[dijit]

> I’ve yet to find a service that saves 5 hours per week per employee. How do I estimate the actual savings in time?

Sure you have, albeit by another name likely.

GitHub/Gitlab as a service easily saves more than 5hrs a week.

I have custom slack bots that easily save me a couple hours a week in aggregate.

Then there’s services such as managed CI or, heck even things like the “search” function on a wiki, those are all things that can be provided by a service.

But a tool like this will show you much much it might be worth investing in a service vs hiring someone dedicated and running something yourself.

> Work expands to fill available time. Will my employee use that time to the company’s advantage?

There’s two points to this argument;

1) if I save an employee time, what value does that give me?

2) if I’m an employee, and efficiency is improved; I still have to be in the office 9hrs per day.

The first argument is at odds with the notion that most knowledge worker jobs tend to only be around 40% productive.[0]

There’s no evidence that it goes lower than that; most of the reasons that percentage is so low, though, is friction. Friction can take many forms such as a bureaucratic process for approvals to change things- all the way to “needing to talk to that one guy who knows the thing, and teams is having an outage”. It’s hard to quantify, but there are so many frictions and there is evidence to suggest that removing these frictions increases productivity, not lessens it. (To a value of 80% which represents a significant increase).

(I will supply citations when I get to my pc, this comment is from a phone)

Problem 2 goes into the expectation that if you’re in the office you must be busy- there’s no value to you the employee of the company gets more efficient! Except obviously that’s not true in a more macro sense; I wouldn’t argue that. I would instead argue that the feeling of empowerment that comes with doing actual work and not busywork will make people more engaged and not less.

You wouldn’t feel motivated in your job if you had to assemble your chair each time you wanted to sit in it, it would be tedious and not challenging and certainly cause you to mentally check out.

[0]: https://talentculture.com/how-knowledge-workers-really-spend...

[lucideer]

> Sure you have, albeit by another name likely.

> GitHub/Gitlab as a service easily saves more than 5hrs a week.

> I have custom slack bots that easily save me a couple hours a week in aggregate.

> Then there’s services such as managed CI or, heck even things like the “search” function on a wiki, those are all things that can be provided by a service.

These are three highly subjective, very unconvincing statements. I use Github, Gitlab, multiple slackbots (some I wrote, some others wrote), managed CI, and a few search services internally in my company every day. I have no confidence that any of them are timesavers in the way that you state.

Slackbots in particular have been shown to use more time than they save (context switching is extremely costly). Github is a source-code host that ads distracting social features, notification queues, etc. all which can add to an employees distraction load and decrease productivity when compared to a more basic code-host.

I'm not saying they don't save time in aggregate, but there's certainly arguments on either side.

[dijit]

I know a company which I will not name, that was passing source-code around with a USB drive.

Their version of an SVN lock was basically 'who has the USB stick right now'. And while this is an absurd and extreme example (and a true one, crucially) you can't deny that github would have saved those developers countless hours.

Maybe even enough hours to pay a person full time to manage something on-site: but that's why this topic (and the OP's site) is interesting, how do we quantify it?

[jonfw]

You're supposed to compare a service against it's competition, not the least productive thing you could possibly imagine in it's stead.

If you walked into a car dealership and the best thing they could say is "It's WAY better than walking!"- that wouldn't make a great pitch.

[dijit]

This is such a bad faith comment I don't know where to begin.

Of course no reasonable person is saying that you have to compare against the worst thing, that's stupid- I was simply stating that I've seen things that have easily quantifiable returns.

I'm comparing it against the 'nothing' that I would otherwise have.

If you're comparing something then that's yours to compare, and this is a tool for doing that.

if you're not running github or gitlab, what are you running?

Maybe SVN+jira? or gogs? or gitea? what about teamcity?

I'm not going to break down the cost savings and expenses of each of those, I'm just saying we're all already using services that have saved us many hours a week compared to those services not existing in any form.

Its up to you to debate the 'many forms' a service takes, and remember that server hosting and human time is not free, so something self-hosted that requires some hours of time to maintain needs to be cost controlled for.

[mattmanser]

How is it bad faith? You're wrong, he's right.

You can run your own code repository, people did for decades, and took backups home. These days you could just send one to A.N.Y.Other cloud service.

It doesn't take 5 hours per week, and if it takes 5 hours per year I'd be surprised.

[dijit]

It’s bad faith because it speaks to the content of what I said and not the point.

I am one of those people who managed code hosting repositories: but it’s completely absurd to assume that code hosting was without any cost involved at all, and to remove all of the other integrated features too? No. Absurd.

And anyway. The point was making is that we are already paying for services that save us a lot of time- they are of incredible value, and thus universal.

And yes, you might not spend 5hrs a year on _just_ code hosting but code hosting and web view and merge request portals and issue tracking and so on- should those services not be provided somehow (or be provided by something like jira/swarm etc); would easily cost more than that in time.

Hell, even running gitlab is an hour/w job just ensuring that backups are well tested and CI machines are purged, running updates and so on.

It has a low cost, because it’s just one person doing it, but the overall _point_ was that we already have some services that save us time and the parent was not speaking about them.

[projektfu]

Before git and mercurial I used CVS. It’s pretty easy to set up a server if you have ssh or telnet access. For personal things I used RCS because it was integrated nicely in emacs. So I think one needs to at least compare it to those.

RCS had almost no configuration required. It just wasn’t easily shareable. So, going to git costs cycles for personal projects with the hope of a return from better management of sets of changes, etc.

[throwaway43234]

> notification queues ... which can add to an employees distraction load and decrease productivity when compared to a more basic code-host.

If you think your employees would be less productive if they didn't have a queue of work items to address and co-worker comments to look at and respond to... let's just I wouldn't work at your shop. You running a room full of air-gapped code production units or something?

[lucideer]

I think you're oversimplifying. There's a world of difference between "all notification queues decrease productivity" and "some notification queues can decrease productivity in some cases".

Seems something of a divide in interpretation of comments here between people who get subjectivity and people who insist everything is necessarily a binary.

[mattigames]

> all which can add to an employees distraction load and decrease productivity when compared to a more basic code-host.

Is a bit naive to think those features are the ones distracting the employees, when if you watch a developer at work 98% of the time-wasters are the strongly attractive ones: YouTube, Instagram, etc.

[lucideer]

Dunno where you work, but this doesn't correlate in my workplace.

[wil421]

How does Github save time over the predecessors like self hosted a Git or SVN repository?

[dijit]

This is reductionist, I am not the first to fall in line to support github here, but lets expand on what github is instead of assuming.

Github is, primarily, managed source code hosting- but it has more components, so lets break them out:

1: Source code hosting

2: Web view

3: Issue tracking

4: Project management tracking

5: Search

6: Authentication and identity (oauth2 and "applications")

7: Documentation rendering

8: Web hosting (a-la github pages)

9: CI pipeline (github actions)

SVN replaces point 1, and poorly. How many hours do you need to manage an SVN server? Very few I would wager but it at least behooves to ask the question. And it certainly behooves to understand that you're not replacing 1:1; you're replacing many things with one, and managing the above yourself can be done cheaper, but poorer, so that "poor" imitation could cost more time for using it too.

It's a very nuanced topic but an interesting one, and reductionist questions like this are not helpful.

[deviantfero]

Pull request management, code reviews, CI flows, github actions, github does have many features that other git clients don't, certainly not a self hosted git or SVN repositories, and those features save a lot of time just in the ease of initial setup.

[jeremymcanally]

Permissions. I specifically remember talking to Tom and Chris about how that was really the main point of rage that spawned the idea of GitHub (and its original slogan “Git hosting: No longer a pain in the ass”).

[ori_b]

Did you spend 5 hours per week per employee managing permissions in git?

[jeremymcanally]

Setting up and managing keys, spinning up new repos, etc. would suck up an inordinate amount of time. This was especially true when I worked at consultancies. Blocking developer work with “Sorry gotta wait on Todd to add you to the server and set you up on Git,” when Todd would avoid it like the plague because it sucked would lose a lot of time.

If you averaged the time out, it may not have hit the five hour per week mark, but it combined with other time savers (linking Git directly with the issue tracker for example instead of having to figure out how to cross reference them) easily did/does.

[ori_b]

> Setting up and managing keys, spinning up new repos, etc. would suck up an inordinate amount of time.

The claim was 5 hours per employee per week. I found that it took far less than 50 hours a week when working at a 10 person company. And if it did, I'd question the competence of whoever was doing that maintenance.

I'd pin it at closer to 3 or 4 minutes per employee per week, with maybe 8 hours of setup once. That's still a bunch of work, but it's a whole different ballgame.

[cmbell715]

Maybe not managing permissions directly. But I could easily see a number like that, or higher, being plausible if you consider clean up from junior devs accidentally pushing to development/main branches thinking they were on their feature branch. Which is an issue directly solved by permissions.

[projektfu]

To the extent that a service avoids yak shaving, definitely that value can be quantified. How do I know it won’t cause yak shaving though?