3. Tough Mudder PA #1 by The 621st Contingency Response Wing
http://www.flickr.com/photos/621crw/6984091812/
4. KEYWORD DATA
THE END IS NIGH
REGULATION
CHALLENGES
PRIVACY ISSUES
Etna Volcano Paroxysmal Eruption Jan 12 2011 by Gnuckxhttp://www.flickr.com/photos/purpcheese/4031751643/
iGaming is right at the forefront, and it’s tough
Can’t afford to pass up opportunities, make mistakes, or get left behind.
Challenges aren’t radically different, but the stakes are high.
Panic, ignore it.
Fine - lets you get on with day to day, but creates competitive risk
Keyword not provided ate away at our data for months and months, and it wasn’t until it hit ~90% that most people truly considered the implications and got on board with new ways of thinking. If you’d got their earlier, you could have had a competitive advantage. Spot opportunities and act. I’m a big fan of doing.
My role as Head of Insight; background in SEO and Analytics; an understanding of what needs to be done, how to measure it, and how to get it done.
I’m not an igaming expert. I’ve experience in running campaigns, assessing performance, but it’s not been a big part of my world.
Got me thinking about… what would I want, need? What tools, tips, tricks would be useful?
Let you get hold of my toolkit...
GET MORE FROM CHANNEL DATA and the tracking infrastructure you have in place.
- Already got loads of data, but missed opportunities to get more actionable stuff
- How do you get more/better/more connected?
- Lots of things we overlook with channel data!
Easy competitive advantage. Stuff you can do now.
Tips, tricks, tools, processes around tagging, tracking, analytics.
Who has a phone number on their website?
Who has a phone number on their offline collateral, TV?
Who’s using phone tracking?
Who’s using unique numbers per visitor?
Who’s tying this back into their analytics?
- Multiple numbers for multiple endpoints
- Expiry windows to generic number pools
- Output type classification
- Offline and API extension
- Call listening and lead scoring
Small fee + cost per call
Routes to destination numbers
Layers on top of existing processes
Opportunities to get clever with data feedback
Infinity = Great GA integration, API, data layer stuff
Offline, too?
Response Tag (formelly AdInsight) great technical tinkering for call expiry time pools, etc.
Auto channel classification - great, but terrible for GA etc
Logmycalls
Not used, but does intelligent call listening and lead scoring
Humanise your data
Demographic Data in GA. Not using GA? Put this on anyway.
Avoid a filter bubble effect
Other solutions are black box / machine learning. This is similar, but more transparent
- Apply advanced segments
- API
* higher chance to be blocked by adblockers; use a fallback script
Personas
Market to estimated persona, rather than actual
Who’s running GA / UA? / Done stuff with it?
Weather API, pass as custom dimension @ SASCON. Not unique. Process is back to front.
- The underlying approach is how most web analytics software has worked for years
Weather *might* affect your users behaviour.. And you might even come up with ways to capitalise on this. But this is a limited way of apporaching things.
Think about external factors which influence digital behaviour. Grab them. Do the research first; don’t expect to have an epiphany about what to track
- pass odds of bets placed with conversions. Segment GA traffic by riskier/safer behaviour?
- Which football team is top of the league?
Think about offline events which fundamentally affect your worldview
- Account cancellations / closure
- Big spenders / big winners / higher-than-averages?
Most of your direct traffic probably isn’t direct. We train ourselves to ignore it. “That’s nice”
- Desktop software
-- Social platforms
-- Un-tagged emails
--- Automated system emails!
- Apps
- Your own people (including you)
- Bots / uptime checkers
Segment it. Define sub-channels of direct based on identifiable patterns. Landing page. Technology. Visitor type. Engagement.
Who does TV advertising? … or has a competitor who does?
Mostly Direct. Vanity URLs don’t work. “Search For X” campaigns are weak.
- You’re trying to change behaviour in order to track it!
Do real-time analysis in the minutes around TV ads
- How does behaviour deviate from standard direct traffic (based on your sub-channels). Obvious stuff: Device. Browser. Resolution. Landing page. Network. Location. Visitor type (new/repeat). What’s left?
- Chances are that a chunk of this is your TV audience. Treat them differently. Assess their behaviour; they’re hyper-engaged, where do they drop out?
- Rinse, repeat, refine. Hone in on behavioural clues of TV advert responders to produce a closer segment.
Direct TV segment + demographics.
- Who do you think you’re advertising to, and who is responding?
API into a hub (analytics platform), API out to enrich individual arms. - Get phone conversions into your CRO platform, or your TV audience data in your CMS.
Enrich processes
- Market/audience/keyword research
- Design
- etc.
DirectionMonster.js
Carry campaign parameters through to destination in hashtag, collect and process at destination.
Re-attribute from dark web
Single Ecosystem?
Domain-centric technology
- Multiple domains?
- Localisation/Internationalisation?
- Device-specific environments?
- Distinct apps?
- Single ecosystem, multiple views
- from an analytics/measurement/performance perspective, you need the ability to see variances as single entities, or distinctly
What’s the conversion rate of your ‘about us’ page?
Page value / dollar index is a weak metric. Tells us what worked eventually, not whether each page did its job.
Because we don’t stop and assign jobs. “This page’s job is to convert!”. Not necessarily. It might be to ‘convince’, or to ‘route’.
2006
Different types of pages
Different types of metrics
Router pages - resurface rate. High means that the navigational process has issues!
~16 from ‘advertisers’ to ‘reassurers’ to ‘closers’
Reassurance exit rate (people who bail from a form to check delivery, terms, details, etc.)
Tag up at page-level, or crunch in Excel, etc.
Either blindly relied on, or dismissed.
Different context for different types of content, behaviour.
Can be hugely powerful.
The real issue is that it’s incredibly technically sensitive. In 1995 it was fine. Ish. No, that’s a lie.
Video. Assets. Social engagement. Device considerations & changes to content consumption types.
What we’re really trying to understand is “what proportion of our traffic, for this segment[!], did we fail to engage?”.
Turn it on its head. Use “engaged non-bouncing visits”, rather than ‘bounce rate’.
Event tracking (GA supports passive events); tag up all asset interactions, media consumption, page clicks, scroll+read page completion, time thresholds, the lot. Work out what ‘engaged’ looks like and tag it up.
Much better metric, and allows you to take baby steps to influencing top level conversion metrics
Task Completion Rate
Satisfaction Rate
“What did you come here to do today”
“Were you able to do it?”
“How would you rate your experience?”
Combine with demographic data; then you can practically automate site optimisation.
Identify weak performance by task, page, demographic, interest group, and commission content/outreach/advertising/whatever.
Test.
Thinking of testing….
Traffic which enters the site on a non-page asset (e..g., PDF, image) doesn’t fire any javascript tracking.
Implement server-side script, e.g., PHP GAPI vs HTACCESS using rewrite rules.
Serve your files through a central script, but rewrite to a pretty URL
Something like site.com/example.php triggers site.com/?file=example.pdf, which triggers the tracking then servest the PDF
Challenges with session/user stitching, but can overcome with Universal Analytics; see me after!
Referral traffic; generally a link click.
API call to GA data for referrals. Compare against your link data.
If not found, check if it's indexed.
If it's not indexed, link to it.
Remove when indexed?
Watch out for URL fragmentation, etc.
Programmatically monitor your rankings for mid-value terms which have slipped to the second page. Adjust your internal anchor text, and/or content, to pull these rankings back.
Detect copy/paste actions
Fire tracking events
Drop outreach phrasing into paste body
Add tracking parameters
Getting traffic for terms you’re not interested in?
Fashion site getting lots of adult traffic.
Used to be able to 301 it by keyword to an adult affiliate site. Gradually lose the rankings/visibility. Rinse and repeat.
Now by page - if you’re getting the traffic anyway, focus it. Get a big dirty landing page up. Then 301 the URL. Same principles apply.
How much money do you spend on PPC, and how much of that is lost/inflated?
Which organic keywords drive you the best traffic?
How does your offline activity - and the wider world - affect your digital performance?
Where does your traffic really come from?
How well does your website convert when you think of conversion in a more sophisticated way?