Joel Pinkham

It’s always exciting to make a big change on your website – the people involved on the project always have an interest in making the site look “good” and “improving” it.

What happens so often though – we get it wrong.

The problem is these are very subjective changes. A lot of the time the good look changes harm functionality and results. Which in my mind, are more important.

I got it wrong

In the last few weeks I’ve been through this scenario. I’m confident I went about it the right way and know what I’m doing, but plain and simply I got it wrong.

I started off looking at the Analytics – seeing how people were using the homepage and understood the metrics that needed improving. Around these I creating a strategy to “improve” the stats and along the way make it “look better”.

In reality, bounce rate has risen from 40% to 55%. There has been some positives out of it, but 15% of your traffic now leaving the site instead of browsing is a huge amount of the audience lost.

My point is – what you think should be done on a website is often wrong.

Solution – Don’t think, TEST

Solution – don’t think, test. I don’t mean focus groups or asking user feedback. This particular homepage tested well in that regard. When the Analytics figures came in though, it failed.

The Amazon Test

Amazon are a perfect example for testing before implementing. When any change is to be made, they sample a small amount of their traffic and measure the impact.

For example before changing the location of the “add to cart” button on their product page, they would try 5 different options over 2% of their visitors and see which one increased conversion rate the most.

Unlike the real word though, they are fortunate enough to have the resources and the traffic to pull it off. If you’re only getting a couple of hundred visitors a day, this kind of test would take a long time. So do what you can within your means, but don’t assume anything.

How nice would it be to take benefits of online advertising to offline?

I’ve just come back from Napier for Easter weekend. As well as enjoying wine and Jack Johnson, I got thinking about the ton of roadside signs attempting to get you to stop for their cafe (Coffee Addicts Ltd) or second rate tourist attraction (Puzzle World). They get a ton of traffic driving past and seeing their signs, what if their message is completely missing people? What if it’s attracting people who want to buy a single drink instead of a family buying a meal? Same principal applies to more than just roadside signs, they’d never know without testing and measurement. The kind of things I guess online marketers take for granted.

Most of these places have no evidence of a brand or reputation so they rely on their signs being seen at the right time and giving people the push they need to pull over. But surely an optimised message could show huge improvements on their revenue.

If only they could setup something to measure the number of cars driving past – something similar to what I’m sure Transit use. Then they could compare this to either the number of cars actually stopping and the sales figures. They could then calculate metrics like Pull over Rate (e.g. 0.1% of traffic visited the shop) and Sales per Vehicle (e.g. the store made $3.25 per 100 cars driving on SH1).

Once they’ve got a benchmark they could begin testing different signs. Maybe in a few years it could be done digitally when a giant LCD is cost effective for a tiny roadside cafe, but for now they could try 3 different signs up on core flute and see how the figures are impacted. I’m certain they would see obvious results, with the traffic numbers being so large especially over holiday periods they’re going to get statistically significant figures in a short amount of time.