Possibility and Probability

Coding Python and making businesses…

12 January 2016

Ways to test your experiment

by admin

In the last few posts we’ve talked about why experiments are important (you are doing some right?), and how to use some very simple tools to get very insightful information. But one thing that we have not discussed is how to get people to actually use your experiments. In the first post the idea of statistical significance was presented. This basically states that you can’t decide between two things until you have had enough people look at them. For example, if you are testing out 2 different “call to action” buttons, you are going to need more than 10 people to look at those buttons (and click on one or the other) before you can make a real decision about which one is better. So using Evan Miller’s A/B site we can see that most of the time we will need several hundred users (or visits, I’m using the terms interchangeably here) to look at your experiment. The next question you need to ask yourself if is “How long will I have to wait until I get that many views?”. You do know how many people visit your site, right? If not, stop and install something like google analytics or addthis right now!

Look at my experiment!

Let’s say you decide you will need 1,000 views to determine the outcome of your experiment. But, your website only gets 100 views a month. That means it will take 10 months to determine the winner of the experiment! That is way too long to wait, in that time all kinds of external factors could come into play and affect the outcome. Not to mention if your business depends on this answer, you could be out of money well before then! Obvisiouly you need more people to visit your site. This is the eternal struggle of all businesses both on the internet and in real life. Here are some ideas on how to drum up some visitors to your site. Please note that these ideas will get people to your site, but that does not necessarily mean that they will purchase things from your site.

Buying ads

One way to quickly get quality views of your site is to purchase ads that will take users to your content and experiment directly. The advantage of this is that as long as you set up your ads correctly, you should be getting very targeted people. This means that they will be more likely to engage (e.g. do something) with your website. The downside is that this can get expensive over time. In short timeframes though, this is a really good approach to quickly testing something out. As far as cost, it is a case of you get what you pay for. Costs are typically per-click on the ad, so if you get a ton of clicks it will cost more. You can also set how much you want to pay per click. The details for this are specific to each ad vendor, so be sure to check with them for the fine print. Here’s some of the heavy hitters in this area:

Buying traffic

This is a cheaper alternative to buying ads, but the quality of the traffic is much lower. Depending on where you buy the traffic from you might be able to request a certain type of person visit your site (think: “someone interested in games”), but from what I have seen the overall traffic is typically not very good: The visitors don’t seem to engage or stick around for very long. There’s several potential reasons for this which are beyond the scope of this post. The market for supplying this kind of traffic is… “fluid” so I’m not going to recommend any links. A quick web search will probably be able to get you in touch with the current industry leaders.

Experiment with Social Media

Using social media (like Facebook, twitter, etc.) is great for generating traffic to your site and your experiment. Its free, if done correctly its authentic (in that it reflects your brand), and it can be long lasting. The downside is that it can take a while to get going. Not every posting goes viral, so this is a long game approach. Setting a schedule and posting good content regularly is key. Sharing the content with your friends and connections is also important and can help spread the word. Just be careful not to spam your friends too much, that can be very bad for your reputation and stats. There are services out there that can help boost things on various social platforms. While these can work, they do require great content from you to get going. Bad posts, ugly pictures, uninteresting stories simply will not work no matter how much money you throw at them. If you do decide to use they types of services, I highly recommend you spend some time working on your content first. Content is king!

Wrapping up

In this post we talked a little more about statistics and why its important to get people to visit your site and interact with your experiments. There’s lots of choices when it comes to getting people to look at your content. I personally recommend taking a long game approach and building up great content, and then augmenting it with ads periodically to get focused attention to specific experiments. Good luck and let me know in the comments what works best for you!

tags: