Monday, June 21, 2010

Segmentation and Personas Part 1

The Web is all about choices and Analytics is all about understanding the choices by analyzing the behavior patterns:

Who are my right customers? Why do they make the choices that they do? What opportunities are worth chasing? What features will provide the most value? What is the best time-to-market, and most important which customers are most important.

While it’s difficult to make ALL the right choices, one has to make most of them right for the campaign, site, or product to succeed.

Enter Personas

This is a must have tool for marketers for helping them make the right choices. Personas is essentially a technique for capturing the important learning(s) from analyzing users and customers and identifying and understanding the different types of people who use the site.

It is a description of an imaginary but very plausible user that “personifies” these traits. The three key traits which personas help identify are:
Behavior, Attitudes, Goals.

How can Personas Help

  • Rallying – Personas can help you build a common vision. If you look at any website, there is so much information that most marketers look at it at an aggregate level. There are literally thousands of details about a user (think pathing, content affinity, referring sources, tools usage, yada yada yada). An analyst, can’t possibly measure or conclude key behavior traits by analyzing these data points. Personas can help group these types at a high level and provide a human face to the types of people.
  • Testing and Optimization – As a side benefit, you can test specific creative tactics on these personas. Messaging for the Net Generation (Gen Y) would be different than the messaging for Gen X (Baby Busters) . If you can identify the key behavior traits by type then you can build a test plan around it to drive conversions on the site.
  • Targeting – Anil is one of the great thinkers around the topic of One on One personalization. His entries on Behavioral Targeting are a delight to read. Imagine, armed with “statistically significant” behavior data (or traits) you can proactively market or provide content to drive usage and ultimately convert the user. That’s the ultimate promise and the holy grail of marketing isn’t it?

Are there any cons or pitfalls

There are obviously things to watch out for. The biggest ones are prioritization and validation:

As a practitioner I have seen that a lot of research, thought, and homework done in building personas. Usually Strategy gets involved along with Product/Brand Owners and Senior members of the client and is handed down to executors.

What is important to understand is that your website is not for everyone. People will come to your site, tease you, perform competitive shopping, look for products, research about products etc.

It’s not okay to say that your website is for everyone. If you think that way you are deluding yourself. This is extremely difficult for most marketers to grasp. I try hard to explain to my clients is to focus your efforts or release or a landing page on a single persona. It doesn’t mean that the website or landing page will not be useful or usable by others, but if you gradually build the user experience around each type or user profile you will ultimately do a great job in attracting the highest quality users to your site. That’s the promise of Qualitative surveys and Quantitative analysis.

Another pitfall I have seen is that teams create personas based on their "assumptions" which usually comes from the Highest Paid Person In the Organization (HIPPO). In my mind this is guesswork and not based on any analysis; the right thing to do here is to take time to analyze your web data, interview/take time to talk to real users and verify if these personality types or personas really exist.

So are you using or thinking about personas for your website? Does your agency recommend taking the approach? Share it with us.

Next time we will talk about different techniques both qualitative and quantitative to measure personas.

Thoughts? Comments?

This is a guest post from my friend and ex-coworker Kanishka Surana. Kanishka is currently the Head of Web Analytics for Ogilvy. He runs Ogilvy's Web Analytics group in North America, he has been in the Web Analytics space since 2002 and has worked in London, Seattle, Greece, Seattle, San Francisco, Salt Lake City, and most recently in New York.
Before that Kanishka worked at Gerson Lehrman Group a pioneer in proprietary research space. Kanishka and his wife Mini live in New Jersey.
This is first of a series of posts that Kanishka will be doing on this blog.


Looking to fill your Web Analytics or Online Marketing position?  Post your open jobs on Web Analytics Job Board.

Current Open Positions

Thursday, June 10, 2010

5 Web Analytics Misconceptions

There are several misconceptions in web analytics (created by some author/bloggers/experts) though many others have tried to clarify them from time to time but they keep reappearing. I recently had a conversation with someone who was so much in love with one of the misunderstood metrics, listed below, that it prompted me to write this blog post. So without much delay, here are the most common five misconceptions that I come across all the time:
  1. More Page Views are good – Unless you are an ad supported site that sell advertising via CPM (cost per thousand impressions) more page views might mean that the visitors are lost on your site and can’t find what they are looking for. More pages views/visit could indicate issue with your site navigation. For effective analysis, set your baseline and then watch for significant deviations (up or down) from the baseline.
  2. All that bounces is bad – I have written 2 detailed posts showing why all that bounces is not really bad. Bounce rate is one metrics that people overly obsess with. Keep in mind all bounces are not bad. The things that cause high bounce rate are:
    1. Links to external sites that you want visitors to click
    2. Ads on your site take visitors out of your site
    3. Returning visits might bounce because they might come to your site to read your daily/ weekly/monthly update
    4. Visits that are for a specific reason e.g. find your phone number
  3. Focus on reducing the bounce rate and everything will be ok- Well that’s the advice many people give without even looking at the other data points and analyzing if reducing the bounce rate will really help you achieve your goal or not. Reducing the bounce rate might not be the most effective way to increase ROI. You should create a monetization model and determine the impact that reducing the bounce rate will have before you start creating different version of a high bounce page to A/B test to reduce your bounce rate. I have seen cases where you won’t get positive ROI even when you reduce the bounce rate to 0%.
  4. Time on site (or page) shows how much time people are spending on the site – As I wrote in my blog post titled Understanding the "Time Spent on the Site" Metrics there are many issues with measuring the actual time spent on the site or a page. One of the main reasons is that the last page that a user views/reads on your site is not counted in this calculation. So if you have a non-ecommerce sites then the chances are that the visitors spend most of their time reading the last page but that page won’t not counted in this metrics and hence your time on page and time on site metrics will be way off. As long as you know that you need to watch the trend instead of using this metrics as a absolute measure of time spent on site then go ahead and use this metrics.
  5. Referring Sites report shows all the traffic sources including campaigns – Well… not really. There are a lot of reasons for the referring source to be lost from the time the visitor clicks on the link to the time they arrive on your site, two big reasons are
    • Server redirects – This happens a lot with ad serving. Suppose you buy an ad though a 3rd party company who then uses an ad network to place your ads on a publishers site e.g. yahoo, each party does some processing and redirect of its own. In doing all these redirects the referring information is lost or shows one of the sites that does the redirect. For example, you might see atdmt.com showing up in the referring sites which means you were serving ad via Atlas even though the ad might have been served on MSN.com. Many URL shortening services used on twitter also show up as referring domain instead of twitter.
    • 3rd Party Apps – This is a big issue with Twitter URLs. A lot of twitter users use 3rd party apps and any clicks to your URL posted on twitter from these 3rd party apps will show up as direct traffic.

    If you are running a campaign or posting links in social media, blogs, forums etc, make sure to tag them with campaign identifiers so that you can use campaign reports instead of relying on referring sites report.

Thoughts? Comments?

image source: ct4me.net

Looking to fill your Web Analytics or Online Marketing position?  Post your open jobs on Web Analytics Job Board.

Current Open Positions

Wednesday, June 02, 2010

Compound Metrics in Web Analytics

Should you use a compound metrics in your web analytics reporting? This was a topic of discussion in one of my classes at UBC Web Analytics course.

What is a Compound Metrics?

Before we get into answering the question, let’s look at what a compound metrics is.
Simply stated a compound metrics is when you take two or more simple measures and combine them together to form one metrics.

So should you use it?

Short answer is - why not? Sometimes simple metrics such, such as visits, page views, clicks etc., are not enough to explain a complex concept and that's when you need a compound metrics, e.g. engagement metrics, visit quality measure etc.

But isn't compound metrics hard to explain?

It depends on how you define it but then a lot of people still struggle with
visits, visitors, hits and pageviews.
With that in mind, I agree that initially it might be a little hard to grasp the compound metrics. Over long term, compound metrics actually helps simplify the measurement of a complex concept.

Some of the common uses of the compound metrics are
  1. Credit Score (Everybody has one and it affects them every day but how many actually know how it is calculated?)
  2. Google Page Rank
  3. Twitter Resonance (it will use several factors such as retweets, clicks on links etc.)
  4. Twitter measurement. Many twitter measurement tools use compound metrics since it is not easy to explain Reach, Impact, Engagement in simple metrics.
  5. Facebook's "Likeability Index". Ok, I made this one up but I am sure that's coming soon.

Example

Let's look at an example and see where such a metrics will make sense in your current web analytics reporting.

Lets take an example of a product information site. The products are sold offline via 3rd party retailers and since this company does not want to compete with its retailers it doesn't sell anything online. This site provides information about the various products that this company sells. It provide white papers with pre-purchase information and post purchase information/support. The site has some videos and some stories that are published every now and then. Well there is a sign-up form to allow users to save their product information.

The whole goal of the site is to provide information to current and future customers.

Now your big boss asks "We spent thousands of dollars to build this site, is this site working?".

You reply, well... Visits are down but repeat visits are up so seems like people like it and are coming back. However, page views/visit are down. More white papers are downloaded as compared to last month. However, video views are down and sign ups are also down. Seems like some things are up and some things are down.

Boss goes...."What does that mean? Is it working or is it not? Are customers finding information?"

How do you measure that?

As an analyst you can look at all the metrics and come to a conclusion but you have to be able to convey the end result to the VP of marketing. He needs to know if the site is successful or not.

This is where a compounded metrics comes in handy.

A simple formula for this could be

(% of visits viewing X pages or more + % of visits viewing video + % of visits downloading white papers type 1 + % of visits downloading white papers type 2)/4

we used 4 in the denominator because we are using 4 different metrics with equal weight

Now you can add other metrics that matter to the business and also assign different weight to each, so your formula could be something like:

(1*% of visits viewing X pages or more + 4*% of visits viewing video + 1*% of visits downloading white papers type 1 + 2*% of visits downloading white papers type 2)/8

1,4,1 and 2 are the weights assigned to each metrics based on their importance to the business.

Once you develop a baseline for your metrics, you can confidentially tell you boss if the sites performance is better or worse than the last month. Keep in mind that you are an analysts and you should always get under the hood to analyze each component and find opportunities for improvement.

What do you think?

Questions? Comments?

Other articles on similar topics are


Looking to fill your Web Analytics or Online Marketing position?  Post your open jobs on Web Analytics Job Board.

Current Open Positions