New & Powerful: AdWords Search Funnels

Last week Google announced a new AdWords reporting feature called ‘Search Funnels’ that has been getting a lot of attention.  At Enquiro, we were given a sneak preview of this new tool a few weeks ago and were able to use it with one of our major clients to get some more complete insight into paths that searchers are taking as they interact with advertising and the client’s site.  This enabled us to generate some ideas for optimization based on data, rather than conjecture.  So this increased visibility is most welcome!

Some brief notes on Search Funnels:

• collection of 7 reports that include Assisted Conversions, First and Last Click Analysis, and Path Length
• encompass only paid search activity (do not include clicks on organic results)
• provides info not just on clicks, but also ad impressions seen
• require AdWords conversion tracking or goals imported from Google Analytics into AdWords
• reports look back 30 days from the conversion event
• will be rolled out to AdWords accounts over the next few of weeks (accessible via Reporting > Conversions > ‘Search Funnels’ link on the left, below ‘All conversion types’)

Using ‘Search Funnels’, AdWords advertisers will finally be able to look beyond ‘last-click’ attribution to answer questions like:

• how many times did a visitor click on an ad to visit the site before completing a conversion?
• what keywords did the visitor search on and use to visit the site (via paid search) prior to completing a conversion?
• what keywords did a visitor search on and see an ad for (without clicking) prior to completing a conversion?
• for a given keyword (ad group, campaign), what is the ratio of assists to last click conversions? (closer to ’0′ indicates the keyword is a ‘closer’, while a high number indicates the keyword is pulling searchers in to the top of the funnel)
• after visitors clicked on a given keyword to visit the site, what were subsequent keywords that they used to get to the site prior to completing a conversion?
• how many hours or days passed before the first ad they clicked on and the eventual conversion?

Especially for B2B marketers, with relatively long (and sometimes twisted) conversion paths, often involving a string of generic and brand terms, having access to this information is golden.

To see how it works in action, check out this Google video:

Worried About Google Analytics Opt-Out? Nah (At Least Not Yet)

On March 18, there was a brief announcement on the Google Analytics Blog giving a “head’s up” on Google’s plan to release a browser plug-in that will allow web users to opt-out of Google Analytics tracking.  Naturally, this sparked some vigorous commentary, with opinions ranging from ‘disaster‘ to ‘non-issue’.  (Haven’t actually seen anybody – at least any internet marketers – suggesting it might be a GOOD thing.) Eric Peterson had probably the most complete coverage of the issue, with some astute observations regarding Google’s privacy motives being tied to their interest in collecting data from US federal government sites.

My own inclination is to side with those that believe this will have little impact on web measurement for those employing Google Analytics on their sites, for the following reasons:

1. Low Usage: The opportunity opt-in or opt-out is largely ignored by humans, who tend to go with the default, as Dan Ariely has so convincingly pointed out in recent years.  If it works that way for organ donation, we can be pretty confident that is how people will respond to analytics tracking.  Especially since Google Analytics is already set up to not collect personally identifiable information.

2. Existing Limitations: Web analytics data is already fraught with limitations caused by use of cookies, javascript, and half-baked implementation.  These will likely continue to add up to more impact than any opt-out system.

3. Trends: Even if there was some initial adjustment as masses of users opted-out, there would still be enough data for most sites to establish valid trends moving forward. And data trends are arguably more valuable in web analytics than raw numbers.

4. Strategic Interest: Google has a strong interest in encouraging widespread usage of Google Analytics and has made huge efforts in the past to make this tool as attractive as possible to as many site owners as possible. Unlikely they are going to put all that marketshare at risk.

So this is definitely something we want to keep an eye on in order to determine the implications as details are revealed and the program actually rolls out.  We’ll watch, but we’re not worrying – at least not yet.

After all, as Mark Twain said: “Worrying is like paying interest on a debt you don’t owe.”

Why Pie Charts are Evil

Pie charts are evil in the same way that the devil is evil: through mental trickery they beguile you, clouding your judgment while giving you the sense that you are making perfectly rational decisions.  Pie charts obscure the very data story that they are meant to tell.  And that – for visual data representation – is truly evil.

Breakdown of Pie Chart Evilness

Everybody knows that pie charts look great – especially 3D! – and no doubt there are some situations in which they can be used to get the message across.  But they come up short on a critical component of data visualization: context.  The only context present within a pie chart is the relationship of individual data points to each other. But this is usually only part of the picture. To gain an understanding of what is going on in our data universe, we almost always need to be able to monitor trends over time. And this is where pie charts fall apart.

It is interesting to know, for example, the distribution of website traffic among various sources, as represented in a pie chart.  But this is not a static relationship – it is likely to be changing and the important thing is to be able to understand how it is changing.

Recent example:

At a presentation to reviewing the progress of search marketing efforts, one slide included the following chart:

All well and good, provided out-of-the-box by Google Analytics so it is easy to grab, and it gives a snapshot of where things stand.  But it tells us NOTHING about any progress that has been made (or not) in driving traffic from search engines to the site.  Not quite as sexy, but your basic bar chart can tell us what has been happening much more effectively.  Here are a couple different bar chart options, depending on the aspects of the story that you want to emphasize:

Alternatives to pie chart - ah, now I see it!

Bar chart on the left shows percentage breakdown and now we can a) clearly see the relationships, while b) observing that traffic from Search Engines and Referring Sites is growing, as a percentage of total traffic, at the expense of Direct traffic.  Bar chart on the right shows us more clearly the degree of change from one month to the next in each of the categories. Now we can see what’s going on and base our decisions accordingly. Each one takes up about as much space as the pie chart, yet delivers much more valuable information.

More often than not, when you’re tempted to use a pie chart, a less glamorous visual tool may do the job better. This is why hard-core data visualization masters such as Edward Tufte and Stephen Few (.pdf link) renounce pie charts in all but very particular circumstances. And so should we all.

Choosing KPIs: Visitors or Visits?

Recently had a client situation where we were providing the client with monthly organic visitor numbers to their ecommerce site. One of my colleagues showed me a report received from the client, which had the same data, but marked as ‘visits‘.  So that got me thinking…what do they really want to measure here: visitors or visits? And do they know? And have they thought about what difference it makes? And, of course, what recommendation can we provide?

Both numbers are important (although they may not be critical – depending on the outcomes you need to measure) and they provide similar information, but there are some important differences. Leaving aside the argument over whether either of these satisfies the criteria to become a real KPI, let’s consider the uses of each metric in the context of this client.

Visitors

I’m pretty sure that since the reporting was done on a one month period, that the tool the client is using reports ‘unique visitors’. (i.e. People – or at least browser cookies – that are only counted once during the period.)

[Side Note on 'visitors' in Google Analytics:

For Google Analytics, apparently the term 'visitor' is not enough and they even go beyond 'unique visitor' to insist they are reporting  on 'absolute unique visitors'.  Of course, this over-states the case, given the limitation of cookies. But, ok, we get it, this is your best count of individuals visiting the site during a given time period.  More confusing terminology is used in the 'New vs Returning' report.  This is reporting visits, rather than visitors (as explained on the Google Analytics Blog) but the term 'visitor' is also used.  So maybe it would clear things up to refer to 'New Visits vs Return Visits'. ]

It is good to know how many people have come to your site, just as it is good to know how many people walk into your store in the mall. It gives you an idea of the total number of customers/potential customers that you are drawing in, and allows you to compare trends over time to spot opportunities or problems.

But there’s a big difference between a person poking their head into your store on their way to the food court, then never to returning again, and a person who repeatedly makes the trip to your store, even if they don’t purchase something every time. And this is where I think visits may provide more relevant, actionable information than visitors for this client.

Visits

As always, metrics that warrant attention vary depending on the nature and goals of a site. The client I’m talking about has a B2B ecommerce site that sells a broad mix of commercial products, including many that represent ‘repeat‘ or ‘modified repeat’ purchases in Buyersphere terms.  So, yeah, it is interesting to know how many people visit the site and to hopefully see this grow over time, but more critical in this case is the number of visits.

We are looking at organic traffic, and we are trying to use search engines to drive as many visits on as many relevant search terms as possible to the site. New visitors, certainly, but if we can capture visits from searchers who already know the site, so much the better, giving us the opportunity to further build on a relationship already established.

Further, we are already using this logic in measuring paid traffic, by counting ‘clicks’.  Not necessarily the same as visits, but likely to be closer to visits than it is to unique visitors.  So comparing organic visits to paid clicks may not quite be apples-to-apples, but it is at least apples-to-pears and pears are more like apples than oranges. (Visitors being oranges…you get the idea.)

Where the Rubber Meets the Road: Conversions

We all know – because Avinash has drilled it into us with his trinity approach – that it is essential to move from clickstream data to outcomes. (And, to be fair, virtually all leading web analytics advocates promote a similar philosophy.) So the number of visitors is interesting, the number of visits may be more so, but we need to get to the real reason our site exists: conversions. In this case, purchases.  And to make decisions about optimization and resource allocation, we need to understand the efficiency of various channels bringing visits to our site and this means: conversion rate.  And to get a conversion rate that makes sense, we need to have the most appropriate denominator.

Which brings us back to visitors vs visits.  Yes, it can be useful to know what percentage of unique visitors in a month made a purchase, but wouldn’t it be more useful – in the case of this B2B ecommerce site selling repeat purchase products – to know the percentage of visits that resulted in a purchase?  For a lot of B2B sites, the purchase pattern may resemble that of a car dealership: long consideration phase involving multiple visits, probably multiple decision-markers, (hopefully) culminating in a purchase that will serve the buyer’s needs for a lengthy period. This particular client has a site that is more like an industrial grocery store.

So in terms of organic traffic, it is quite possible that the same visitor may return to this site several times during a month searching for different products (in fact, there could be several different searches during the same visit, so visits are not the same as searches, but probably close).  If we really want to understand how efficient our site is in converting organic traffic, we should be calculating conversion rate = orders / visits.

This also helps us compare organic search engine traffic with paid search traffic, where conversion rate = orders / clicks.

Conclusion

Focus on visitors or visits, as appropriate to site type and objectives, but do so consciously. Recommendation for this client: switch from a focus on organic visitors to organic visits.

It might even be worthwhile to consider tracking and analyzing conversions against visits for some keywords (‘repeat’ purchase) and against visitors for other keywords (larger, less frequent, or ‘blank-slate’ purchases).

For other sites, visitors – or a visitor segment – may be more relevant.  With reference particularly to non-ecommerce sites, Anil Batra has a great blog post on how to dive in and select the appropriate denominator for your conversion rate.

3 Charts for a Potent PPC Dashboard

Whatever tool(s) you use to manage your PPC campaigns – search engine interfaces, desktop clients, campaign management software, or some combination – you get a lot of analysis and reporting power.  However, virtually all of these tools still come up short in terms of providing concise yet insightful overviews of PPC activity that can be presented to stakeholders on a periodic basis and quickly comprehended.

So you inevitably end up exporting data to Excel for manipulation. But then what?

Here’s an approach relies on 3 easy to interpret charts that cover all the essentials: overall volume metrics, cost efficiency metrics, and operational efficiency metrics.  While these charts may not answer all the questions we need answered, they provide a comprehensive summary of campaign activity that can be used to quickly:  a) get a sense of what is working (or not) and b) figure out where the next level of analysis needs to go. (The example below is for a lead generation site, but you could easily add revenue/ROI metrics for an ecommerce site.)

1. Cost & Volume

First thing everybody wants to know is: how much money are we spending? Second thing is: what are we getting for the money?  This chart answers those questions in a straightforward fashion, using 2 vertical axes as necessary.

Pretty obvious what is going on here (the point is to make it obvious): spend is up dramatically in Dec and clicks are increasing even faster.  In this case, a result of broader use of the content network. Cost has been close to budget – may be within tolerance or may require further explanation. Conversions also increased strongly, although we have to use some caution in comparing the rate of change, as conversions are on a different axis.  Still, positive movement all around (hence the green number disc), so let’s move on to the next chart.

2. Cost Efficiency

Here we move from the ‘how much bang?’ to ‘how much bang for our buck?’  Cost per click has come down due to our careful bid management and, more importantly, cost per conversion is trending downward as well.  However, there is still work to be done, as we can see the cost per conversion is above target (even though the target increased slightly for the Christmas season).  Let’s keep that in mind as we move on to the next chart. (The number disc is yellow – reminding us that we need to watch this one.)

3. Operational Efficiency

Now we take a look under the hood to see how the engine is running (to switch metaphors in mid-stream). Our click rate was already disturbingly low (not good for quality score) but we are somewhat okay with that as we are aggressively qualifying visitors for this campaign.  Not surprisingly, with more content network we are seeing even lower click through rate in December than November.  More distressing: slide in conversion rate.  Previously hovering around our target, it has now dipped way below.  So our engine is in need of a tune-up – and the number disc is marked in red to focus our attention.  All things being equal, if we can get the conversion rate up, our cost per conversion will decrease, fixing the problem highlighted on chart 2.

Next Steps

With this clear overview of the status of our PPC campaigns, we can identify the next steps to take in order to figure out how to improve performance.  These are likely to include:

• Looking at the traffic sources to see if we need to pull back advertising spend and/or lower bids on search engines that are not performing.  Maybe the content network is not giving us the quality we need at a cost we are willing to pay? Maybe we can improve this by weeding out high traffic/low conversion placements.
• Drilling down into the campaigns to zero in on the campaigns/ad groups/keywords that are contributing most to the lower conversion rate.  Does messaging need to be adjusted?  Do we need to refine keyword match types?
• Following through to landing pages and looking at bounce rates to see if there the messaging is appropriately aligned and conversion path unobstructed.

The main thing is, we have a solid foundation on which to proceed and a consistent framework that we can use for assessing next month’s results.

Numbers are Not Enough

The numbers are critical, but they rarely tell the story by themselves.  It’s important to include analysis along with the charts.  One of the reasons for number the charts is so that comments can be associated directly with a chart.  It’s best when these comments are maintained on the dashboard over time for continuity.

Dissatisfaction with Web Analytics Vendors

The Web Analytics Association presented its 2010 Web Analytics Industry Outlook in a webinar on Jan. 13.  The information presented was based on a survey of WAA members as well as the broader web analytics/online marketing community conducted in Nov, 2009.  Lots of interesting stuff on this growing, evolving industry, with an overall sense of optimism about future potential for web analytics as emphasis gradually moves away from data collection to data-based business decision-making.

Here’s one chart that I found particularly interesting:

Source: web analyst survey undertaken in Nov, 2009 presented in webinar "WAA Survey Outlook 2010 Report", Jan 13, 2010

This was presented as a positive result for large scale solutions, with the commentators noting that the large players have a higher ratio of satisfaction than almost all other categories, except for free analytics.

If I was a large vendor, though, I wouldn’t be too quick to pat myself on the back. In fact, I’d be very nervous about this result.  Applying the philosophy that ‘only the paranoid survive’, I’d be horrified to note that almost 25% of customers are not satisfied: almost 1 in 4! That is not a foundation on which to build a successful business.  

Even scarier: more users are satisfied/less users are not satisfied with FREE solutions!  People who are paying thousands of dollars per month for web analytics solutions are LESS SATISFIED than those who are paying $0.

Now, maybe this is a function of expectations being higher for expensive, large scale solutions. Perhaps those using such tools are more sophisticated in their requirements.

Any way you want to spin it, though, with the narrowing of the gap between the capabilities of Google Analytics, Yahoo! Web Analytics, etc and SiteCatalyst, Coremetrics, etc, if I were one of the latter, I’d be working harder than ever to keep my customers happy…while looking over my shoulder.

[Note: WAA members can access an archived version of the webinar, and/or download the slide deck in the members area of the WAA website (eventually).]

No! Basic Google Analytics Tracking on Your Sub-Domain is Not OK

Big No-No

Widespread access and ease of use, not to mention lack of price tag, have made Google Analytics ubiquitous on modern commercial websites.  It’s easy to get up and running, fun to watch the data come in, and has the look of very sophisticated software.  But what many companies are failing to realize, is that anything beyond a simple website will likely require modifications in order to ensure accurate, relevant data and to get the most out of Google Analytics.  For example, one issue that has come up with a few clients recently is the use of basic Google Analytics Tracking Code on a main domain (www.site.com) as well as several sub-domains (blog.site.com, support.site.com, shopping.site.com…).  This is a big ‘NO-NO’.

Why It’s Not a Good Idea

The problem is that using the plain vanilla tracking code will lead to data distortions.  This is because, by default, Google Analytics considers a domain to be a separate entity from a sub-domain.  A visitor that moves from a domain to a sub-domain will be: a) counted twice and b) identified in the ‘referring sites’ report as coming from your domain – neither of which are particularly helpful.  The situation is further complicated if the visitor goes back from the sub-domain to the main domain.

The 14,000 visits indicated from ‘xyz-software.com’ are actually visits from the main domain to the sub-domain – i.e. double-counting of visits to the site.  Here how the data breaks down in the above scenario:

* Google Analytics will keep track of all the referrals used in a session (through multiple trips in and out of the site, all within 30 minutes of the next), but it will only attribute the visit to the first one in the session.  Subsequent referrals will be assigned to ’0′ visits. See an example of multiple referral tracking here.

What To Do About It

Since this is a fairly common situation, Google has provided an easily implementable solution for sub-domain tracking that results in your domain and sub-domain being considered all part of the same site. All it takes is one line added to the GA tracking code:

The addition of this modification tells Google Analytics that any sites within the domain identified (‘.example.com’ in this case) should be considered as one. No duplicate counting, no self-referring. Done.

Note that in addition to the small change to the tracking code, Google also provides information on a filter that can be implemented to identify the domain or sub-domain of a given page in content reports.

Key things to remember:

1. This same code should go on ALL your pages of your main domain and any sub-domains you are tracking.

2. If you use the same file names for pages on your main domain (www.yoursite.com/index.htm) and sub-domain (blog.yoursite.com/index.htm), apply the recommended filter to distinguish them in your content reports.

3. If your site has been running for a while and has accumulated a significant number of duplicate visit counts due to traffic between the main domain and sub-domain, be prepared for lower traffic numbers under the new tracking.  It may be necessary to explain to your boss that this does not mean that actual traffic has dropped off, but that it was artificially over-inflated previously.  (Unfortunately, old data can not be retroactively adjusted.)

What If  I Want to Track My Domain and Sub-Domains Separately?

In some cases, you may consider your sub-domain to be a different site than your main domain and you may want to track them separately. That’s cool. It just means that you should set up a separate profile for the sub-domain using the ‘Add a Profile for a new Domain‘ option.   This method will provide you with a new UA number that will be used in the tracking code to differentiate the sub-domain profile from the main domain.  They will then be tracked as independent sites.

Now you know what to do to make sure Google Analytics is set up to track domain and sub-domain traffic in the way that best suits your needs.