An interview on best practice Ecommerce analytics techniques with Dan Barker
Much of what's written about successful Ecommerce sites focuses on the big brands we all know like Amazon, Argos and Asos. But within each sector of retail, there are many Ecommerce outfits outside of the spotlight - some very successful, some looking to improve practice. With this in mind, I have been looking for a while to hook-up with an Ecommerce specialist who understands the ingredients for success.
In this interview, I'm pleased to introduce Dan Barker who manages Ecommerce for two companies (B2B and B2C) operating in the Children's products sector.
Some of you may know Dan for his willingness to offer detailed advice via Twitter and digital marketing blogs - he's also a regular contributor on Smart Insights and also writes his own blog. We're republishing this post from March 2010 as a refresher for Jan 2012 as part of a series of posts related to "New Years resolutions".
Reviewing your visitor mix
Q1. How would your review the effectiveness of the mix of traffic to an Ecommerce site?
In a big nutshell: I'd segment this by channel and, within each channel, segment further by new vs existing customers, then further still if necessary by types of customers. The reason being, when you can do that, you can get a much better picture of ROI.
At that point, you can then find the balance between growing the channel (spending more money to acquire more customers) and keeping it efficient (making sure you make at least £1 back for every £1 you spend).
I think that puts you in a nice position too: If you decide you need to grow faster, you can use the excess money from very profitable channels to subsidise others that aren't profitable. Or if you decide you need more margin, you can then change your ROI targets to say "I want £2 back for every £1 I spend".
When trialling new tactics, it's useful to categorise them by things like:
- Technical setup needed
- Organisational setup needed
- Likely potential scale
- Return (vs cost)
Trial the easiest, least risky, highest ROI potential tactics first. Trial them as small as you can, and then "if they work " scale them up as much as you can & measure ROI the whole way.
Measures and dashboards
Q2. For managing those channels ongoing "is there anything that would apply to any type of ecommerce site to improve results?
Yes, one thing I like to do that's absolutely key is to set up a big grid for regular reporting. For example, across the top I may have key segments and channels including things like:
- All Visitors
- New (unknown visitors/those who haven't purchased before)
- PPC Brand
- PPC Non-Brand
- Affiliate (New)
- Affiliate (Retained)
- Natural (New)
And then down the side a long list of metrics, including some key ones like:
- Current Monthly Orders
- Last Year Orders
- Projected Orders vs Same Period Last Year
- Overall Conversion Rate
- % of Overall Orders
So if I want to see if Non-Brand PPC is doing ok in terms of ROI & year-on-year sales, I just look at its column, scan down to those cells, and the info is there.
A nice extension of that is to include graphs over time so you can look for peaks & troughs by channel, asking the question "why has that dropped?" or "why has that gone up?". Including multi-point attribution is nice if you can (eg. Looking at sales where the "first click" came from a particular channel, vs sales where the channel contributed "last click").
Reporting sometimes gets a bad name, but I think if you can get it to a point where it's actionable it's so useful. At a glance, you can see which channels are over/under-performing, where you can allocate extra budget, and where you need to dig deeper.
Conversion optimization for Ecommerce sites
Q3. How do you approach 'conversion optimisation', a hot topic at the moment?
"Conversion Optimisation" is interesting. A lot of the information out there around conversion rate optimisation has come from membership sites, information product sites and the like "who are all about driving traffic to a single page for the first time and "converting" those visits immediately into sales.
That gets muddy very quickly with E-commerce sites for 4 reasons:
- You have thousands of products rather than just 1 or 2.
- You have new visitors mixed in with old, loyal customers.
- You have any number of channels bringing traffic to the site.
- The journey through the site is often 10 or more pages, rather than a single landing page pushing straight to purchase.
That means immediately you have to break things down into chunks. Rather than saying "what's the overall conversion rate of the site?" and then scratching your head wondering which lever to pull to make it move, you break it down into "micro-conversion goals". For example:
- The purpose of my homepage is to get visitors to the category page.
- The purpose of the category page is to get visitors to a product page.
- The purpose of a product page is to get visitors to add to basket.
- And so on...
So you might say "I've got 100,000 new visitors a month landing on the site on my search results pages. The purpose of the search results page is essentially to show the visitor a product they're interested in and persuade them to click it for more information". From that point you ask "what can I change about my search results pages to make that more likely to happen?". And then you trial that change, measure to see the impact, and repeat if it works.
Though working to increase conversion is hugely important, there are a few caveats that don't get mentioned much:
- Not all pages are made equal. Perhaps 20% of buyers will see a search results page, whereas 100% may see a product page, 100% a basket page, etc.
- Conversion rate is a really misleading KPI on its own. For example: If you increase sales to 200% and increase visits to 400%, your conversion rate has halved, but you may be very happy.
- It's very important to segment your visitors when looking at conversion. For example - conversion is more important for new visitors than it is for older, loyal customers.
Starting a conversion optimization programme
Q4. And when looking to make these improvements - where would you start?
It's perhaps a crude analogy, but in any user journey and on any page, there are various levers you could pull to try and change the way visitors behave: words, images, layout, price, trust factors (reviews, testimonials), links to other pages, among many others.
The hard bit is putting together a process to identify where to pull those levers & which ones to pull. Digging into analytics and gathering user feedback are the two best ways to start that.
- Analytics - looking for anomalies is a great place to begin. For example, if most of your category pages have a 20% exit rate, but one has a 2% exit rate, take a look & ask "what's different here?". Google Analytics has 3 excellent features that make this process so much easier: Advanced Segments, Advanced Filters, and Pivot Tables.
- User feedback - Running surveys regularly and usability testing are both essential & can throw up all kinds of headsmacking insights. These are so easy now with sites like surveygizmo, usertesting.com & whatusersdo.com.
Surveys are also great from another angle: They can give you metrics to focus on other than just direct sales. Eg, "Customer Satisfaction" which you can benchmark and work to improve.
Something I once heard which highlighted why we should always ask our visitors questions: Walking round a shopping mall, we'd never assume we could guess what all those hundreds shoppers were thinking, yet we routinely do exactly that with our own websites.
Product merchandising in conversion
Q5. Would you include product merchandising in 'conversion' - where does that fit into this?
Yes, definitely. I'd say it's one of the most important things to work on. Most visitors come to ecommerce sites to buy products. Working to improve where you place key products on the site should be a given.
(A couple of "offline" books I love in this area are Paco Underhill's "Why We Buy" & Richard Hammond's "Smart Retail".)
If you're merchandising totally manually on your site, the trick is to set up a process for continually updating featured products that (firstly) keeps the site fresh for regular, loyal visitors and (secondly) shows your most likely to sell items as early as you can to new visitors.
An easy starter for that is to refresh your homepage/category pages weekly, and run them on a 3 or 4 week cycle. Each week, feature products that sold well last week, but haven't featured in the prior 3.
One simple metric for tracking that is "performance of merchandised products vs site average". (ie, what was your site average increase/decrease in sales vs last week, and how did your merchandised products compare to that).
And a few extra things you might want to look at online are:
- Seasonal trends - what sold this time last year? If you merchandise it a week earlier does it bring sales forward at all?
- High converters - if you can create a report showing you "views for each product vs sales", you can easily calculate conversion at a product level. Fix your lowest converters & show your highest converters more prominently.
- On-site search - what are visitors looking for?
- Add in other metrics: merchandise by highest revenue, margin, unique sales, highest product review, based on your priorities.
Moving on from that there are plenty of interesting automated tools for this, but that's a whole article in itself!