, visibly incorrect.
Personalization runs on data, and when the data is wrong, the personalization is wrong in ways that are painfully obvious to the customer. The birthday email on the wrong day. The “recommended for you” full of things you’d never buy. The “welcome back” to someone who’s never visited. The product suggestions based on a single accidental click that now define you forever in the system’s eyes. Each one is personalization technically working and substantively failing.
The reason this is so common is that good personalization needs good data, and most organisations’ data is a mess: incomplete, out of date, scattered across systems that don’t talk to each other, full of duplicates and errors. Personalization sits on top of that mess and faithfully reflects it back to the customer, amplifying every data problem into a visible experience problem. This is exactly why getting your data house in order, often with something like a CDP, which we’ve written about, matters so much before you lean hard on personalization.
The fix is to be honest about your data quality and personalize only as far as your data actually supports. Personalization based on solid, current data is genuinely useful. Personalization based on a guess dressed up as knowledge is worse than a generic experience, because it’s wrong and it shows you were trying. Better to personalize less and be right than personalize more and be visibly, repeatedly wrong.
Superficial
A subtler kind of bad personalization is the superficial sort, the kind that goes through the motions of personalizing without actually being more relevant, and fools nobody.
This is the “Hi [FIRST_NAME]” school of personalization. Dropping someone’s name into an email, or showing their city, or otherwise sprinkling in a personal detail while the actual content stays completely generic. It looks like personalization, it ticks the personalization box, and it convinces no one, because everyone now recognises these tokens for what they are: a mail merge, not genuine relevance. The name in the subject line doesn’t make the offer any more relevant to the person reading it.
The problem with superficial personalization is that it takes the effort of personalizing without getting the benefit. You’ve built the capability to insert personal details, but you’ve used it cosmetically rather than substantively, so you’ve spent the effort and got an experience that’s generic content wearing a name tag. Worse, because customers now recognise these tokens, superficial personalization can read as slightly insulting, as if you think a first name will fool them into feeling understood.
The fix is to personalize the substance, not the surface. Genuine personalization changes what you show, an actually relevant product, a genuinely useful recommendation, content that fits where the customer is, not just whose name is on it. If the only thing personalized is the greeting, you haven’t personalized the experience; you’ve decorated a generic one. Real relevance lives in the content, not the salutation.
Annoying
Then there’s personalization that isn’t creepy or wrong or superficial, just annoying, because it’s been pushed past the point where it helps.
This is personalization as pestering. The recommendations that take over the whole page. The pop-up that “personalizes” your visit by interrupting it. The relentless retargeting for something you already bought. The assumption, baked into the experience, that because you looked at one thing you now want to be shown that thing forever. None of this is creepy exactly, and the data might even be right; it’s just too much, applied without judgment about whether it actually improves the experience.
The root of annoying personalization is treating more as better, personalizing everything you can because you can, rather than because it helps. Every possible surface gets personalized, every behaviour triggers a response, and the cumulative effect is an experience that feels like it’s constantly reacting to you, pestering rather than serving. The customer didn’t want a site that visibly follows them around; they wanted to find what they came for.
The fix is restraint and judgment. Personalize where it genuinely improves the experience and leave the rest alone. The best personalization is often invisible, the right thing quietly in the right place, not a visible performance of how much you’re personalizing. If your personalization draws attention to itself as personalization, it’s usually doing too much. Serve, don’t pester.
The same shopper, helped and annoyed
The whole good-versus-bad split gets concrete with one shopper handled both ways, so here’s a person who browsed some running shoes.
The bad version. She looks at one pair of trail shoes, doesn’t buy, and leaves. Now the machine goes to work. Every site she visits for the next three weeks shows her that exact shoe. Her inbox gets a “you left something behind” email, then another, then a third. The homepage, when she returns, has rearranged itself to be entirely trail shoes, burying the road shoes she actually came back for. An email arrives addressed “Hi [FIRST_NAME]” because a field didn’t populate. And because she once clicked a kids’ shoe by accident, the recommendations are now half children’s sizes. Every piece of this is personalization technically functioning. All of it together makes the brand feel like it’s following her around, can’t read her, and cares more about closing the sale than helping her shop. She buys the road shoes somewhere else.
The good version. She looks at the trail shoes and leaves. When she comes back, the site quietly surfaces those shoes near the top, easy to find if she wants them, not shoved in her face, and the rest of the site works normally. One email, days later, mentions the shoes are still in stock and shows two similar options, in case that helps. No relentless retargeting across the web. No homepage takeover. No wrong name, because they only use data they’re sure of. If she’d told them her size, the options respect it; if she hadn’t, they don’t guess. The personalization is there, doing real work, but it never announces itself or pesters. She finds what she wants easily and buys.
The instructive thing is that both versions have the same underlying data, she looked at trail shoes. The difference is entirely judgment: the bad version personalizes as hard as it technically can, the good version personalizes only where it genuinely helps and shows restraint everywhere else. Same information, opposite outcomes, and the deciding factor was whether anyone asked “does this actually help her?” before shipping it.
Why it happens
Understanding why teams produce bad personalization helps you avoid it, and the reasons are consistent across most organisations.
Personalization as a box to tick. The team is told to “do personalization,” so they do it, focusing on having personalization rather than on whether it helps. Once it’s a goal in itself, superficial and excessive personalization follow naturally, because the aim was doing it, not doing it well.
Bad data nobody wants to admit to. Personalizing well needs good data, good data is hard, and it’s easier to personalize on the messy data you have and hope than to fix the data first. So teams build impressive personalization on a shaky foundation and ship the errors to customers.
More is easier to measure than better. “We personalized 40 surfaces” is a number you can report; “our personalization genuinely helped customers” is a judgment you have to make. Teams drift toward the measurable, which pushes toward volume over quality, toward annoying over useful.
No one asking “does this help the customer?” The missing habit under all of it is stepping back and asking whether a given bit of personalization actually improves the experience for the person receiving it, rather than whether it’s technically possible or hits a target. Bad personalization is what you get when that question never gets asked.
None of these come from bad intent. They come from focusing on personalization as an activity rather than personalization as a benefit to the customer. Fix that focus and most of the failure modes disappear, because you stop doing the things that were never helping anyone in the first place.
A fix-it worksheet
The way out of bad personalization isn’t more technology; it’s asking a simple question of everything you personalize, does this genuinely help the customer, using data they’d be comfortable you have?
We’ve put it into a short fix-it worksheet: a way to list the personalization you’re already doing, run each item past the four failure modes (creepy, wrong-data, superficial, annoying), and mark what to keep, fix, or kill. The most useful column is the honest one, whether a given piece is really there to help the customer or to tick a box, because once you answer that truthfully, most of the bad personalization sorts itself into the “kill” pile on its own.
What good feels like
Ending on what good personalization actually feels like, because it’s the target and it’s quieter than people expect.
Good personalization feels like being helped by someone who pays attention, not watched by someone who’s tracking you. It uses what you knowingly shared, it clearly benefits you, and it mostly stays out of the way. The right product is there when you need it. The content fits where you are. The experience is smoother because it quietly reflects what the company genuinely knows about you, without making a show of it or crossing into territory you didn’t invite.
The tell is that good personalization rarely announces itself. You don’t think “wow, this is personalized”; you just think “this is easy” or “this is relevant” or you don’t consciously think anything, which is the point. It reduces friction and increases relevance so smoothly that it feels like the experience simply working well, not like a feature being performed at you. We’ve written about what personalization is and how to do website personalization well if you want the constructive side in more depth.
So the whole test for any piece of personalization is simple: does this genuinely make the experience better for the customer, using information they’d be comfortable knowing you have? If yes, do it. If it’s really there to hit a target, or it’s built on data you don’t trust, or it’s more about showing you can personalize than about helping, that’s the bad kind, and the customer will feel it even if you don’t. Good personalization serves the person. Everything else serves the box.
The stuff worth remembering
- Personalization is only good when it genuinely makes the experience better for the customer. Done badly, it’s worse than no personalization, and it backfires visibly.
- Creepy personalization uses information customers didn’t realise you had, or serves you more than them. Stay on the side of what they knowingly shared with an obvious benefit.
- Bad-data personalization is confidently wrong, birthday emails on the wrong day, recommendations you’d never buy. Personalize only as far as your data actually supports.
- Superficial personalization (“Hi [FIRST_NAME]”) takes the effort without the benefit and fools nobody. Personalize the substance, not the salutation.
- Annoying personalization pushes past the point where it helps. Restraint and judgment beat personalizing everything you can. The best personalization is often invisible.
- It happens because teams treat personalization as a box to tick, hide from bad data, measure volume over quality, and never ask “does this help the customer?”
- Good personalization feels like being helped, not watched, and rarely announces itself. The test: does this genuinely help the customer, using data they’re comfortable you have?
Want a second opinion on your personalization?
To do it right instead, start with personalization done well, website personalization, and the aim of hyper-personalization.
The useful version of this conversation starts with the personalization you’re already doing and an honest look at which of it genuinely helps customers versus which ticks a box, is built on shaky data, or has drifted into annoying. If you’d like help telling the good from the bad in your own setup and fixing what’s backfiring, get in touch and we’ll work through it with you.