If I was an average consumer shopping for broadband only I would have looked at the Plusnet website, not seen it and moved on elsewhere.
Which is something that's been considered as part of the testing. Each version of the site will drive a certain amount of sales. It's entirely possible one will perform significantly worse compared to others. Split testing isn't a newly devised tactic and we're certainly not the first company to do it.
This isn't a reply to Bob - but rather to everyone currently criticising the problem that some versions of the website prove to be "less nice" than others.
This is indeed the whole point of split-testing - it is to send a proportion of your potential customers down different sales/marketing journeys, and to measure the outcome.
The outcomes measured, for commercial companies, tends to be about the lifetime value of a customer - so includes the initial sign-up, and many future purchases. However, it works by averaging the outcome over a large number of visitors (or "prospects")
The downside of this kind of testing is that *some* customers end up going down a "worse" journey, and fail to sign up, where they would have joined if they followed a different path. The statistics will tend to show these things up - and the worst cases will be rapidly discarded from the test.
In this case the test will show the balance between a possible increase in income because people move phone too, against a loss of income because the people who want broadband only walk away. You don't know which will win until you test it.
I've seen split-testing happen over many aspects - headlines, text content, page layout, colours, and even on the raw price of a product (though haven't seen this with Plusnet). Imagine that you *could* have bought a product a lot cheaper, if only the computer had chosen a different version of the website for you!
And no - raising the price doesn't necessarily drop sales. But you only know if you test it.
As Bob says, split-testing is happening everywhere, in all forms of advertising, marketing and promotion.