Allison Hartsoe from Semphonic
Conrad Saam from Avvo
Allison Hartsoe from Semphonic
+ Passing data into web analytics tool requires configuration, doesn’t always give data you need by default.
+ In order to create measurement framework, need the full picture
+ For retail sites usually not a problem because most carts are relatively optimized and metrics are straightforward
+ Use proxies for success for softer conversions such as time on page – relevant for media, government, etc
+ Total value of pages viewed a good proxy for success OR a measure of engagement such as time on site
+ Measure to lifetime value – not just looking at visits, need to understand what visitors are doing over time
+ Tracking to lifetime value can be very difficult – can take a very LARGE set of data or look at sales cycle to look for key metrics that signal conversion behavior. This is essentially creating a segment.
+ Once segmented, understand and test to see how early users will eventually convert
+ Branded Search –> directed product research; Non branded search –> Early stage research
Coding for Measurement
+ To effectively use the analytics tool you need to ensure key variables are getting passed from your search program. This may take the form of a campaign code (such as utm parameters for Google Analytics). Can deconstruct traffic source, campaign name, keyword, etc.
+ In Omniture you can upload a decoder to translate query parameters
+ Omniture: pull the s campaign variable by the GetQueryParam plugin. Set tracking code id in admin.
+ AdWords Inside Google Analytics – ensure accounts are linked; recommends Tatvic GA Excel plug in that allows you to automate advanced query requests and populate charts.
+ Integrated report data – aggregating data from various sources and calculating metrics for a more comprehensive view of campaign performance and tracking lifetime value.
+ Key SEM capabilities – for broad match consider tracking actual in addition to the keyword that converted, for ad networks track content match source, for long tail – collapse search terms and group them (branded vs. non branded for example).
+ Site behavioral reports – ability to path over time at the event level (x visitors entered on PPC search and Y entered on visitor searched); Over time report (behavior of PPC visitors during a particular time period in a subsequent time period)
+ Types of Analysis – SEM traffic, Organic vs Paid, Traffic by Engine, Traffic Trends
+ Traffic by Search Engine example – understanding organic effect on PPC. When you look at source by engine it shows that certain engines are underrepresented. What is the identification of the SEO problem by engine? Do we need to better optimize better for Bing?
+ Analyze by Channel – organic cannibalization example: shut down PPC altogether (dark period) to assess organic traffic and PPC traffic cannibalization. Are you overspending on certain PPC terms?
+ Analyzing Quality – engagement example: different qualification rates for the visit (submitted lead form vs. download whitepaper). Reviewing drop off rate to optimize performance of campaign.
Things to Remember:
+ Make sure you are optimizing to the RIGHT metrics
+ Carry through as much info as possible through proper configuration
+ Range of analytics prospects to ensure success
+ Campaign stacking and how to easily set up – set up as custom evar to allow to track event; insert easy separator to stack and be able to separate out. Be careful how long campaign names are, usually limit is 5 or less.
+ How long would you turn off PPC for “dark period”? Minimum of 2 weeks however possible to do less for clients that have a TON of traffic. Need enough volume for a reasonable measure.
+ When do you see Google Analytics not being adequate and you have to use Omniture for example? GA is an amazing tool however it falls apart in terms of sampling. Omniture and Webtrends have a limit in tables of about 500k. For Google, their sample is about 50k per table however it is boosted when running AdWords. The other area is the fact that you can’t aggregate data the way that you can with Omniture. Although they have custom variables, it’s usually not sufficient considering the amount of custom information often needed to integration. In short – data sampling and robust conversion tracking.
Conrad Saam from Avvo
+ Coming from in house perspective vs. agency
+ Ranking Report Rant: personalization, local, query variations (ranking reports for one targeted keywords can result in missed opportunity for great keywords that are ranking and driving traffic), the long tail (ranking reports don’t capture)
+ Think about your ranking reporting structure: 1. ubercompetitive keywords and online reputation management – track rankings, 2. Head terms – measure traffic, 3. long tail – think about traffic, # keywords, # of entry pages and # of indexed pages
+ How much SEO traffic do i have? MUST separate branded from non branded traffic. Direct traffic and branded usually perform quite similarly.
+ Is my site what I think it is about? Not necessarily – check Google Webmaster Tools for most common keywords, is there a disconnect from what you expected?
+ Is my site authoritative? There are tons of metrics to help determine and track (SEOmoz is great but so many numbers – consider choosing just a couple and stay consistent); look at log files and track number of pages crawled.
+ How does my site’s authority stack up to competitors? Your site’s data alone with no comparison doesn’t mean much. Look at the total number of links and number of domains linking (diversity).
+ A data warning: many link counting tools report inconsistently which ends up comparing apples to organges data. Graph % of links and % of domains over time which is great but sometimes data is skewed, however no good solution so far.
+ Am I a good link builder? Calculate % of links that go to home page. If you have really valuable content more links should be deeper within the site.
+ Was that a valuable link? Look at the traffic it generates. Basic but true.
+ Have I lost some links? In GWT, are links going to 404 pages? Check these and fix!
+ How spammy is my link profile? Look at the difference between your mozRank and mozTrust – if big discrepancy then maybe a problem.
+ Should I give up on this keyword?
+ Do I Nail the Long Tail? Track the NUMBER of keywords and landing pages driving traffic. Also track how many pages are being indexed. Search engines may be crawling but not necessarily indexed – this represents the possibility of wasted crawl budget.
+ Does my site look fat in these jeans? Site performance in GWT as one of ranking factors, also YSlow is an amazing tool for this.
Will prospective partners take my call? Using PageRank to determine authority of your site doesn’t mean much however everyone takes a cursory look at this metric so it does matter on some level.
+ Are there any commonly overlooked methods to increase your site’s authority? Think about your PR agency as your best link building friend. Finding ways to marry PR with link building can be surprisingly effective.
+ Why do you have to separate branded data when many product sales are based on that? Because behavior is extremely different (friends vs. strangers) and they will not convert the same.
+ How did KPIs change over time? Yahoo Site Explorer was garbage – too much data fluctuation. For a long time obsessed with ranking reports until realized correlation between ranking reports had nothing to do with traffic and engagement.
+ Pages crawled vs. pages indexed, how did you get pages indexed? Work on authority and link building, realized that certain template pages weren’t providing that much value – worked to increase content and value of those pages to get them indexed.
Rachel Andersen works for the Portland based SEM agency Anvil Media, Inc. She has expertise in all aspects of search engine marketing and specializes in SEO for large sites. Andersen has been responsible for the development and execution of dozens of search and social marketing campaigns over her time spent with Anvil. She also is a regular contributor to SearchEngineJournal.com