Code of Ethics and Beyond
Ethical Analysts . . . Then What?
- Nothing to Fear
- Accuracy Paradox
- Predicting the Future is Hard
- Predicting by Guessing
- Decision Optimization
- Dynamic Challenges
- Dynamic Web Analysts
- Challenges for the WAA
The release, and public support for, the Code of Ethics gives me great encouragement that the community is serious about policing our own. I have, of course, signed up to support the recently finalized Web Analytics Association Code of Ethics.
Self-examination never being a bad thing, perhaps the issue of the Code of Ethics will lead to a larger discussion of:
- Who is/should be our membership?
- What we can all do to prepare ourselves for the future?
- What exactly is the Web Analytics Association?
Nothing to Fear
As Eric T. Peterson points out on his Web Analytics Demystified blog post of December 2nd, 2010, there is little to fear from losing granularity in tracking. In truth, better models may be built as a result of pulling back the focus from individuals.
Covered previously, as we try to zoom in closer and closer to the individual level visitor, i.e. increased precision, we lose confidence in our measurements and subsequent predictions, i.e. accuracy.
Predicting the Future is Hard
I interviewed John Mount and Nina Zumel of Win-Vector for an upcoming post on statistics as applied to web analytics data, and during our discussion Nina had one of the best quotes about predictive analytics:
Predictive modeling as it is currently being done is largely a correlation of factors rather than truly a prediction of the future. Predicting the future is hard.
Why make it harder by trading away accuracy for precision?
Predicting by Guessing
Perhaps a better model would be to make predictions based on guessing, such as a human would make. What may be most relevant is the portability of the previous experience into the next experience, whether correct or not.
The discussion of the human brain used the analogy of entering a new airport, the human brain would, perhaps erroneously, assume that the new airport would have a similar layout to the airport with which it was most familiar. How to apply HTM, or any other algorithm, to predict human behavior is a topic for another time.
The attractiveness of mPath is that it utilizes, in real time web analytics data along with external econometric data to optimize the user experience.
What is more valuable to your organization, incorporating in nearly real time all the publicly available external data such as:
- Local Events
- Census Data
Or exactly how many times visitor XYZ1234567890 came to your site, which pages they looked at, and other highly granular data?
Probably makes for nice charts, but charts don’t buy things.
Consider that the internet, with all those web service Application Programming Interfaces for just about every relevant site, is essentially a distributed computing platform.
The ability to source, clean and even prototype connectivity with disparate external data appears to be overlooked at many companies. In contrast, companies such as LinkedIn, Google, and Facebook seem to have the concept of rapid experimental prototyping built into their DNA.
That old manager of mine loved to say that people could either be busy with work, or be busy at work.
Dynamic Web Analysts
This could, as my predictions guaranteed to be correct state, lead to dynamism in the role of the web analyst.
Instead of valuing experience with flexible data, which are increasingly important, many companies currently place an extreme valuation on experience with inflexible data presented in an inflexible format: the tool which is deployed at their business.
Challenges for the WAA
Should my predictions guaranteed to be correct in fact be correct, and organizations increasingly desire people with skill sets outside the traditional tools, the challenge is:
- Where ultimately does the WAA and #measure community sit?
- Are we solely analysts?
- Do we ever build stuff?
- What is our role?
I thoroughly encourage more people to sign up for the WAA Code of Ethics, and for employers who consider looking at the Certified Web Analysts list to also look at the signatories to the Code of Ethics.
My hope for a discussion of the WAA at large is based on the perspective that we, all too often, are outsourcing what we could be doing quite well. The more difficult the challenge the more I enjoy it, so I am not a person to shuffle predictive modeling over to that group, statistical analysis over to that group and something else to a third group.
The concern long term would be that once all the work is distributed, what is left for us to do?
Drop me an email if you have an challenging challenges or private comments.