Reading Kaushik (Part 2): Digital Analytics

Data Quality Sucks, Let’s Just Get Over It

  • Data quality in the web is not great
  • Six step plan:
    1. Don’t dive deep into the data to find discrepancy between data
    2. Assume a level of comfort with the data
    3. Start making decision that you are comfortable with
    4. Over time start understanding small areas of data
    5. Get more comfortable over time with your data
    6. Absolute numbers rarely matter, segmented trends do

Tips for Web Analytics Success for Small Businesses

  1. Get top key phrases from search
  2. Get top referring URLs
  3. Which content is popular on your site?
  4. Percentage of Visitors on the homepage
  5. Check segmented click densities
  6. Learn about your site’s bounce rates

Measuring Success for a Support Website: A Point of View

  • Moment of Truth: hold or lose customers — web: often support problems
    1. Don’t measure unique or total visitors
    2. Identify top methods to customers find information
    3. Click Density Analysis for the Top FAQ pages
    4. What percent of site visitors call the support phone number?
    5. Measure: Problem resolution, timeliness, likelihood to recommend
    6. Check if the top solutions are corresponding to the top real problems (call center, user forums, blogs, etc.)

Seven Steps to Creating a Data Driven Decision Making Culture

  1. Go for the bottom-line (outcomes): What motives the people around you?
  2. Reporting is not Analysis
  3. Depersonalize decision making
  4. Proactive insights rather than reactive: Deliver insights before someone asks
  5. Empower your analysts: They are not reporting monkeys. Give them strategic objectives
  6. Solve for the Trinity: What, Why, How much?
  7. Create a understandable, repeatable framework for making decisions
  8. Business/Strategy should own web analytics

Five “Ecosystem” Challenges for Web Analytics Practitioners

  1. Lack of relevant talent / skills: No real format training; too much experience requested (5+ years) although the field is moving fast
  2. Entrenched mindsets: Decision makers still thinking in the old way
  3. The web is no longer a monologue
  4. It’s not about you, it’s about your customers: Less clickstream, more experimentation, usability, integration of multiple sources
  5. Web analytics is the first step

Web Analysis: In-house or Out-sourced or Something Else?

  • In the long run: in-house team
    • strategic implementation of WA can’t exist in a silo
    • Qualitative analysis is also needed
    • Tribal knowledge helps the decision making
  • But, it depends on the stage:
    • Stage 1 – No WA -> Implement and show promise from data
    • Stage 2 – Too much data -> Hire WA, customize dashboards, tag everything
    • Stage 3 – WA rocks; Asking the why -> Start testing, collect qualitative data, expand your team
    • Stage 4 – Trinity implemented -> new data, more people, a self-sustaining process

Five Rules for High Impact Web Analytics Dashboards

  1. Benchmark & Segment: Provide context
  2. Isolate your critical few metrics: less than 10 metrics
  3. Include insights
  4. Don’t create more than one single page
  5. Constantly adapt your dashboard to changes in the environment

I Got No Ecommerce. How Do I Measure Success?

  • Visitor Loyalty: How often does one person visit my website?
  • Recency: How long has it been since a visitor last visited your website?
  • Lengths of Visit: How long does each session last?
  • Depth of Visit: How many pages did they visit?

Convert Data Skeptics: Document, Educate & Pick Your Poison

  • Understand how your tools actually measure
  • Document your findings
  • Present everybody touching the data your findings
  • Report high level trends between the tools
  • Pick the best tool for each metric

Is Conversion Rate Enough? It’s A Good Start, Now Do More!

  • There is more stuff than just conversions
  • Esp. for non-ecommerce businesses, ask:
    • Have you found what you were looking for?
    • Will you go to our store?
    • Will your recommend our website?

History Is Overrated. (Atleast For Us, Atleast For Now.)

  • Value of web analytics data decays in time
  • Your visitors change too much: browsers, cookie deletion, etc.
  • Your computations change too much: new computations, maybe other tagging, etc.
  • Your systems change too much: Other hosts, new technology, etc.
  • Your website changes too much
  • Your people change too much
  • => no real tie to legacy tools and data
  • => keep some major benchmarks for comparison

“Engagement” Is Not A Metric, It’s An Excuse

  • Engagement is unique – therefore say what you actually measure instead of saying that you measure engagement.
    1. What’s the purpose of the website?
    2. How do you measure success?
    3. Define your metrics
    4. Call them what they are
  • Ideas:
    • Question: Are you engaged with us?
    • Question: Likelihood to recommend website
    • Use primary market research
    • Customer retention
    • RF of customers

In Web Analytics Context Is King Baby! Go Get Your Own

  • Never report data in aggregate, or by itself. Always always always test to see if you are including context!
  • Compare trends over different time periods
  • Compare key metrics and segments against site average
  • Report multiple metrics
  • Benchmark
  • Tap into the tribal knowledge

The “Action Dashboard” (An Alternative To Crappy Dashboards)

  • Why do dashboards suck?
    • Don’t provide a interpretation
    • Readers don’t trust the providers of dashboards
    • Not enough company context in the interpretation
    • Providers don’t have enough experience
  • How to make a good one?
    • Report only 3-5 most critical metrics
    • Show a trend in a graphic for a metric (segmented)
    • Give key trends and insights
    • Recommend actions
    • What are the impacts of the company

Multichannel Analytics: Tracking Offline Conversions. 7 Best Practices, Bonus Tips

  • Track your online store locator, directions, etc.
  • Use unique 800 numbers
  • Use unique coupons / promotions
  • Connect online and offline – e.g. club cards, delivery, etc.
  • Ask your customers (survey)
  • Conduct controlled experiments
  • Do primary research

Multichannel Analytics – Tracking Online Impact Of Offline Campaigns

  • Use vanity urls: Permanent redirects help you to differentiate between offline and online referals
  • Use unique coupons / offer codes
  • Survey, survey, survey
  • Correlate traffic patterns with offline ad patterns
  • Experiment

Slay The Analytics Data Quality Dragon & Win Your HiPPO’s Love!

  1. Change your boss
  2. Compare web data with their favorite source
  3. Put the data quality problem aside and give them actionable insights
  4. You get trends rather fast even without a complete WA implementation
  5. Start with the outcomes
  6. One WA is probably enough
  7. Realize if the data quality is good enough
  8. If you don’t have enough traffic, care about more traffic first
  9. There are inaccurate benchmarks and illegal customer behavior – don’t care too much about it
  10. Fail fast, i.e. test

Barriers To An Effective Web Measurement Strategy [+ Solutions!]

  • Note: Tools aren’t the real problem, still lots of people talk about them
  1. Lack of resources (45%): Start for free and ask your right for a budget
  2. Lack of strategy (31%): Change your job. If your a VP maybe you can help create one
  3. Siloed organization (29%): Start small and offer value
  4. Lack of understanding (25%): Do and show
  5. Too much data (18%): Limit yourself on the critical few metrics.
  6. Lack of senior mgm buy-in (18%): see previous summaries
  7. Difficulty reconciling data (17%): Whatever.
  8. IT blockages (17%): Show lost revenue by delay
  9. Lack of trust in analytics (16%): previous summaries
  10. Finding staff (12%): Don’t be to narrow minded
  11. Poor technology (9%): Concerns mostly only brand new technologies / platforms

Who Owns Web Analytics? A Framework For Critical Thinking

  • The biggest problem is most often the organization structure.
  • How long has the company been doing WA? -> New: find accepting division to embarrass the seniors – Older: see next point
  • Analytical maturity? New: find accepting people – Older: identify power centers
  • Who owns the power to make changes to the site?
  • Which model: Centralized? Decentralized? Something else? Agile team with satellites
  • Which division / department is best for WA? Marketing if they have the power -> Ideal situation: own department

10 Fundamental Web Analytics Truths: Embrace ‘Em & Win Big

  1. If you have more than one clickstream tool, you are going to fail: Implementing, understanding and communicating one tool is hard enough
  2. Expensive tools won’t give you insights: Real problems are lack of skills, terrible orga, no structure or no courage.
  3. It is faster to fail and learn then wait for an “industry case study” or find relevancy in a “industry leader white paper”
  4. You are never smart enough not to have a Practitioner Consultant on your side (constantly help you kick it up a notch)
  5. Your job is to create happy customers and a healthier bottom-line
    • Go to your own website
    • Read your own email campaigns
    • Buy something on your own website
    • Return a product on your own website
    • Do the same stuff on competitor’s websites
    • Do usability studies
    • Be a customer and ask yourself: What will create happier customers tomorrow?
  6. If you don’t kill 25% of your metrics each year, you are doing something wrong
  7. A majority of web analytics data warehousing efforts fail. Miserably: Too much irrelevant data, mostly anonymous, full of holes, BI are bad at answering WA questions and DWs are too slow
  8. There is no magic bullet for multi-channel analytics: see previous summary
  9. Experiment, or die.
  10. The single most effective strategy to win over “stubborn single-minded” HiPPO’s is to embarrass them.

Measuring Incrementally: Controlled Experiments to the Rescue!

  • Test everything
  • You may need additional personnel for conducting and analyzing the experiment
  • It gives you excellent insights in different methods
  • Act on your results as fast as you can

Reading Kaushik (Part 1): Digital Marketing

I read Occam’s Razor for quite a while now and I really like Avinash’s style and insights. I thought it would be nice to reread most of his stuff and as a nice extra, I will post my notes on here.

I oriented each section by the section defined in his overview of all articles.


The 10 / 90 Rule for Magnificent Web Analytics Success

  • There is lots of data but no insights
  • Rule: 10% in tools and 90% people/analysts
    • may seem over the top but
    • med-large websites are complex
    • reports aren’t meaningful by default
    • tools have to be understood
    • there is more than clickstream to analytics
  • If you don’t follow the 10 / 90 Rule
    • Get GA account
    • Track parallel to expensive solution
    • Find a metrics multiplier, so you can compare GA to old data
    • Cancel your contract and hire an smart analyst which will probably deliver more insights for less money

Trinity: A Mindset & Strategic Approach

  • The goal is to generate actionable insights
  • Components:
    • Behavior analysis: clickstream data analysis
    • Outcomes analysis: Revenue, conversion rates, Why does your website exist?
    • Experience: Customer satisfaction, testing, usability, voice of customer
  • Helps you understand what customer experience on your site, so that you can help influence their behavior

The Promise & Challenge of Behavior Targeting (& Two Prerequisites)

  • We have so much behavior data but you get the same content regardless whether you are here to buy or get support
  • There are BT systems but you have still think about the input
  • You have to first understand your customers good enough to create suitable content
  • Test content ideas first to learn what works and as evidence for HiPPOs

Six Rules For Creating A Data Driven Boss!

  • Paradox: The bigger the organization the less likely it is data driven in spite of spending lots of money on tools
  • It is possible to achieve this but you have to actually want to do and fight for it
  • 1. Get over yourself: Learn how to communicate with your boss and try to solve his problems
  • 2. Embrace incompleteness: Data is messy, web data is really messy but still better than completely faith based initiatives.
  • 3. Give 10% extra: Don’t just report data, look at it. Give him insights he didn’t asked for. Make recommendations and explain what’s broken.
  • 4. Become a marketer: Great analysts are customer people. Marketer as internal customer (like account plannner)
  • 5. Don’t business in the service of data: Data should provide insights not just more data. Ask: how many decision have been made based on data that have added value to the revenue?
  • 6. Adapt a Web Analytics 2.0 mindset:

Lack Management Support or Buy-in? Embarrass Them!

  • HiPPOs may be don’t listen to you but they better listen to customers & competitors
  • 1. Start testing
  • 2. Capture Voice of Customer: Surveys, Usability tests, etc.: Let the customer do the talk
  • 3. Benchmark against the competition, e.g. use Fireflick
  • 4. Use Competitive Intelligence
  • 5. Start with a small website
  • 6. Ask outsiders for help

How To Excite People About Web Analytics: Five Tips.

  • 1. Give them answers
  • 2. Talk in outcomes / measure impact
  • 3. Find people with low hanging fruit and make them a hero
  • 4. Use customers & competitors
  • 5. Make Web Analytics fun: Hold contests, hold internal conferences, hold office hours

Redefining Innovation: Incremental, w/ Side Effects & Transformational

  • 1. Incremental innovation, e.g. Kaizen
  • 2. Incremental innovation with side effect, e.g. iPod or Adsense
  • 3. Transformational innovation, e.g. invention of the wheel
  • Web analytics can’t probably create 3
  • Clickstream alone is also not enough for 1.
  • generally the more the better (Web analytics 2.0)

Six Tips For Improving High Bounce Rate / Low Conversion Web Pages

  • Purpose gap between customer intent and page
  • 1. Learn about traffic sources / keywords(!)
  • 2. Do you push your customers against their intent? Identify jobs of each page and focus on your call to actions.
  • 3. Ask your customer what they are looking for
  • 4. Get insights from site overlays
  • 5. Testing!
  • 6. Get first impressions from people, e.g. fivesecondtest

Online Marketing Still A Faith Based Initiative. Why? What’s The Fix?

  • Faith based initiatives like TV, magazines, etc.
  • Online marketing gives us useable data
  • and allows us to test easily
  • The web is quite old yet it is not in the blood of executives
  • Old mental: shout marketing, instead of new inbound marketing
  • Lousy standards for accountability
  • Let the customers speak
  • Benchmark against competition

Win With Web Metrics: Ensure A Clear Line Of Sight To Net Income!

  • Focus on the bottom line, i.e. profits
  • 1. Identify your Macro Conversion
  • 2. Report revenue
  • 3. Identify your Micro Conversions
  • 4. Compute the economic value
  • Net income = Unit Margins * Unit Volumes
    • Unit Margins = Price – Cost
    • Unit Volumes = Market Share * Market Size
  • Because Net Income is the goal, you have to measure Price, Cost, Market Share or Market Size
  • Which metrics help doing that? And if not, why do you track/report this metric?
  • They also depend on the strategies or more general goals of the organization
  • — let your “boss” decide what matters most to him/organization
  • identify clear metrics / KPIs for each used strategy
  • use the web analytics measurement framework as a reporting foundation (more to this later)
  • find actionable insights with segmented analysis

Digital Marketing and Measurement Model

  • Marketing with measuring helps you to identify success and failure
  • Digital Marketing & Measurment Model
    1. Set business objectives (should be DUMB)
      • Doable
      • Understandable
      • Manageable
      • Beneficial
    2. Identify goals for each objective
    3. Get KPIs for each goal
    4. Set targets for each KPI
    5. Identify segments of people / outcomes / behavior to understand why things succeeded or failed
  • What scope has the model to cover?
    1. Acquisition: How do people come on your site? Why? How should it be?
    2. Behavior: What should people do on your site? What are the actions they should take? How do you influence their behavior?
    3. Outcomes: What are the goals? (see previous summary)

11 Digital Marketing “Crimes Against Humanity”

  1. Not spending 15% of your marketing budget on new stuff
  2. Not having a fast, functional, mobile-friendly website
  3. Use of Flash
  4. Campaigns that lead to nowhere
  5. Not having a vibrant, engaging blog
  6. “Shouting” on Twitter / Facebook
  7. Buying links is your SEO strategy
  8. Not following the 10/90 rule
  9. Not using the Web Analytics Measurement Model (previous summary)
  10. Using lame metrics: Impressions, Page Views, etc.
  11. Not centering your digital existence on Economic Value

Video: Successful Web Analytics Approaches by Avinash Kaushik

Again a great video, this time about Web Analytics by Avinash Kaushik. I just love his no-BS style.

  • Ask the metric: So what? Three times, if it don’t give an action it’s useless
  • Data should drive action
  • Give people the information they need – don’t send them everything => no death by data
  • Home pages of websites, are no longer the home page you want
    • Where do people come from?
    • What are they looking for?
  • Context matters: previous months, years, etc.
  • Relative numbers more important than absolute numbers
  • Compare different metrics, e.g. conversion rate and page views
  • non e-commerce sites:
    • averages hide truth effectively
    • How often do they visit?
    • How recent did they visit?
    • Depths of visit
    • => Understand the value: Loyalty
  • Segment people
  • Survey people: What do they think about the content?
  • Bounce rate: Came and left
    • Segment by source, entry-page, landing pages, etc.

Rules for Revolutionaries

  • 10/90 Rule: $10 Tools, $90 People: Understand Data & Business, Able to analyze => to extract value
  • Reporting is not analysis: Reporting -> provide data; Analysis -> prove insights
  • Data Quality can be low, but is still better than other data
  • Faith-based initiative: e.g. magazine ad without tracking
  • Make decisions, don’t argue about the quality of the data
  • Over time understand why quality is different -> confidence will get better


  • Decision making is a journey, not a destination
  • => Put some level of process in place, mostly for tasks, e.g. what happens to implement a test, etc.
  • if HiPPo (highest paid person’s opinion) makes the most decisions
  • => make experiments
  • Learn to be wrong, quickly
  • => You probably don’t know what your customers want
  • => Experimentation