Personalization got sold to businesses as an artificial intelligence problem. Buy the engine, feed it behavior, and the site rearranges itself for every visitor.
For most corporate and professional service websites, that is the expensive answer to a question you can solve a simpler way. The useful version of personalization is mostly segmentation, and it needs no machine learning at all.
Most of what gets called personalization is just segmentation
A first-time visitor reading about who you are has a different job to do than a returning one comparing case studies. You do not need to predict that. You already know it from where they came from and what they clicked.
Three signals cover most of it:
- What brought them. The search term or the ad tells you the problem they think they have.
- Which page they landed on. Someone arriving on a service page is further along than someone arriving on the home page.
- Whether they have been before. A returning visitor rarely needs the introduction again.
That is enough to make a site feel like it was written for the person reading it, and none of it requires a model.
The version that works without any machinery
- A page per thing you actually sell. Not one services page listing nine offerings. Someone searching for one of them should land on a page about that one, not on a menu.
- A call to action that matches where they are. A first-time reader is not ready for “Request a quote”. “See what it costs” respects the stage they are at.
- Industry pages, if your work genuinely differs by industry. If it does not, do not fake it. A thin page per vertical reads as thin.
- Follow-up email triggered by what someone did. Downloading a guide and visiting a pricing page are different signals and deserve different follow-ups.
All of that is ordinary content and configuration work. It is also the part that actually moves enquiries, because it changes what the person reads rather than how cleverly it was chosen.
Two things to leave alone
Pricing that changes based on behavior. Showing a different price because of how someone browsed is not personalization. Customers find out, because people compare, and the reputational cost lands all at once.
Tracking that you would not be comfortable describing out loud. If explaining the mechanism to the visitor would make them uneasy, that is the answer. Under GDPR and CCPA it is also a compliance question, not only a taste one.
Personalization and privacy are the same decision
Every piece of tailoring is built on something you know about the visitor, so the two cannot be separated.
- Opt in and opt out should both be easy. A consent banner where declining is harder than accepting is a dark pattern, and regulators now treat it as one.
- Prefer first-party data. What people tell you and do on your own site, rather than what you buy about them.
- Say what you collect in plain words. People engage more when they understand the exchange.
Transparency is not the cost of personalization. It is the thing that makes people willing to give you anything to personalize with.
Where the effort actually pays
If you do one thing, make the page someone lands on match the reason they clicked. An ad about emergency repair should not land on a general home page. A search for one service should not arrive at a list of nine.
That single alignment does more for enquiries than any recommendation engine, and you can do it this week with the pages you already have.
At Velant we build the site and then keep it running, which is what makes this kind of change practical: adjusting a landing page or a call to action is a request, not a project.