SES London 2010 | Search Analytics Deep Dive

Brian Clifton is moderating this time and our panellists are John Marshall of marketMotive, Tami Dalley, Acronym Media and Neil Mason, Fovian applied insights.

Brian’s a bit “phew!” from the last sesh so we’re pretty quick over to John.

John starts off with his experience, which is actually his Netscape user analytics experience (at Netscape) that he is drawing on most here.

At Clicktracks, built a web analytics tool as complete neophites. Quickly learned it goes like this… build massive tool, loads of data, spit out reports that nobody uses…

History of WA – fits in two worlds, data mining vs web analytics. John could not understand why these two worlds did not seem to talk to each other.

Reason – data mining tools deal with deep data (age, gender, postcode, household income.) Few people, but huge amounts of detail. Web Analytics is almost the antithesis. Massive amount of people but at much less depth. Web analytics is shallow (I like, so totally knew that.)

So… Why deep dive? How deep can we really get from WA?

Too often the reason is need driven (we need better data, fatter paychecks, better reports). There is still value in deep dive, but remember you CAN NOT turn web analytics into data mining.

Make your data deep. Find out and collate more levels of info from…

  • Login
  • Email
  • Volunteed info
  • CRM
  • Website


  • This is a complicated process
  • Not for everyone
  • Big companies truly suck at this

John puts this down to choice of tool. If incumbent tool exists path is dictated by the tool. Existing tools may not be what you want. In most cases, to go really deep – the company CRM is often the first port of call.

John summarises – Basically make sure you have enough data to warrant the exercise.

Q from Brian – what IS the problem with large companies like P&G? John thinks it is because these co.s have the right tools but approach web analytics from wrong angle.

Next up is Tami Dalleyfrom Acronym, who focus on WA, MV testing and UX

Tami disagrees slightly with John in that she feels it is possible to go deep with web analytics without too much additional data.

What is deep dive analytics? Mining additonal information to find the story.

Good analysts love to spend time wallowing around in data, sometimes without objective. So process needs to go…

  • Hypothesis
  • Define scope
  • Pull data
  • Follow the crumb trail
  • Translate into action

Popular Techniques

Segmentation (Geo, High value visits)

Example: Geo insights for PPC

Q – How should I change my market budget allocation by geography?

Process – focused on KW cluster (we’re looking at a bubble graph where size of bubble is size of order value, with CVR on vertical and CPC on horizontal.) Tami shows us a comparison where we compare the same graph per market – which shows dramatic difference.

So: let’s look at messaging within KW clusters to understand more.

Example “free shipping” had 2x ROAS than EU and N.America (for concerts but not gift keywords.)

We’re moving on to a really cool representation of demand by KW cluster by regions; where each region has it’s own colour on a bar, and colour area width denotes demand per region. Looking at conversion by KW cluster, should also be considered against volume demand per region.

Travel client asked them to help improve conversions. Conversion funnel analysis example (obviously important in a climate of doing more with less): looked at select package stage, identified high converting packages and made recommendations.

Final good practise tips:-

A/B test, not just usability to inform recs. Good data sample size, garbage in = garbage out.

Brian is introducing our final speaker of the day – Neil Mason of Foviance.

Segmentation is the theme of this presentation, which is a particulalr passion of Neil’s. Neil thinks this has it’s roots in 15+ years of offline marketing.

In WA we tend to focus on metics from a site perspective. We use bland meaningless metrics like time on site, page views – which creates a one size fits all experience.

In a fitting median point between John and Tami’s differing perspective Neil is kinda in the swimming pool.

What do we mean by Segmentation?

Leveraging knowledge and insight to make things more interesting and relevant to users. So in that case – what “segments” are useful?

Traditionally; gender, age, household etc

Another way; attitudes, buy again liklihood, need states, brand empathy, category empathy.

Third classical way – behavioural; browsing, purchasing, visits (recency, frequency). (WA professionals are good at this!)

Neil urges a slice and dice approach. We need to “get into the guts” of it to understand it. Tools are getting better so there is no excuse! Adding in demographic data is not easy but it is do-able. Find out where data exists within organisations. Attitudinal and behavioural survey data may exist elsewhere in the business.

Real value comes from data synergy. As an example one publisher segments “daytime online” and “weekenders” as these two types have wildly disparate behaviours.

Lesson: look for differences, in data and behaviour. Segment attitudinally, segment behaviourally and segment demographically.

Brian wraps up with asking the audience for questions. Nobody is forthcoming so John gives us a bit of food for thought. Just because a deep dive is useful, going deeper may not be more useful. It might just be a total waste of time.

Leave a Reply

Your email address will not be published. Required fields are marked *