Discovery: The Most Important Part of Conversion Rate Optimisation
Estimated reading time: 12 minutes, 39 seconds
There is a common mistake that many people make when they first get started with conversion rate optimisation (CRO). I’ll admit, I’ve made this mistake in the past too. This mistake involves diving straight into testing without gathering the appropriate data first that informs that testing. It is easy to make the wrong assumptions about what to test, these assumptions can mean that you test the wrong thing which then makes it unlikely that the test will be successful. Or even if it is successful, whether it is successful enough to really move the needle for the business.
I really like this case study from Michael Aagaard because it illustrates this point very well. One could assume that telling users that you will never spam their email address would lead to an increase in signups. However this test proves otherwise:
I saw Michael speak about this at the Conversion Conference in London and he guessed that the inclusion of the word “spam” may actually plant negative thoughts into the mind of the visitor. This appeared to be true after he ran the following test which had a positive result:
This case study illustrates the point quite nicely and shows that sometimes, the expected outcome may not actually happen.
Because of this, I wanted to go into detail on what I call the discovery phase of a CRO project where you gather as much information as possible, with the goal of using this information to inform your hypotheses for testing. This phase breaks down into the following areas:
- Understand the company
- Understand the website
- Understand the customer
Let’s take a look at these in more detail.
Step 1 – Understand the company
There is a difference between the company and the website and this difference will become clear to you. It is important to frame them as separate things so that you can distinguish between them and take data from each one appropriately. For the company, you need to be asking questions like the following:
1. What are the company goals?
Company goals should cascade down throughout the company which includes the website. For many companies, the website may not be the only way that they generate revenue. They may have offline stores, telesales or remote sales teams or any number of other ways of generating revenue. The goals of the company will drive each of these and it’s important to get an understanding of these so that you can make sure that the website aligns with them. The website is where you’re likely to have most influence when it comes to CRO, if the website isn’t aligned with the goals of the company, then a few things may happen:
- The website isn’t taken seriously as an asset – meaning you’ll find it painfully hard to get sign-off and budget
- Any improvements you make to the website performance won’t cascade up the company
- Any improvements you make to the website won’t impact the overall goals of the company
I really like the model that Avinash uses here which illustrates the point nicely:
The bit in the middle is the bit that can be directly influenced by a CRO project. When you can clearly see this model in place, you know that the work you’re doing will be valued because it sits within the measurement of success model for the whole company.
2. Why does the company exist?
I’ve talked before on Moz about this in relation to using link building to hit business goals. The principles are exactly the same for using CRO to hit business goals. Here is a snippet from that blog post which still applies today:
If you can do this, it will make your job much easier with the following:
- Understanding their industry and what makes this business different
- Knowing how to get things done internally by understanding what matters to the company
- Defining a strategy that is in line with their business goals
- You can talk the same language and they will trust you because of that – once you have trust, it’s a lot harder to fire you, too
There is another subtle benefit here, too – if you’re pitching to win their business, you’re far more likely to win if you show a genuine understanding of their business.
3. What are the USPs of the company?
The point above leads nicely into the importance of identifying the unique selling points (USPs) of the company. Identifying these early can make a big difference to your CRO projects because it is ultimately these USPs that will drive customers to you over your competitors. Incorporating these into the website is important for persuading them to become customers. If you’re struggling to identify USPs, then you may have an altogether different problem!
Even if you have selling points that aren’t particularly unique, you should make sure that you’re displaying them prominently across your website. Including on mobile devices too, I really like how Schuh have a consistent experience across their desktop site and their mobile version. As you can see, they show things such as next day delivery, student discount and returns policy across the top of the page:
Whilst their competitors can obviously copy these, it is still important to display these for all users across all devices.
4. What data does the company have that may help us?
This is a key difference between gathering data about the website (we’ll talk about this below) and gathering data about the company. There may be some data that can be useful to your CRO project that you need to get your hands on. A few examples are:
- Cancellation rates
- Phone support logs and recordings
- Sales support staff guidelines
- Customer surveys
The list could go on, but you get the idea. This kind of data can be used to inform your hypotheses and help you run much smarter and more effective experiments. For example, if you discover from speaking to sales staff that customers are generally concerned about price and delivery time, you can gather data from the website using tools such as Qualaroo and CrazyEgg to see if people care about them online too. If they do, then perhaps you can emphasize them more to get more people into the sales process.
Step 2 – Understand the website
Next, we need to focus on the website itself and really get to know it and importantly, how customers are currently interacting with it.
1. What are the goals of the website?
Always start here. Remember that they may be subtly different from the goals of the business. For example with companies in the B2B space, the website may not actually sell anything directly, therefore revenue is not a direct goal of the website. Instead, lead generation may be the goal of the website which becomes the thing you focus on. Don’t forget micro-conversions too which, although, secondary, may still play a key role in the sales process. A few examples could be:
- Whitepaper downloads
- Email addresses
- Callback request
Once you have a complete list of website goals, it’s time to look at how (if at all!) they are being tracked.
2. How are goals being tracked?
Fingers crossed, you don’t need to worry too much here! However I’ve worked on too many websites where either goals aren’t being tracked at all, or they are not being tracked correctly. This seems to be the case a lot with goals being tracked via Google Analytics which can sometimes be a bit fiddly.
There are a few ways of measuring the goals of your website, some will be tracked via a backend system such as a custom order system. This is fine and is better than nothing, but ideally you want goals integrated with the analytics platform of your choice too, such as Google Analytics or Omniture. Without going too deeply into this, there are broadly a few ways to do this:
- Goal tracking using URL confirmations – for example if you have a “thank you” page after someone orders a product or contacts you, you can track loads of this page as a goal
- Event tracking – this involves tracking clicks on sections of a page such as a download link or email address link
- Ecommerce tracking – if you generate revenue via your website, this is an absolute must
For the purposes of CRO, it is essential that you are measuring one of these and to make sure they are tied to the goals of the company. If you don’t ensure this, then you will have no way of knowing if your CRO experiments are successful or not.
3. Benchmark and look for opportunities
Now is also the best time to benchmark the performance of the current website so that you have a starting point for improvement. This also allows you to look for opportunities for improvement in your experiments.
For example, I looked at the exit rate report in Google Analytics for a client a few days ago and discovered that one of their checkout process pages was one of the highest pages on the site for exits.
Quick sidenote here: Before jumping to conclusions when viewing this type of report, make sure that Google Analytics code is installed on every page on the website. If it isn’t then clicks to pages without the code will count as exits. This can skew your data quite badly and lead you down rabbit holes!
You should also pay close attention to the audience technology section of your analytics platform. Look for differences in conversion rates between different types of browser or browser versions. This can help you find bugs with the checkout process on certain types of browser.
4. Map out the sales funnel
If you’re lucky, funnel reports will already be setup for you and you’ll just be able to go in and start taking insights from them. From experience though, this is rarely the case! So you will need to setup reports that will track the sales funnel and importantly, show you which parts of that funnel visitors are dropping out at. Once you get this data, you can start looking deeper into why people are dropping out at those stages and running experiments based around those.
If your sales funnel includes offline steps too, you should be sure to incorporate this into your work. Again, I really like how Avinash visualised this in one of his blog posts:
This shows the steps quite nicely and if you can track this effectively, you can easily see where the problems lie and work on improving them.
5. Setup and use appropriate customer segments
Segmentation is absolutely crucial when it comes to CRO – or any form of analytics for that matter! Segments give you more context on what data you’re seeing and you can make more intelligent decisions if you segment appropriately. For example, this is the conversion rate of a website I work with:
But then look what happens when I segment by new visitors and returning visitors:
Now that changes things a bit! It turns out that returning visitors are not converting anywhere near as well as new ones. This means that I could run experiments that are specifically targeting returning visitors to try and improve this. It also gives me the insight to dig deeper into why returning visitors convert so low. It could be because returning visitors come back specifically for something other than buying – such as a piece of sticky content.
Step 3 – Understand the customer
Finally, you need to spend lots of time understanding your customer. It is ultimately your customers than differentiate between CRO test successes. For example, a successful test on a website selling $5 DVDs may totally flop on a website selling $2000 laptops. This is partly because of the difference in mindset between a customer looking to buy a cheap DVD and one looking for a decent priced laptop. Their motivations are different, as are their objections and concerns.
1. What are the fears, concerns and objections of the customer?
This is crucial to the entire process and the outcome of this work should be a series of hypotheses that aim to overcome them so that the visitor is more likely to convert. There are a few ways to do this.
I love Qualaroo, it is simple to install and use and can give you some great insights into what customers really think of you. You can ask them direct questions such as:
- What is stopping you from buying from us today?
- What don’t you like about this page?
- What would you like us to improve about this page?
One of my favourite things to do is to run Qualaroo on pages that the customer reaches after they’ve converted. This means you can ask questions to customers who have already converted and therefore there is no risk of distracting them from buying. You can then ask them questions on what may have nearly stopped them from buying or what they didn’t like about the checkout process.
There are a few tools for this such as usertesting.com, Feedback Army, 4Q and Survey Monkey which you can use to run customer surveys which directly ask for feedback about their concerns and problems with your website and products.
One important point to note here is to make sure that you’re identifying the right demographic for your surveys so that you get good quality data. This is particularly true if you sell very niche products or products that are very technical or expensive. These factors are naturally going to narrow the amount of people who you can send your survey to, but that’s fine. It is better to have less high quality data than more low quality data.
Check live chat transcripts
If the website uses some live chat software such as Olark, ask for access to the transcripts so that you can take a look through and try to find trends in problems and questions.
2. Know your customer demographic
I mentioned this briefly above and it is important in terms of designing experiments too. One example of this is if you are a business that sell internationally. If you sell internationally then you need to understand cultural differences which can sometimes be very subtle. Nathalie Nahai goes into a lot of detail on this in her book which I’d highly recommend.
She also filmed a recent whiteboard Friday where she gives an introduction to this area too.
Something else you can do is enable demographic tracking in Google Analytics which, once you get enough data, can give you some decent (although not complete) demographic data on your visitors. Once you have this data, you can then run some segements over the data and see if it shows any differences in conversion rates which is super cool:
From here, you may find differences in conversion rate between ages, genders or even languages which you could then use to form hypotheses and split test.
That’s about it for this post! Let me know if you have any feedback via the comments.