My notes and other stuff


Paper: Don't Let Knowledge Walk Away

So let's try the long form thing here, for stuff that doesn't fit my blog but is too big for twitter-style stuff. I like to review papers for all sorts of stuff, and this one is quite topical.

Given the current Twitter & Musk news, I jokingly asked my Resilience Engineering network if "anyone had papers on the risks of massively firing/losing employees with institutional knowledge?!"

Well it turns out there are, and one of them is really good! Don't let Knowledge Walk Away: Knowledge Retention During Employee Downsizing from Achim Schmitt, Stefano Borzillo, and Gilbert Probst, is a fascinating read across concepts of organizational knowledge and a literature review of various studies documenting the impact of willingly (and strategically) deciding to massively cut a bunch of your people loose.

The paper starts with an overview of various sorts of downsizing (letting people go in a planned manner, excluding retirement and resignations) and their consequences. They first define downsizing as sometimes necessary and useful, namely in industries showing a decline (eg. newspapers print business) or when technological change causes obsolescence. It's also characterized as a strategic move usually aiming at higher efficiency and productivity, and possibly an adaptive move to better fit the organization to changing environmental circumstances.

Their review shows that

While [...] employee downsizing announcements are generally associated with immediate negative effects on the stock price, the long-term stock performance effects are rather mixed. [...] Researchers state that employee downsizing leads to sustainable performance improvement, but also that it has sustained harmful effects on organizational profitability. Equivocal findings were also identified regarding employee downsizing and labour productivity, firm R&D activity and sales growth.


[Researchers] found that employee downsizing’s performance is closely related to the way it is implemented. [One article] identified underestimating employee downsizing’s profound effects on work processes and structures as the main source of negative performance outcomes.


[E]mployee downsizing runs the risk of undermining sustainable competitive advantage. Sometimes referred to as ‘dumbsizing’, employee downsizing efforts may result in the loss of key knowledge and individuals, leading to deteriorating quality, productivity and effectiveness. In particular, foresight and downsizing strategies’ non-prioritized implementation tactics run the risk of causing the loss of valuable institutional knowledge and memory if the wrong employees are laid off.

In short: while good results are possible, how a downsizing is done really defines whether it will work or not. They mention companies often losing people they shouldn't have and then having to bring them back as much more expensive contractors (having this happen to you sort of owns though, IMO). In order to evaluate who you really want to keep, the authors suggest a model.

Organizational Knowledge Model

Retained Knowledge is the result of things known at the individual (experiences, observations, knowledge), group (practices, routines), and organizational (process, culture) levels. It enables a process of sorting, categorizing and sense-making within an organization that lets it apply knowledge in new strategic ways, influences how new information is framed, and in turn allows behavioural changes. All these processes are intertwined, and knowledge is shared in ways that exceeds any individual's own cognitive abilities.

Employee downsizing not only runs the risk of destroying valuable organizational knowledge on the individual and social network levels, but may also profoundly disrupt established procedures, routines and the organizational culture. These more indirect effects can have severe long-term consequences.


effective employee downsizing efforts should be considered part of the firm’s long-term strategy to preserve its critical knowledge for sustainable competitive advantage. The word ‘critical’ knowledge is of particular importance for effective employee downsizing, as an elementary aspect of organizational learning is based on forgetting. Unlearning is often a necessary condition for change.

In this respect, knowledge retention can be viewed as both beneficial and detrimental to organizational functioning.

This is interesting as a point: forgetting is good. Sticking to old patterns can be problematic. Downsizing, if judiciously applied, can help organizations move on and adapt in ways its internal feedback loops wouldn't let it. However, doing it wrong risks disrupting all sorts of tangled organizational knowledge mechanisms and be real negative.

They refer to a 4-levels framework for knowledge, and then derive guidelines about how to have better results.

Intrasubjective level

This includes explicit and tacit knowledge. The paper frames both as mutually dependent: tacit knowledge guides the application of explicit knowledge. It's what guides hardly conscious decisions every day and lets people adjust heuristics and strategy in the most effective ways. But tacit knowledge is pretty much impossible to surface nor describe, and is therefore much, much harder to replace than explicit knowledge.

Particularly, it can only be transmitted through interactive conversations. The supposition is that dense employee networks are one of the best places for this transmission, but only if there is a strong level of collaboration taking place. More superficial interactions don't necessarily do much, nor would adversarial ones either.

So aiming for high levels of collaboration is a way to foster conditions to help keep tacit knowledge alive.

Intersubjective level

This covers people learning together, and sharing knowledge. Knowledge sharing across groups is what enables organization to sustainably maintain it.

In organizations with "structural holes" (where silos don't communicate), diverse groups will have to acquire new knowledge from independent sources that don't overlap. These groups, since their knowledge bases don't overlap, end up increasing the diversity of information in an organization and grow its overall variety of interpretations. But weak relationships can be created, which will allow the transmission of these perspectives and will function as vital "transactive memory systems":

employees with only strong ties in an organization are very likely to be part of dense networks with a closed system of information flows. Conversely, employees with weak ties are connected to multiple sources and thus have access to multiple networks of distinct information. These weak ties often appear between dissimilar people and fuel creativity since employees are exposed to perspectives that differ from their own. Moreover, multiple connections to other organizational networks and subunits enable employees to understand the entire internal organizational context and allow them to position their activities in a global context. From a resource perspective, weak ties between employees are therefore more valuable than strong ties.


[Losses of specific individuals] are particularly severe if the individual can be labelled a ‘central connector’ (an employee with high technical expertise, organizational memory and critical relationships), ‘broker’ (an employee with broad retained knowledge of the organization’s functioning and capable of connecting, coordinating and integrating networks), or ‘peripheral player’ (an employee with niche expertise, often well connected within external networks).

So downsizing that ends up losing you these individuals can have a very outsized impact over a longer time period.

Collective level

This is where the knowledge from intrasubjective and intersubjective levels gets transformed into discrete procedures and routines, whether formal or informal, through rules or customs. It's essentially how people consider that things are done here.

Even if they are based on substantive knowledge and may have been chosen by people who were very intentional, they are often applied in an automatic and inarticulate way.

While routines allow organizations to retain and use knowledge effectively, they could also create certain dynamics that could automatically transfer past experience to inappropriate situations, thus creating inefficiencies and failing to create new knowledge.

A lot of this is influenced by incentives and the need to properly align them. The ability to adjust and change routines depends on Psychological Safety, which is the sort of spirit where your management and team give you the shared belief that well-intentioned experimentation won't land you in trouble and won't be punished.

The authors point out that changing the organizational structure drastically during a downsizing may impact and change these elements and harm the capability of further adjusting routines, and therefore turn your collective level into a stiffer, less adaptive structure.

Extrasubjective level

Organizational culture refers to a shared common frame of reference that employees learn socially and transmit, thus providing rules for organizational behaviour. This covers things like shared identity, how likely departments are to share information with each other, and so on. It also covers abstract concepts like morale and attitudes.

Stating the somewhat obvious:

Employee downsizing’s psychological effects have been proven to influence employees’ morale, work productivity, motivation, commitment and job performance. Consequently, there is a risk that employee downsizing could disrupt the culture of trust and safety, which, in turn, could reduce the employees’ willingness to apply their retained knowledge to present decisions. An accumulation of negative perceptions increases the risk of turnover among the survivors, which will undermine the envisioned savings through employee downsizing.

[...] Survivors often evaluate an organization’s overall fairness in terms of how employee downsizings are carried-out.

I sort of guess anyone who has been through downsizing knows this, but I guess it's good to prove it when published in a thing called "Management Learning."

Things like "cutting a fixed percentage of the workforce across all departments", while intended to communicate fairness, can end up having the opposite effect and lowering overall trust. How the people being cut are chosen, and how they are treated once they've been selected, all have a lasting impact on the extrasubjective level and your ability to use the knowledge of whoever is left.


This yields the following suggestions from the authors:

None of these proposals have been tested in practice. The authors note them as suggestions that may help, but need testing.

The authors do mention that a lot of the effects of a downsize are not going to be visible to quantitative metrics (eg. monetary value) but will only be visible to qualitative analysis (innovation, voluntary turnover, reputation, quality levels in work output) and that measuring a better bottom line may not be as much of a slam dunk as the numbers may suggest.

They also provide this helpful table that sort of TL:DR;s a lot of the content:

To better choose who to keep or not, they suggest just asking people to characterize the relationships they have with other people they interact with. This lets you see past the org chart to find the various influence structures:

The loss of individuals with very few links to other organizational employees could mean that the tacit knowledge embedded in these employees may be lost. [...] Additionally, managers can identify those individuals who form a unique link between otherwise unconnected networks in the organization. Maintaining these employees during employee downsizing prevents the destruction of critical network connections between organizational units.

There's also a caveat for organizations that have cooperative vs. competitive cultures. Cooperation helps form links in dense networks and maintain knowledge across networks. Competitive environments instead tend to have a poor ability to share that knowledge, so if you have a high-performing unit there, cutting from it risks creating irreparable damage (whereas cooperative teams could repair more damage). In these low-cohesion environments, you also doubly want to keep people who provide the few rare links to other parts of the organization as well, since they’re critical and there’s so few of them.

So anyway, that's why I'm watching Twitter go so attentively right now. Case study being written right in front of our eyes.