The Soil Cultivation Metaphor

Posted on by Stuart Murray
hand-1-by-david-pacey
Hand 1.  Attributed to David Pacey.

One of the greatest pleasures of home ownership is the opportunity to work in the garden.  Gardening is fulfilling for several reasons.  The accomplishment is satisfying and tangible, unlike a lot of office work.  Gardening is great physical exercise, involving a range of low-impact and core-intensive body movements.  Gardeners get time outdoors, bolstering vitamin D intake and exposure to fresh air.  By handling soil, our body develops a resistance to the germs and bacteria, or so rumor has it.  The work is solitary and meditative, improving mindfulness.  The home-grown and fresh-picked produce tastes better and is higher in nutrients.  Gardening is a kind of cure-all for wellbeing.

Soil cultivation is extremely similar to the practice of cleaning a data set.

My first round of soil cultivation was a patch of land that previously had a garden shed sitting on top of it.  There was no organic matter in the soil at all; just a dense patch of dust.  The soil required a major intervention to become useful.  This is also true about a new data set.  It’s great to have new data, but I just know it’s going to require a lot of attention before I can use it properly.  The data will lack clear labels, there will be columns or fields that are useless in some way, and some data points need to be converted.  Sometimes it simply arrives in the wrong format, such as on paper or ascii, or built around a different software environment.  The certainty that I must put work into it in order to get something back, turns this into “real” work.

With data I typically find that some fields are all wrong, and I need to track down or create “lookup” tables that convert the raw data into something that I know is accurate.  I don’t like throwing out data.  I prefer to just keep the dirty data on the left hand side of a spreadsheet, and to the right of a thick, vertical line, create a modified column or field.  The raw data is black, and the modified data is has color, so I know what I’m working with.  I also give the modified data more explicit labels, in succinct but plain English.  Many dubious data fields are suddenly rendered “accurate” by a good label.

As with my fully-remediated soil, my fully-amended data set means that I am ready-to-roll.  I can work with a converted and color-coded batch of cultivated data, culled of garbage and meaningless fields, and turned into something useful.

I know that some people think of a garden as a place where plants grow.  And some people think of data as something that is capable of producing analytic insights.  In both cases, there is something deeply human about taking a mess – soil or data – and turning it into something more.  We advance civilization one step at a time, one cubic foot at a time, one data point at a time.  Sometimes we just need to break a sweat, get some sun, work our bodies, and build immunity.

Millennial Turnover Similar to Prior Generations

Posted on by Stuart Murray
hipster-attributed-to-rodger-evans
Hipster.  Attributed to Rodger Evans.

It is my pleasure to draw your attention to a great paper produced by three students at the University of British Columbia.  Grace Hsu, Geoff Roeder, and Andrew Lee produced a paper for their Statistics 450 course with Dr. Gabriela Cohen Freue which was put in for a student research contest.  The paper, Analysis of Factors Affecting Resignations of University Employees won an honourable mention for the contest.

The paper identifies that “Millennials do not exhibit a practically significant different length of employment compared to other generational groups.”  That is, that although those born after 1975 have a high quit rate right now, they are passing through a high-turnover age group.  Prior generations that passed through the 25-34 year old age group in years past, themselves had high quit rates.

Getting more to the point… “This finding disrupts stereotyped representations of generational factors in the workforce and suggests that younger employees resigning sooner can be better explained as a feature of their age rather than their generational group.” My guess is that age 25-34 is when people figure out their career, partners, and housing, with some things changing a few times before getting stable.

Working with twenty years of data covering 7000 staff who quit, their data model chose “years of service” as the variable that would be explained by other data points.  If we could predict the number of years a new hire would stay, this might be something an employer could improve.  That is, assuming it was not illegal to pre-judge.  Thankfully, their findings suggest we should not pre-judge.

Years of service prior to quitting averaged 1.2 to 1.9 years for 25-to-34 year olds, and 4.3 to 5.5 years amongst 35-to-44 year olds.  There were small differences between generations, but not in a manner that strengthened a stereotype.  For example Generation X quit more quickly when they were younger, but stuck around for longer once they were 35-44.  Baby Boomers were not always big on job loyalty, being the quickest to quit in the 35-to-44 age bracket.

One more thing… men and women do not have a big difference in their length of service.  When sizing-up job candidates for staying power, it is not just unfair and illegal to favour men; it is wrong on the facts.  Keep that in your back pocket next time you help with hiring.

The Mathematics of Love

Posted on by Stuart Murray
coffee-hearts-photo-attributed-to-peter-burka
Photo by Peter Burka.

In this fascinating TED talk by Hannah Fry, the speaker describes three mathematical algorithms that explain love.  One of her findings was that the number of romantic advances someone receives is improved if there are polarized views of whether they are attractive.  That is, you will approach someone if you think that many people other than you will think they are unattractive.  How would we apply this to human resources?  I would point out that the Oakland Athletics baseball team was successful at snapping-up rookies that they knew were great and everyone else had passed over.  Using better metrics, the team found high-performers who seemed wrong, but were factually very good.  In many cases the Oakland A’s got first dibs on diamonds-in-the-rough.  Can you apply this insight to your own workplace?

The Math Behind Basketball

Posted on by Stuart Murray
Kobe Bryant, Photo by Alexandra Walt (public domain).jpg
Kobe Bryant.  Photo by Alexandra Walt.

In this TED talk by Rajiv Maheswaran, the speaker describes the translation of basketball moves into a series of moving dots, looking at games played by professional basketball players.  Using machine learning they were able to identify the difference between the baseline probability that a shot would be successful, and an individual player’s personal odds that they would make a shot. This distinction allows us to tell the difference between those who have more opportunities (or create more opportunities), as opposed to those who perform well based on what is in front of them.

The speaker invites us to consider other applications for the analysis of moving dots.  In my opinion, this means we can think of activity levels, workplace layout, and injury statistics alongside other workplace wellbeing indicators.  It might be that physical movements have a substantial impact on non-physical workplace performance.

The Mountain of Gold

Posted on by Stuart Murray
camp-millar-gold-mine-hut-by-smudege-9000
Camp Millar Gold Mine Hut, photo by Smudge 900.

Many human resource people are bad at math. But, even if you regard yourself as one of these people, a great event is ahead you.  It’s helpful to place HR in the context of other strategic business pillars – those major units inside large firms that have their own professions or their own Vice-Presidents, such as accounting, production, and marketing.  In the recent history of business, computers made it easy for each of the pillars to apply math to their data.

Finance and accounting got good at this in the 1970s and 80s.  Engineers changed production lines using data throughout the 90s.  Marketing and customer service saw big data happen from the millennium onward.  These fields achieved great business success applying new computer tools to fresh data.  In the meantime, human resources twiddled its collective thumbs for fear of the data itself.

I like to think of data as deposits of gold in a mine.   And, as data miners, it is up to us to use new tools to bring this treasure to the surface. Although many mountains of data have been already mined, and have produced lots of gold, data miners have started hitting a lot of rock. The yield just isn’t what it used to be. Beginning around 2010, someone found a large, new, and un-touched mountain of gold.  To many this mountain appeared to have nothing inside it.  That is incorrect, though. This mountain is filled with gold, but you just need to know how to drill for it.

This gold is human resources metrics.  Decades ago the finance function insisted that we automate payroll, create a line-of-sight into spending, spend the correct amount of money, and minimize risk by obeying all laws.  A fancy system of rules was created to regulate salaries including collective agreements, pay policies, rules-based pension plans, and human rights laws about fair pay.

Deep inside this alleged payroll data exists a large and accurate dataset about age, sex, length of service, union, rate of pay, and job code.  And the other pillars can’t get into this data because they don’t know enough about the people.  Some classic examples would be that there may appear to be a large number of women quitting, or a modest number of accident claims, or a growing bureaucracy.  By lining up the apples-to-apples comparisons and creating better ratios, things may look different.  You might find that women’s rate of turnover is average, that one unit’s accident rate is high, or that the size of the bureaucracy is modest.  Then, people can make better choices about where to devote their efforts.

By starting small, building basic skills, cleaning the data, and creating a steady growth of analysis tables, there are ways to make gold.  Strangely, you can only do this the human resources way, full of stories and feelings and a sense of fairness and the motivation to strive.  The data just takes you there; because this is still human resources.  So start digging.  There are treasures to be found.

Rainer Strack TED Talk

Posted on by Stuart Murray

 

stack-of-rainier-no-rights-claimed
A Stack of Rainier, no relation.

This TED Talk is one of the most compelling explanations of why workforce planning is so important to the business success of major employers.  He takes it global and forecasts into 2030.  With jokes.