Accidental Factors

The Kirkpatrick Learning model helps with evaluating the effectiveness of training. It proposes evaluating at four levels:

  1. Reaction: What did learners think of the training
  2. Learning: Did learners improve their understanding or skill from before the training
  3. Behavior: Are learners applying the new/improved skill contextually
  4. Results: Have the organizational outcomes changed

Training design goes in the other direction, starting with the result, identifying the behaviors that need to change, and the learning needed to drive those behaviors.

My impression is that step 3 is the really tricky one. A lot of training interventions have a clear idea of the results they want, but don't do enough to consider what behavior change would be needed, and whether the environment actually supports that behavior. I suspect for most effective training the learning needs to be accompanied by environmental changes that incentivize the right behavior.

I’ve recently read two great pieces of writing that start with a well known exceptional thing, and largely show that it is predictable, given a proper understanding of the circumstances. The first is Human Advantage by neuroscientist Suzana Herculano-Houzel. The book tackles the notion that the human brain is exceptional: exceptionally large, exceptionally high in neurons, exceptionally high in energy cost. It turns out that the human brain is all of these things - large, costly and complex - but not in an exceptional way: it is pretty much as expected for primates of our size, and the really interesting evolutionary deviation was that of primates from other mammals.

Similarly, Random Critical Analysis looks at healthcare spending in the US, which seems high, but turns out not to be as a percentage of household disposable income. It is much more in line with other service spending, and the experience in other countries even under different organizations of healthcare spending.

Both a great reads, and I found the rhyme between the two particularly interesting - something that seemed exceptional and special case was in fact part of a general case pattern. That changes how you look at it, and the kind of interventions you might try to improve a deficiency. I suspect this pattern is a common one - if we think some large scale result is somehow unique to the group being looked at, we are likely to be wrong, and should at the very least investigate whether, in fact, this is just "another one of those".

There are quite a few organizational maturity levels, but I've generally found them either too specific to a certain domain, or too speculative to apply properly. I recently read a Miller Heiman Group publication that included the one they have developed for sales enablement processes. It seemed to strike the right mix between practicality and useful, and I've found it helpful in getting an orientation on a few process-related questions.

The four levels (at least in how I interpret them) are:

Random: While the process is happening, its generally due to the actions of individuals or small teams, and there is no formal coordination across the organization in this domain.

Organized: There is some cross-team coordination in at least one part of process, there is a general recognition of the need for such a process, and the core audience the process serves is well defined.

Scalable: The process is organized across all parts, there is a governance method determined, and a broader set of audiences/stakeholders is considered.

Adaptive: The process has visibility at high levels of the organization, and there is a second level of direction setting to reflect and redirect on it.

Its pretty easy to map these to questions to ask to ask or things to look for. For example, if I were looking at a hiring process at a company I might consider:

  1. Random - is hiring handled by individual managers? Are standards and procedures ad-hoc, or shared manager-to-manager?
  2. Organized - is their some formal involvement by an HR function or similar? Are they involved consistently in at least one part (sourcing, interviewing, onboarding)?
  3. Scalable - are there consistent practices across the company, and across the whole hiring pipeline? Is there a group which is the keeper of the truth for those standards?
  4. Adaptive - is company-wide hiring a consideration at the executive level? is there a process in place to review how well new hires do and suggest improvements?

Assessing the maturity level gives you a guide on where to apply pressure, and what kind of tactics will improve the situation and help move to the next rung of the ladder.

Taylor Lorenz at the Atlantic has been doing some really interesting reporting on Instagram influencers, getting in to a lot more depth than the normal breathy coverage of any new Youth Marketing Trend.

Influencers are interesting to me as they are a pretty good example of one of the new styles of work that are emerging. The basic unit of value - endorsing a product to your fans - is far from new, but almost everything else is in new territory. Unlike with a celebrity endorsement, many of the influencers monetise their taste directly, rather than as a side business to an entertainment or sports career. Influencing has a low barrier to entry (4 digit follower count), but getting big deals requires some real audience building and investment, and some real risk of the landscape changing underneath you.

Brands identify and contract directly with individuals, which open up all sorts of interesting incentives. Up and coming artists are placing peer to peer promotions and it seems likely upcoming influencers them selves are buying cross-promotion the same way. Similarly, influencers are posting up fake sponsored content to build credibility - both a valid response to social signalling being used to evaluate social signalling. While this is likely frustrating to some buyers - particularly traditional brands - they've certainly set up a situation where its important to differentiate:

"Teenagers are more affordable to work with because of their follower count and age," says Christy Oh, an 18-year-old who handles marketing for Doux Lashes, which sells fake eyelashes. "They’re not doing Insta as a full-time thing; they’re just trying to make extra money, so it’s not super expensive to partner with them."

Though it is equally true, I cannot imagine an Uber rep describing their drivers this way - but for an 18 year old in charge of social marketing, its not strange at all. The fake sponcon is also amusing note on regulating new sectors - while disclosing promotional content is now required on social media, there is no standard on presenting something as promotional content when it isn't.

Took me a while to get round to listening to the Sinica interview with Kevin Rudd on Xi Jinping's worldview, but it's excellent and worth a listen even if you don't have a specific interest in China.

One of the points Rudd makes is that we shouldn't act as if the the politics in the "mysterious middle kingdom" work fundamentally differently than in countries we are more familiar with. It's important to understand the pressures on a Marxist/Leninist government, and the thought process behind that, but there is much in the politics within the party and the government that would feel familiar to any politician.

Churn is, generally, seen as a bad thing within an organization - teams wants stability and predictability. These help make the organization more legible, and they are part of generating the feelings of psychological safety and mutual trust that are needed for groups to be effective. Churn also produces work which does not feel productive: recruiting, staffing, training and so on.

My anecdote based estimate though is that most white-collar workplaces would be better with higher churn, a more aggressive balancing of people to projects, of skills to needs. More churn sounds pretty awful in part because the mechanisms to manage it are significantly underdeveloped - churn is hard because churn is something to be avoided, not embraced.

It is often easier to do nothing - leave a high performing team in place, even if the teams member would be more effective elsewhere.

To use a sports metaphor, a football team may field a pretty consistent squad match to match, adapting the specific positional role each player fills to the strengths and weaknesses of their opponents. Season to season, the squad will evolve as the nature of the competition changes and players developer and mature. Decade to decade, the positions themselves evolve as the game develops (due to fitness standards, rule changes and so on).

At each timescale the things that aren’t changing benefit from being drilled and perfected, but if that results in an inability to react to the longer timeframe they switch from being a benefit to being a drag on performance.

In general organizations, getting "the right people on the bus" means some degree of churn. Getting it right doesn’t end the churn though, it just pushes the question up to "is this the right bus".

At the end of his Patterns of Conflict briefing, Boyd talks about how grand strategy is centered around maintaining and improving our adaptability to change.

One of the key components of that is being able to effectively:

  1. Generate new ideas and possibilities
  2. Rapidly adapt to them

This got me thinking a little on diversity. Diversity within a group requires rewarding and incentivizing different ways of being - what Boyd calls promoting the unconventional. To generate a range of possibilities we need a range of backgrounds, a range of thinking.

Rapid adoption of new possibilities when they work requires harmony and alignment - tight, effective groups that can "turn on a dime".

There is a tension there, as one of the easiest ways to get that high cohesion, high alignment harmony is to suppress diversity, and optimize for fit. This generally means a group with more similar backgrounds and cultural experience. This allows for a higher level of implicit communication, a key part of rapidity. Similarly, alignment is easier without diversity of thought - ideas are more easily assimilated when they come from similar starting points.

A group that relies too much on background for harmony will be able to move rapidly, but will be unable to generate a really wide range of possibilities - they will be effective until they are suddenly and catastrophically unsuited for the world as it has become.

Building a group with both diversity and harmony is significantly harder. The leader is required to have quite a high tolerance for discomfort and ambiguity - their team should be producing ideas which they are not going to like. The team will also have to find their own basis for the implicit communication and high levels of trust.

This is easier the smaller the group and the less complex the structure of the organization. Boyd describes the idea as multiple cooperative centers of gravity.

The more a group can be introduced to a range of conditions, the more they are forced to adapt and explore new ideas. That means choosing adaptability over medium term optimisation, on the basis it results in better long term outcomes. In the medium term may well look like a bad idea, which is a risk the group has to be comfortable taking.

Interesting post and comment thread on Arnold Kling's blog about construction productivity being higher in the Great Depression.

It reminds me of a note in Patty McCord's Powerful about Netflix going through a round of lay-offs. Afterwards, they found that their output dropped less than expected - they were more relatively productive than they were before the layoffs because the people they had retained were the strongest fit for what the company was doing.

There is a huge range between the most and least productive workers, but the causes seem to often be industry, geography or firm specific. I suspect there is some systemic weakness in work-to-worker matching (a large scale coordination problem?), but I wonder whether that problem is even worse as companies scale. Matching for a 100 person company is likely easier than a 100,000 person one, in part because a larger company allows for more specialised roles, and in part because effectiveness becomes more tied to firm-specifoc knowledge.

The New Yorker has a piece on Jeff Dean and Sanjay Ghemawat, who are directly responsible for a many of Google's technical innovations. They have a very close collaboration and friendship that has spanned two different companies and a number of projects.

It's a great article about an extroardinarily deep and productive relationship. Towards the end the article whistfuly notes that they only spend a day a week coding together. The remarkable thing for me is that they spend a day together at all - Jeff leads a significant number of people, and Sanjay is a sort of uber-tech lead. It's rare to find senior folks able to regularly get in to the details in that way, even one day a week.

Financial portfolio theory has the idea that owning too much of a particular stock is bad as it exposes you to greater company-specific risk, which unlike systemic risk is not generally compensated for by higher returns (assuming investors are demanding proper risk premiums etc.) This leaves room for a particular kind of human error: you have to sell things which are doing well, as they come to represent a greater proportion of your portfolio that is optimal. This is emotionally difficult.

Societally, we have the same kind of bias. One of the influences on the 2008 Great Recession was a view that home ownership has been good for most home owners, therefore we should incentivize home ownership. Clearly home ownership isn't the only route to good living situations - Germany is fine with much lower rates - and perhaps the conclusion should have been "home owners are doing fine, let's incenticize rentals" or some other option. It's much easier to reinforce something that is going well, even if in that action you are taking on significantly more risk for the future.

On a personal level, this feels true about careers too. The most direct way to greater returns for people with white collar jobs in large companies is, generally, to do well at their job and get promotions or other opportunities. This encourages building up localized skill sets, as differentiated from more market-available skills. This approach brings with it a risk though - you, as an individual, are exposed heavily to your company’s risk profile. Someone who moves around a little more may take on more earnings risk (they might not perform as well at the new firm), but increases their resilience in case of specific company failure.

I remember hearing a story from a friend about an ex-Nokia software developer coming to them after the old Nokia software teams imploded, proud and certain that their accomplishments (which were notable) at Nokia would land them any job they picked. They quickly discovering that the needs the market had were quite different than the skills they could easily demonstrate, leading to a long and somewhat painful sounding job search.