Let me start off with a story:
I once had a call with a potential customer, and during the call they asked me to tell them what’s wrong with their site, and how I’d improve it. Using typical heuristic analysis, I threw around a few ideas – but I said ultimately we don’t know whether my ideas would work, so we need to conduct proper research. I went on about telling them which data we’d need, and how I’d gather it. “We don’t know what’s working and what’s not working without data”, I told them. Their initial response: “Why do you need research? We talked to this other company, and that other guy knew immediately what was wrong with our site.”
The really sad part was that “the other” guy was a quite well-known digital marketer.
When you’re doing conversion work, you need to have the right mindset: there are no absolute truths. Conversion optimizers should have no dogmas. There is no “this always works”. What’s working for one website, does not necessarily work for the other.
Ego and bias – and we all have them – are not our friends when it comes to boosting website conversions. There is no single right way to do things. You can’t know how a website is doing just by looking at it. We need to “know”, not guess. It’s true that if you’ve worked with a bunch of sites, you develop an “intuition” or “quick thinking” (as per Kahneman), but you should use that only as input for data gathering. For instance if you come across a complicated website, and you think “this is too complicated to use!” – make sure you check that in usability tests. It might also be that it’s not a problem at all for the target users. So get some data behind every “hunch”.
That’s why it’s much more important to master frameworks and methodologies than dogmas (e.g. “never use automatic sliders as they suck”) as every website is different, its audience is different, and the world is in constant change. Frameworks teach you how to think, not what to think.
You want to ask the right questions, and find the data to prove you wrong or right. Sometimes the only way to “prove” something is by testing it, but quite often you can validate it already way before, and not waste precious testing traffic on it.