Optimization is all about data and the confidence to make sound decisions backed by numbers. But that doesn’t mean there is always a simple, clear cut answer. Even with a sophisticated experimentation program, you still have to juggle priorities. And there are a lot of them to consider, both internal and external. So, how do you navigate the ins and outs of optimizing your digital channels? What makes a really successful optimization program?
Chris Hedick is a long-time optimization professional with experience working both in-house and as a consultant, and is a seasoned SiteSpect user. I spoke to him about ways to maximize your approach to data-driven marketing.
Keys to Successful Optimization
As brands get more and more tech savvy and optimization becomes more ubiquitous, the challenges that optimizers face become more and more nuanced and range from navigating internal dynamics to really technical approaches to data analysis. These are some of the key points from Chris Hedick.
Your data has to be accurate
The more you optimize, the more important data accuracy is going to be for you. If you’re painting in broad strokes, maybe you can afford to sacrifice some detail. But, if you really want to understand customer behavior data accuracy will make or break your business. Hedick and I spoke about how websites now are to some extent templatized — customers have an expectation for what a product listing page is going to look like, for example. And, as optimization becomes a more sophisticated practice, experimentation gets finer and finer.
As Hedick says, “As the web gets more and more competitive it gets harder and harder to get new wins out of a website. So the accuracy of the data is really important. SiteSpect being the old dog for server-side A/B testing, the data is just more reliable. So I think SiteSpect is definitely an industry leader in that area, and I would recommend it to any enterprise client.” Any faults in your data mean faulty decision making. Really audit your marketing tools to understand how they collect and organize data — tag-based tools are prone to misfires, and also lose data from Safari and Firefox anti-tracking features. Don’t just set it and forget it.
Look at the entire customer journey
Hedick says, you need to take “everything from engagement and the first Google Ad or Display Ad, or whatever that first engagement metric is, and optimize and every single point of the journey. Just focusing on the shopping cart funnel, or just focusing on the homepage or a landing page is not going to cut it anymore.” If you’re newer to experimentation and just starting to optimize your site, you likely will see some initial big wins focusing on conversions like purchase, registration, or other bottom of the funnel conversions. But, as you iterate, the smaller steps along the way are crucial for constant improvement.
Part of this process is also dealing with strings of flat or negative results — results that are still incredibly valuable to your business. You have to learn from the “losing” A/B tests too: Hedick says, “even if we produce 10 A/B tests and 8 of them are losers, out of those 8 losers we’ve still learned something. We’ve dug into the data, we’ve come out with something that we now understand about our customer and how they use our website, and now we have to iterate to create future wins.”
Know how the data fits into your business
In optimization and data science driven marketing, you hear a lot about the challenge of dealing with the HiPPO (Highest Paid Person’s Opinion). Hedick says, “this gets into a political problem at a lot of organizations.” It’s also something we talk about sometimes at SiteSpect — being able to empower all of your employees to take smart action with a good data background. The recurring theme in our conversation around the HiPPO was the possible conflict between data-driven decision making and brand-driven decision making. But these two sides of the scale should actually work together. As Hedick says, brands don’t want to “rock the boat by doing something that may be mathematically correct but might actually injure the brand.” Optimization and experimentation is a great way to understand and validate a brand image.
Ultimately, Hedick says, “At the end of the day in ecommerce the decision to buy something or not buy something is an emotional, human decision. We’re not robots. And creating a good marketing message and A/B testing that marketing message, creating a good user experience and A/B testing that user experience, that still matters. And I don’t think that’s going to go away anytime soon.” No set of best design practices or web templates can account for the specificity of how users behave on your website. With the ability to A/B test and personalize with really sophisticated data, this type of data-driven marketing isn’t going to go away.