An idea came up in several conversations in the last few weeks: the analogy between externalities in economics and the location of activity and costs of software development. This post just dumps the thoughts I had around it.
In economics, an externality is
an indirect cost or benefit to an uninvolved third party that arises as an effect of another party's ... activity ... Air pollution from motor vehicles is one example. The cost of air pollution to society is not paid by either the producers or users of motorized transport to the rest of society.Externalities may be beneficial; my neighbour might buy and plant flowers that I can enjoy from my side of our shared garden fence at no cost. But the recent discussions have been more about negative externalities.
Everywhere that software is developed by more than one person there is process around the development of the software. In any given shop it will be more or less defined, refined, and front-of-mind, but it exists and has two major aims: design, create, and ship artefacts and expose, transfer, and store information.
Artefacts are straightfoward. That's the software, user documentation, change logs, and so on. Information is messier. In typical processes it might include things like
- who wants this piece of work
- what they want it for
- what constraints there are on it
- what is blocking it
- what risks are perceived with it
- what depends on it
- when it's needed
- what has been tried before
- what was implemented this time and why
- what was tested
- what was found during testing
- what wasn't tested
- when work started
- when work ended
- who spent time on it
- how much effort was spent on it
- which work package it was part of
- which commits were made
- which branch commits made into
- when the code was merged to the main branch
- when CI first ran on it
- what the results of the CI run were
- which release the work went into
- ...
In each case there are producers and consumers of the information.
Sometimes the producer will be the tooling and there will be no local consumer. We can igore this case, just like we probably ignore the raft of pointless reports and charts most process management tools generate.
More interesting are the situations where the producer and the consumer are people. For example, a company board producing goals for their product team to consume, a product owner producing a list of desired features for a development team to consume, or development teams producing documentation for an auditor to consume.
Since we are human, and busy, and tend to prefer doing over recording, we'll often skip or skimp on the information. We've all seen (and made, be honest!) drive-by one-liner tickets, empty commit messages, and lazy PRs.
But each time we do that we push a cost to someone else: the PO who has to second-guess what the board want, the development team who have to second-guess what the PO means, and the business who put all hands to the pump tidying before an audit and rectifying after it because auditors aren't paid to second-guess.
These feel like externalities and, in market economics, there are some standard approaches to tackling externalities including taxes, regulation, and compensation.
In software development processes, taxes and compensation don't have direct analogs. Although producers are paid for their work, it seems unlikely that board members will have their salary docked if they don't take the time to communicate their needs well.
Regulation feels more familiar, perhaps in the form of reports, templates with mandatory fields, and delivery metrics. Of course, we've all seen these and we all know that it's easy to regard them as friction and it's common to want to make the minimum effort necessary to get past them and build more things.
Given all that, could this model of information flow might have some value in software development? If so, we might ask how we could give ourselves more than an outside chance of reducing negative externalities and
- help to produce the right information at the right time
- keep costs of production proportionate
- have the costs paid when they are cheapest to pay
- locate the right costs with producers
- motivate the producers to pay the costs
- increase positive externalities
My gut reaction includes shortening the distance between producer and consumer, working in smaller increments, minimising handovers, and asking for information when it's needed rather than pushing non-information when no-one is ready to receive it.
It's interesting to me that these all feel like cultural aspects of process, requiring collaboration and personal responsibility.
Your thoughts are very welcome!
Image: https://flic.kr/p/Eghfa4
An interesting article - thank you. I hadn't thought of externalities where it comes to data, but had thought of user experience (blog waffle: http://randomtechthoughts.blog/2021/01/23/user-experience-ux-and-data-quality/).
ReplyDeleteOne part of the solution is, as you say, to reduce the friction or cost to produce useful information. This can be addressed by improving or creating internal tools.
E.g. a form with a field that auto-completes with helpful (often context-sensitive) suggestions, or lets you pick from a list, rather than forcing you to type. Tools that talk to other tools behind the scenes to spare you the grunt work / find helpful suggestions for you. Or tools that create thing A by applying code to thing B, rather than a human having to create thing A and thing B separately. One example that springs to mind is things like API documentation being auto-generated by Swagger / OpenAPI from source code + comments + annotations.