Posts filed under 'Web Analytics'
Narrowly focusing on key performance indicators can have unintended consequences,
optimizing for the wrong things at the expense of something else.
For nearly every metric that an organization can focus on with web analytics, there is potential for backfire. Here are a few:
Engagement is a buzzword these days that refers to making an experience “sticky”, or that the user finds content so compelling that they develop a stronger relationship with a brand. Metrics like average time on site and page views per visit are often used to measure engagement but these metrics can be wrongly translated as frustrated visitors who can’t find what they are looking for.
Increasing visits to a site can’t be wrong can it? Not unless in an effort to reach a certain level of visits you engage in spammy tactics like click-bait headlines or spend lots of marketing dollars that in the end produces lots of unqualified traffic. Why praise increased visits if none of it takes the actions on the site that make you any money?
Measuring your advertisements reach refers to how many people are seeing the ad. The more the better is not necessarily true. In the age of precise targeting online, who sees your ad is more rewarding than how many.
Increased conversion rate is every site owner’s goal, but an unhealthy obsession can lead to marketing that only targets people who are the most likely to buy, and the most likely to buy are those people who are already familiar with your brand. Left unchecked, you never gain new customers as your marketing preaches only to the converted.
A high bounce rate is never good but continually changing content to decrease bounce rate can turn into an inverted curve – bad to better to bad again. Content goes from irrelevant to more generally applicable to banal and watered down.
Increasing revenue at all costs is the most obvious short-sighted metric and the result is what you see in many failed business: lower quality, less return customers, decreased goodwill, lower customer satisfaction, decrease in perceived brand value, the list goes on.
July 22nd, 2015
With web analytics you can see sources of traffic via your traffic sources report, but it’s not really
the sources of traffic.
With this report you can see the different sources of traffic like search, direct, email, etc. and their associated metrics like revenue, visits and conversion rate. This is beneficial to be able to optimize spend so that more money goes towards sources of traffic that have the best return. But there is an inherent limitation here that is often overlooked: the reason behind the user’s decision to access the site.
You know that a user came from paid search, but why did they decide to buy something to begin with? Did the product they were using wear out and they need a new one? Did a new life event happen that required them to make the purchase? Did they decide to buy after seeing the product used by a friend?
I think it’s important to remember that the ultimate motivation behind the user’s decision to come to the site is unknowable. Data from traffic sources is a blunt instrument at best to figure out how you can generate more sales. Directionally, it’s beneficial to invest marketing dollars to the extent that return on ad spend is efficient at the highest volume possible. But don’t be deceived because you can (sometimes) see what keyword they used in a search engine, what device they used, what the email offer was, the coupon code used, the product they purchased, ect.; trying to replicate winning scenarios doesn’t always return like a math problem does.
Online marketing is still marketing, not a math problem that can be solved if you could ‘figure out the data’. Customers are irrational and make purchase decisions based on reasons that web analytics data can’t explain. You can see what happened and as a result improve the site’s usability and marketing investment, but trying to increase why people buy is not as easy as web analytics makes it seem.
July 6th, 2015
On the surface in Adobe Analytics, the hierarchy variable and capturing site sections with channel and traffic variables seem to accomplish the same thing, so why do both? It’s smart to do both because the reports look very different to the analyst and can help them achieve different goals.
First of all, the hierarchy variable (“s.hier1=”) records site structure and is used to determine the location of a page in your site’s hierarchical page structure. It allows the analyst to start from the top of the site’s hierarchy and drill down through it. Once you start drilling down, you can’t see other groupings of pages at that same level outside of the hierarchical chain.
Site Sections are very horizontal in nature. They show you metrics for groupings of pages at a particular level across your entire site. The channel variable (“s.channel=”) is used to identify a section of your site. When sections have one or more levels of subsections, you can use additional Custom Traffic Variables (“s.prop=”) to identify such levels.
You could opt for only implementing the Site Sections via s.props and then correlate the different levels using the correlation function to drill down between levels. The only drawback is that the Page View metric is only available when breaking one Traffic Report down by another. If you wish to see Visits and Unique Visitor metrics at each level, use the Hierarchy variable.
November 4th, 2014
There are so many variables when it comes to ecommerce that I’m convinced that the idea of knowing why visitors do what they do is not really possible. All you can know is the results, not the why.
But not knowing why doesn’t sit well with our human brains. We yearn for patterns, explanations and stories to explain why what is happening is happening. Not knowing why also makes it difficult to get buy-in from others. If you’re trying to convince a manager to make a choice based on your data, you can be much more convincing with a story coupled with data, instead of just the data itself. Storytelling is a powerful tool, but if taken too far it can quickly go from presenting what happened to pushing an agenda.
The problem with creating stories with data that reach too far is called the narrative fallacy, made popular by Nassim Talbm in his book The Black Swan:
“The narrative fallacy addresses our limited ability to look at sequences of facts without weaving an explanation into them, or, equivalently, forcing a logical link, an arrow of relationship upon them. Explanations bind facts together. They make them all the more easily remembered; they help them make more sense. Where this propensity can go wrong is when it makes us think we understand it more than we really do and as a result, become more confident in a story that isn’t true.“
When you think you understand what the visitors on your site are doing more than you really do, you may start to let the data take a back seat going forward, and fall into the trap of confirmation bias where you start paying attention only to information that confirms your story while ignoring information that challenges your preconceived notions.
Somanyblogs are promoting the idea of telling stories with data without a bit of warning on the dangers of that approach. The world is a very complex place, and there is almost never a simple answer or a simple series of events to explain any action. In the end we don’t actually need stories to make decisions. Stop pushing agendas and get over who gets to take credit (I don’t think it’s a coincidence that the ones most interested in story telling are the same insufferable people who want to put off doing anything until after a meeting is held about it).
To make a decision, you simply need the ability to compare numbers and choose the best one. I don’t need to know why variant C was better than B, I simply need to know that it was 10% better and then I can take that insight, apply the change and move forward.
August 22nd, 2014
I recently read Malcolm Gladwell’s David and Goliath: Underdogs, Misfits, and the Art of Battling Giants
where he discusses the concept of the inverted U curve which explains how having more of something can be counterproductive. I think attribution analysis in online marketing applies to the same curve.
In my graph you can see that more information from attribution modeling does in fact produce better investment returns. But each additional piece of information yields less marginal utility – known as the law of diminishing returns – and, at some point, additional information begins to have the opposite effect.
Count of visits before purchase is increasing. Amount of sites visited before purchase is increasing. Time spent researching before purchase is increasing. Amount of devices used to access the internet is increasing. Amount of sales offline influenced by online research is increasing. Amount of time spent online generally is increasing. All these factors combined turn attribution analysis into a brutal rabbit hole.
Eventually, assigning specific value to any individual interaction is an exercise in futility. Gone too far and it will begin to take away value. I think the online marketing manager of the future will employ a kind of marketing mix strategy rather than a channel ROAS strategy.
October 30th, 2013
In Google Analytics under Audience > Behavior > Frequency & Recency you can find the count of visits report. Add the Built-In segment Visits with Transactions. This way you can see how many visits it takes for people to purchase. Compare year over year count of visits for purchasers. There is a good chance the amount of visits it takes for people to convert has gone up year over year.
A few thoughts on this:
- As the internet becomes more ingrained in our daily routines we all browse more. A visit to an ecommerce site is no longer a signal of high purchase intent like it used to be.
- Retargeting is more important. Visitors are taking more time comparing prices and sites. While they’re weighing their options it’s worth it to remind them of your offering with display ads.
- Do you offer discounts so often that visitors waiting for your products to go on sale?
October 14th, 2013
One reason to analyze traffic sources is to identify which sources have the most value and to generate ideas on how to get those high value sources to perform better. If you were to ask what is the value of $100 on AdWords, your analytics tool can give you an answer. Often paid search and other channels are combined along the path to purchase causing a multi channel funnel, but there is still a significant amount of sales that search is solely responsible for.
Not so with direct traffic because unlike other sources, it doesn’t work alone. Direct traffic is not really direct for two reasons. 1. The way all analytics tools work is that if they can’t identify the source, they will call the visit direct. 2. Even if it all really was direct – people typing the url in the browser bar, direct traffic is hard to analyze on it’s own because something else always has to cause it to happen. You don’t go typing in a URL in your browser without learning about that URL from somewhere else. You can never really know what initiated someone to come to the site directly. So direct isn’t really a source, it’s an action. A better label for direct would be unknown.
Direct being unknown is not necessarily a bad thing. Direct traffic should be used in conjunction with analyzing all other channels. All the work those other channels do will contribute to direct. This forces you to think of your site’s acquisition strategy in terms of an ecosystem rather than channels working in silos, independant of each other.
September 5th, 2013
When I used to do sales I would beat myself up about getting rejected. Then I started to keep track of how many people I talked to who were actual decision makers and realized I was mostly getting rejections from people who weren’t the decision makers and couldn’t buy anyway. This made me feel better because I realized my conversion rate was much higher once I got in front of a decision maker. The same applies to websites.
A less than 2% conversion rate seems bad when you think about it; 98% of people don’t buy. But if you were to whittle down the people who were actually there to buy your conversion rate would be a lot higher.
Take a look at all the different reasons someone would come to your site. There are visitors there to check their order status, contact customer service, browse, compare prices, find a store locator or are not interested in buying at all and leave after one page view.
If you can measure all the actions on the site that infer a certain kind of customer cohort than you can build something like this:
In this graph I’ve spit up all traffic into five segments:
- Service – visitors checking order status or contacting customer service.
- No Shot – visitors that show no interest in buying, visit <=1 Page. Who knows why they showed up we have no information on them to infer anything from.
- Browsers – visitors reading blog posts, relase calendars, looking at
- Buyers/Instore – choosing the pick up in store option, visiting the store locator, getting driving directions
- Buyers/Online – behavior that matches that of the Visits With Conversion segment
With this new view in mind you can measure accurately how well your site is converting those shoppers who actually have an intent to purchase. And you can accurately measure the task completion rate of those other visitors there to go to customer service or shop offline. You can also segment these buckets by marketing channel to see how many qualified leads each channel is providing.
August 23rd, 2013
Many ecommerce sites have blogs as a means to drive traffic, help with SEO, drive new visits, engage with customers, make sales, increase credibility and more. All of the goals can fit into two silos: engagement and driving revenue to the site. The two depend on each other – the better the content the more potential people will click through to the site and purchase.<
I built a dashboard for measuring the success of these two silos on a monthly basis. Check it out:
Click to embiggen
Measuring engagement has multiple facets. How well does your content attract new visitors, get people to come back, bring in others via social sharing and just be all around worth reading? The top left column pulls in those metrics of measuring how engaging the content is by using comments, shares and likes as a proxy for quality. Likewise the first row of bar charts shows the count of visits, days since last visit and page depth – pulled from Google analytics, over the last 4 months. Seeing these trended out gives you an idea if your content is getting better or worse over time – same with the line social graphs in the bottom row.
Next, how well does the blog get people to buy from your site? The first step to driving a sale is to get people from the blog to the site, so the Visits to Website line graph and the Blog to Site CTR shows how many, and with what frequency people are clicking through. The middle chart on the middle row shows overall revenue and per visit value.
A dashboard is only as good as the actionable insights one can glean from it. This dashboard shows (the numbers are all made up mind you) that last month the content got better – people commented more, shared more and Liked more. The next step here would be to pull the All Pages report for the month and see what kind of content resonated so much, and then make more of it.
Despite less traffic overall the quality of traffic to the site increased as Per Visit Value increased – too bad not more people aren’t clicking through to the site, maybe more links to the site could help that. Were the links to the site product pages, category pages, the homepage?
More questions: Did the new visits this month convert? The count of visits from last month were higher, was there a theme of content that you stopped using this month? Anyway you can add onto the content that drove the spikes in visits from previous months? How does your cadence of posting affect the volume of traffic?
Download the Ecommerce Blog Dashboard in Excel.
July 31st, 2013
Almost every marketing proposal and brief include one or more of the following meaningless goals: drive awareness and increase engagement.
Of course we want to drive awareness and/or increase engagement. What is marketing if not one or both of those things? The point of marketing proposals are to allow the thinkers on the team to collaborate and then give those ideas outlined to the doers on the team. Proposals, like most endeavors, are garbage in – garbage out. The more vague the objectives and strategy the more vague the results will be. So let’s how we can boil down “awareness” and “engagement” into unique actions that we can measure and get better results.
Objective #1 Drive Awareness
Awareness to who? Certainly not everyone. Moms with kids? Traveling businessmen? Baby Boomers? Teenager jocks? Go one step further and say “Drive Awareness to ________.” Then the doers can measure themselves against how well they are reaching that desired audience. Age and gender reporting are easily accessible in AdWords from the Google Display Network.
Also by using the myriad ways to overlapping contextual, interests one can measure the CPM and CTR of getting in front of the coveted demographic.Targeting new customers or return customers is also a consideration – how well can you reach past purchasers of product x who might have an affinity for the new product y? Remarketing lists can also be a useful tool to measure awareness. Getting the message in front of visitors who browsed certain categories or similar products can be counted as success. All of these tactics drive awareness and give you ways to measure how well that awareness was achieved.
Objective #2 Increase Engagement
Engagement is subjective and web analytics tools are inherently unable to measure the kind, positive or negative, of engagement and are left to only measure the degree of engagement. Measuring the degree of engagement is going to be unique to the experience of the site. Maybe it’s the amount of contest submissions, tweets, comments, video plays, likes, it all depends. There are a few standbys however: loyalty, recency, length of visit and depth of visit (all located under Audience > Behavior in Google Analytics. With all of these metrics see what the site average is and then use that as a baseline to achieve against.
So instead of being meaningless, an example objective on a marketing proposal could say:
“Drive awareness to our target 25 – 35 mothers with children who are trying to save time and increase the value of quality time spent with their families.”
“Increase engagement by reaching more than X contest submissions while driving X new visits through user generated social shares.”
July 26th, 2013