You start with a landing page. Every respondent who arrives at that URL gets the exact same page.

But to improve, you need to test. You know this because every book or blog post you read about landing pages says the same thing: test, test, test.

To get started, you might do an A/B split test (or a 4-way A/B/C/D split test as shown here) to experiment with a few variations of your page. You can picture this as a Ferris wheel, where each page variation being tested is a bucket on the wheel. As respondents arrive, they are randomly assigned to a bucket. You see which bucket that respondents, on average, like best and declare that the winner.

Now, if you want to experiment with more variations, you might implement a multivariate test (MVT) to try many different combinations of headlines, images, body text, etc. on that page. Effectively, this is like adding a lot more buckets to your Ferris wheel — each bucket being a particular combination of elements tested on a page.

It’s important to remember, however, even as you’re testing dozens or hundreds of possible page variations, that as far as the outside world is concerned (i.e., your respondents) there is still only one landing page. Which particular version of the page they happen to get — which bucket on the Ferris wheel they’re plopped into — is simply the luck of the draw.

It’s also important to remember that the winning page — the bucket that respondents seem to like best — is only the winner on average. It’s not necessarily everyone’s favorite.

For an oversimplified illustration of this, take the 8-bucket Ferris wheel above, with respondents in each bucket A - H. Let’s say bucket B is the winner because 20% of its respondents convert, more than in any other bucket. The next best bucket, E, has only 10% of its respondents convert. However, even though bucket B converted more than E, the people who converted on E may not have converted if they had been dropped in B.

When you’re doing A/B or multivariate testing, you’re looking for the best performing landing page on average — but individual respondents may have their own favorites that differ from the winner for the majority. Have you ever had a situation where you disagreed with the majority? Just because someone likes E, doesn’t mean they’ll like B — regardless of how many other people prefer B.

People are not statistics, and it’s almost impossible to please all of the people all of the time with the exact same thing. One size does not fit all.

This is where landing page optimization can start to feel like you’re going around and around a Ferris wheel — making you a little nauseous. As you test more and more tweaks and variations to the same page, improvements to your conversion rate start to slow down. As you make changes, you might win new respondents who wouldn’t have converted before, yet lose other respondents who preferred the previous version. And after running through a large number of tweaks, it’s almost impossible to figure out what you really learned from all those tests.

This is where landing page optimization hits a wall.

The secret to breaking free of this endless loop is to stop focusing on variations in content, and instead aim to better understand the variations in your audience. Out of all the respondents coming to your landing page, it’s highly unlikely that they’re homogenous. They’re often an aggregation of several distinct segments or niches.

To make big leaps forward in improving your conversion rate — the 2X to 5X improvements that make you an online marketing superstar — you need to identify these segment respondents and send them to separate landing pages that speak specifically and authentically to their different needs and interests. Instead of one Ferris wheel with a hundred buckets, you’re far better off with 3 or 4 separate Ferris wheels — one for each of your key segments. You can then run a smaller number of tests on each of these segment-specific wheels to optimize their experience (usually A/B testing is sufficient and easier to manage).

You no longer have one conversion rate trying to be optimized — suboptimally — for everyone. Instead, you have separate conversion rates genuinely optimized for each segment. Instead of one size fits all, you have different sizes for different types of respondents — and as a result, respondents get a better fit for who they are.

This is a Long Tail strategy for landing pages:

• more separate/distinct landing page destinations;
• each landing page targeted to a very specific audience segment;
• a smaller (more manageable) number of tests on each landing page;

This approach to landing page optimization is more about finding the optimal number of distinct landing pages for specific groups of respondents rather than trying to optimize a single page ad infinitum (ad nauseum?) to the lowest common denominator of all traffic in general.