Website personalisation: logged-in experiences that convert

Website personalisation is any change to what a page shows based on something the site knows about the visitor. At the shallow end that knowledge is a campaign parameter or a rough location, and the change is a headline or an example. In the middle it is the visitor's behaviour this session, and the change is which guide or offer is suggested next. At the deep end it is the visitor's identity and account, available only once they log in, and the change is a page that knows their name, their plan, what they bought, where they are up to and what their own numbers say. The deep end is where the value is, and it is the part of the site our client portal guide calls the gated layer.

The industry conversation is mostly about the shallow end, because that is what the tooling sells: rewrite the homepage per segment, swap the hero per source, reorder the products per click. Some of that works. All of it carries a cost most vendors do not mention, which is that a page that changes per visitor is a page Google cannot trust to be the page it crawled, and an AI engine cannot quote with confidence. The rest of this guide is about getting the lift without paying that cost.

Three conditions keep open-page personalisation inside what Google accepts. The default page, the one served with no campaign parameter, no behaviour history and no login, is a complete page that stands on its own, and it is the page Googlebot receives. Personalisation on that page is additive: a headline swap, an inserted module, a reordered list, never the removal of content that the crawled version carried, because a page that shows Google more than it shows people is the definition of cloaking. And the personalised versions live at the same URL, with campaign parameters handled by a canonical tag pointing at the default, so the crawler does not find a hundred near-duplicates of one page.

AI engines make the rule more important, not less. ChatGPT, Google AI Overviews and AI Mode, and Perplexity quote the default page, so the default page has to carry the facts, the definitions and the proof you want attributed to you. A site that hides its best content behind a behavioural rule (show the case study only after three pages) has hidden it from the engine that would have cited it. Our guide to getting cited by AI in Australia covers what the default page has to look like; the point here is that it must exist, complete, for everyone.

Behind the login none of the constraints apply and all of the value appears. The page can greet the customer by name, show them the one thing to do next (which is the single most effective piece of personalisation there is, because the alternative is a dashboard of everything), surface the onboarding video for the stage they are at, show their own report, and offer the upgrade that their usage says they need. Gartner's finding that only 14 per cent of self-service issues fully resolve is largely a finding about generic help: a knowledge base article that does not know which plan the customer is on cannot give them their steps. A logged-in page can. That is the difference between a portal that customers return to and one they phone around.

This is why the two-layer site is the right architecture for personalisation rather than an obstacle to it. The open layer is the same for everyone and complete for the engines; the gated layer is different for every customer and invisible to them. Our membership website guide covers what belongs in the gated layer, and our onboarding video guide covers the first thing a personalised portal should show a new customer.

Levels three and four run on personal information, and in Australia that means the Privacy Act 1988 and the Australian Privacy Principles, which apply to most businesses with annual turnover above $3 million and to some smaller ones. The practical obligations for a personalised portal are the ones the Office of the Australian Information Commissioner publishes: collect only what you need for a stated purpose, tell people what you collect and why in a privacy policy they can find, keep it secure, and let them access and correct it. A portal that shows a customer their own data is, done properly, the transparency the principles ask for; a portal that quietly builds a profile from behaviour the customer did not know was tracked is the opposite. This is general information, not legal advice.

The simplest test is the one from our gated content guide, applied to data: personalise from what the customer has knowingly given you (their account, their purchases, their stated preferences) before personalising from what you inferred. The first is expected, the 71 per cent in McKinsey's research are asking for it, and it is inside the rules. The second is where the 76 per cent frustration turns into something worse.

Measure the open levels by conversion rate on the personalised version against the default, with the default as the control and the change held to one variable at a time; our A/B testing guide covers the mechanics and the sample sizes. Measure the logged-in levels by return rate, resolution and the share of customers who complete the next step they were shown, which are the numbers that say the portal is doing the job. And keep watching rankings and citations for the open pages while you personalise them, because a drop there is the indexing rule being broken somewhere. All three sets of numbers come from the data layer, which on our side is the Snowball SEO platform doing the monitoring other agencies bill by the hour, so the budget goes to the design of the personal layer rather than the reporting on it. Building and measuring it is part of our conversion optimisation service.