The episode explores the tension between necessary on-the-spot decision-making during a software incident and subsequent organizational governance. Andrea’s story sets the scene: a colleague makes a critical, correct decision under pressure, only to have a later, higher-level investigation reverse it without adequately addressing the systemic issue that caused the incident, leading to frustration and burnout. We then further discuss how organizations—whether centralized or decentralized—often struggle with decision paralysis and the human consequences of chronic underinvestment in long-term fixes. The conversation emphasizes that true decentralized architecture requires not only empowering individuals to act but also leadership allyship and robust, non-weaponized processes to support and record their actions.

The core of the governance solution lies in documenting decisions effectively and creating a culture of learning over blame. To avoid "decision paralysis" and preserve the historical context, the hosts advocate for recording critical choices using Architectural Decision Records (ADRs). A key principle is that these records should be immutable; any subsequent change must be documented as a new, superseding decision rather than reopening the original. Furthermore, decision-making health can be assessed both quantitatively (by measuring flow and reversal rates) and qualitatively, by including sense-making questions about team readiness and feelings toward the decision, drawing on insights from practitioners like Rebecca Wirfs-Brock. This approach transforms governance from a bureaucratic bottleneck into a feedback loop that highlights deeper systemic failures.

Key Takeaways for Architectural Governance

• Immutable Decisions: Decisions recorded in ADRs must be treated as read-only history. Reversals or changes should always be documented as new, superseding decisions to prevent "decision paralysis."
• Leadership Allyship: Managers should use their political capital to support and empower the team's expert decisions, acting as facilitators and advocates rather than simply taking over.
• Systemic Focus: Governance must move beyond validating individual actions to addressing the root causes of repeated incidents and failures to prevent team burnout.
• Decision Metrics: Measure the health of the decision-making process both quantitatively (number of decisions, time to decide, rate of reversals) and qualitatively (team's emotional readiness/frustration).

Kenny (Baas) Schwegler: Welcome

back everyone to this another

2

episode of facilitating Stories on

Software Architecture and Design.

3

So we are gonna talk today with

Andrea, who's gonna tell us a story,

4

and with us is Andrew as well.

5

Hello Andrew.

6

this content series, which we don't give

it a name, we don't call it a podcast or

7

a flog or anything, it's just a series

of, conversations about stories where you

8

might be an architect or a designer that

believe that we need to do decentralized

9

decision making or facilitating

software architecture, but you struggle.

10

And these are all stories about

these struggles that we on a

11

daily basis still have as well.

12

So it's also a little bit for us.

13

We learn a lot.

14

in the previous episode, we

say, well, as long as nobody's

15

listening, doesn't matter.

16

We're learning, so we hope

you are learning as well.

17

So, Andrea, it's up to

you to tell a story.

18

Andrea Magnorsky: It's my turn

to tell a story and I'm switching

19

Tuck a little bit, as I tend to do.

20

And, what I'm gonna give you is, a

scenario then I'm, of wanna ask some

21

questions and how would you go about it?

22

So this is the scenario.

23

Imagine you just arrived at work

and you're doing work stuff, and

24

suddenly there's a, a ping in your

whatever messaging platform you use

25

for, communicating with your people.

26

And in there, there's an incident

like a high importance incident.

27

One of your colleagues.

28

Goes off, investigates, you're like

up in your hand and they're fine.

29

And then during thing, colleague,

decides to stop something.

30

They're like, this client can't do X.

31

That's their decision.

32

And it's in the middle an incident.

33

That's what happens that time.

34

Because it's a high, important instance.

35

the very next day, let's say there's a

meeting with your colleague, yourself,

36

and a few other peoples that are higher

in the hierarchy to review this and other

37

things during that scenario, the review

is that the action that your colleague

38

took was correct, which is an interesting

point that you might, want to, comment on.

39

then, that's where it stays.

40

However, there is an obvious thing

that, people that are in the hierarchy

41

a little bit higher than you, they want

to investigate further to see if this

42

is something like, what do we do next?

43

Okay, this was the good decision for

now, but what do we do in the future?

44

Someone volunteers to

do the investigation.

45

Another one of your colleagues, they go

off into it and they come back saying

46

that actually we should allow it.

47

There's another meeting.

48

This same group where this happened,

communicates to the rest of us that

49

the decision has been reverted and

actually precisely the thing that

50

was stopped is with no change.

51

The only thing that

changes the understanding.

52

The hierarchy, has, but with not a,

a lot, extra context, but not much.

53

Your colleague is, not completely

impressed because they are,

54

well, this is gonna happen again.

55

That's the context.

56

That's the story.

57

The story obviously, I

mean, I'm sure it happens.

58

Something similar, probably

happened to you, so.

59

This question that I'm gonna

pose is about governance.

60

We were talking about it a little

bit in the previous episode.

61

And so also about how to, how

would you deal with this or what

62

could happen in option A, a very

autocratic kind of centralized place

63

versus b, more decent central place.

64

Like, can we, can we talk

about the whole experience?

65

Like incidents happen all the time.

66

You make decisions, None of

those things are, strange and they

67

will happen in both, scenarios.

68

But I would love to hear about,

how things could be different or

69

the same, like any similarities and

differences is where we should top off.

70

especially focusing on governments.

71

Whoever wants to go first.

72

Andrew Harmel-Law: The only thing I

wanted to say that I think you implied

73

happened, but one of the biggest things

that I make sure is like, maybe you

74

have to retrospectively create it,

you document each decision separately.

75

So you write an a, DR

maybe retrospectively.

76

'cause the person didn't have time.

77

They just, you know, based on

the things they knew at the

78

time, they took a decision.

79

so that should be represented in a DR.

80

Then the revisiting of that

decision shouldn't reopen that.

81

A DR, after they're decided

acted upon ADR is a read only.

82

That's like a big deal for me.

83

So then there should be a subsequent

decision, which supersedes the older one.

84

So you've got the record of the

history and all that kind of stuff.

85

when you, if you don't have that, if you

keep going back and reopening decisions.

86

That's just like paralysis, right?

87

And then it becomes very emotional

and all this kinda stuff.

88

So I've got other thoughts about shelf

for Kenny, but that's the big one.

89

It's like, right?

90

'cause it's not like one decision you

kept going back to one was actually taken

91

and then there was a subsequent thing.

92

So that's, would recommend it separately.

93

Andrea Magnorsky: Yeah, it's

more like the, I'm making an

94

emphasis on a few aspects here.

95

The governance, the human story and

what happened, also what we like it

96

to be different and how, and feels

like one coherent story itself.

97

Like at the end of the day, the system,

has like is behaving in a certain way.

98

that's kind of interesting too.

99

That's the three kind of.

100

That I'm seeing in the story.

101

So I totally agree on the, on the

decisions themselves and that they

102

should be not, not all encompassing

and, oh, actually, let me just,

103

let me just one more thing and

then you'll never finish anything.

104

Kenny (Baas) Schwegler: I think Rebecca,

and we should invite her once she talks

105

about the earliest responsible moment.

106

You can take a decision, right?

107

And, you have the last responsible moment.

108

But the earliest, I think

she calls it earliest.

109

but she says, take him

as soon as possible.

110

And then you go into, Andrew's not right.

111

We can take him now and then

you can supersede them later on.

112

And I think.

113

It's the same with coding, right?

114

Make small codes.

115

People are very afraid of

making a decision there.

116

So that's the human aspect.

117

For some reason, people are very much

afraid of taking smaller decisions

118

or smaller changes for some, reason.

119

and the second thing that, I

hear you say, and I'm curious

120

and that's a question to you.

121

I, she's not impressed by the decision,

122

People sometimes see decisions

as a single thing in time, but.

123

It might be many decisions.

124

So for some person coming in and,

and maybe it's a new person, right?

125

And they take a decision, but for

that person it might be maybe the 50th

126

time that it happened to that person.

127

Right.

128

I'm not sure if that was the case here.

129

Is that what, what you're, because they're

not, I felt they're not impressed, so I'm

130

like, okay, that there's something there.

131

Maybe, I'm not sure.

132

Andrea Magnorsky: So, for some further

context, in this particular case that

133

I'm thinking of, some of this comes as

a background of we're not investing in

134

actually solving our systemic issues.

135

And when there's an incident, we are

the ones that have to respond back end.

136

this particular colleague was very

tired of like, well, you know, why

137

can't we actually fix the problems?

138

Why do we keep saying yes?

139

And the pressure is always like, well,

because the market moves faster than us.

140

Which is, I get it.

141

Also there's a balance

there that needs to be had.

142

You can only.

143

Uh, plasterers over your wounds.

144

So many times, if you keep

kind of going, like imagine

145

you have a wound there, right?

146

And it's not healing.

147

And so before it starts

healing a little bit.

148

Uh, plaster.

149

Okay, cool.

150

Maybe okay for now, but you have

to wash your hands, et cetera.

151

Soon after you plaster, plaster,

plaster thinking, plasters, and at some

152

point your hand actually gets damaged.

153

I think this is the point at

which my colleague was when

154

they were not impressed.

155

Kenny (Baas) Schwegler: Yeah, I think

that goes a little bit into hierarchical

156

structures, and I have a little bit of the

157

thing against, not against, but the, the

partnership of management roles, because

158

what I usually see is that managers

are like, I'm in this position to help.

159

But what does help mean?

160

So that A implies that people need

help that's the first assumption

161

I see managers take, do they

actually need help with this?

162

And b, is it for me to solve?

163

So, this is what I see a lot of

managers do wrong instead of.

164

Asking first, okay,

where do you need help?

165

How can I help you make this

decision better next time?

166

And give people insight.

167

They rather think, Hey,

I need to help them.

168

That's assumption A, and then they

are already moving forward, so

169

there's not even an agreement on

should we change or should we help?

170

Someone's already moving forward with,

okay, then we need to go forward and that.

171

That's like a mental

model separation, right?

172

One person is like, ah, this was fine.

173

We took the decision, let's move on.

174

The other person is, well, no,

no, we're gonna investigate this.

175

So it should be more facilitation

and collaboration in that point than.

176

You know, it came as a news

flash a week later that again,

177

someone else took the decision.

178

And I've had this with a

manager that says, no, I'm

179

not gonna take this decision.

180

You're gonna take this decision else.

181

We're gonna breed into the

systematic problem that

182

Andrew Harmel-Law: Yep.

183

Kenny (Baas) Schwegler: Teams don't

make decisions because they're, and

184

then a manager takes decision, but

they're not okay with that decision.

185

So

186

Andrew Harmel-Law: Yep.

187

Kenny (Baas) Schwegler: that

manager actually took a stand.

188

I'm not taking a decision anymore because

every time I'm just feeding the beast.

189

Andrea Magnorsky: I think in this

situation it was, really hard

190

to do anything that feels right.

191

Basically, this is why I think it's

such a good place for asking about

192

governance and asking about what

would this situation feel like and be

193

like, and how can we facilitate these

decisions in a more centralized space?

194

That's why I thought this is like

a rock in a hard place, right?

195

So,

196

Andrew Harmel-Law: I think

there's something as well

197

about like allyship, right?

198

And I don't just mean like, like DEI

allyship, like you said, Kenny, right?

199

You could lend your political

capital or organizational capital

200

or whatever, social capital to the

person making the decision and echo

201

and repeat it as opposed to taking it

on, like make sure that the ownership

202

still belongs where that person is

203

the weight of your experience

and maybe extra context, right?

204

Like, you know, this

isn't just happened once.

205

This happens all the time.

206

And here is me as a manager saying 30%

of this team's time is spent restarting

207

this system because of this thing

that's, you know, and so therefore we

208

know that, you know, the revenue from

various customers is, is super key.

209

But also we know that the loss, you

know, there might be access and have

210

privy to access to information, the

person just turning the thing off.

211

Hasn't got, but they might be able to lend

that kind of political capital to stuff.

212

I've seen the anti-pain where people go,

we are all in this room to decide, and

213

if you don't decide, I'll decide for you.

214

And the end result is no one

decides because they don't wanna

215

be responsible because then it'll

ultimately go to the big box.

216

Right?

217

And then that drives

completely wrong behavior.

218

So.

219

Kenny (Baas) Schwegler: I had some

very good examples early on in my

220

career from project managers or

managers that did the Good King.

221

I remember one project manager that came

to us and this was like a, like a, a.

222

Banking industry where, where after the

crisis, where, you know, there was a

223

project that they need all the information

to know that the bank is healthy,

224

So then we had this project manager, I

remember vividly coming to us and he said,

225

what do you think that is needed, right?

226

Because the, COO thinks this and

this and we sat together as a team.

227

This is what we need, this

is how we need to build it.

228

So he asked us, he kept

asking question what we need?

229

Okay, so this is what you need.

230

Okay, I'm gonna arrange it.

231

And then later that day he started

launching with the CEO and you

232

saw him really influence as a CEO.

233

And later on that day, he

came, okay, I arranged it.

234

Let's go.

235

I put full trust in you as a team, and

we were very successful because he put.

236

That person enabled that

success to happen, right?

237

They're saying, okay, this is a different

approach, but you're the expert.

238

We hired us experts.

239

And I think that's for any architect

in that role, that that's the shift

240

in, so I've seen managers, you

can do, choose two sides, right?

241

You can choose the side of the hierarchy

or you can choose the side of the team.

242

Andrew Harmel-Law: Yep.

243

Kenny (Baas) Schwegler: my

main value based heuristic, I

244

always put, value into the team.

245

Andrew Harmel-Law: Yep.

246

Kenny (Baas) Schwegler: And that's hard.

247

But yeah, being in a

leadership position is hard.

248

So.

249

Andrew Harmel-Law: the thing,

if you lose the team, right?

250

Like maybe this time they, you know,

like your, in your example Angela, right?

251

Maybe they go, okay, I'll just do it.

252

And they go along with the decision,

but you are slowly losing that person's,

253

Andrea Magnorsky: Oh no,

they're, they're gone.

254

No, it's actually, in fact I think

it's kind of systematically designed

255

so that people burn out after some

time and they leave and they can,

256

there's enough time to replace them.

257

Andrew Harmel-Law: So the people

who have been no longer, who

258

don't have to live in this world.

259

I mean, this is the thing, like what I

want, I dunno what it is I want to write

260

about next, but like something about

the fact that we live in these codes,

261

sociotechnical systems, that's not, we

don't pay enough attention to that, right?

262

But it does.

263

And we move in and outta it.

264

We change companies, we change job

roles and we moved teams and stuff, and.

265

Are big deals.

266

Kenny (Baas) Schwegler: We

don't pay enough attention to

267

this systematic thing, right?

268

There's this famous comic I share with

Evelyn Hin in our trainings that was

269

like, okay, who wants, who has some ideas?

270

And then someone says,

well, I think we know.

271

Okay, more ideas.

272

I think there's.

273

No, that's shit.

274

Come on.

275

Keep sharing ideas.

276

We're not aware of that systematic, so if

a person gets told like three times, no,

277

of course they're not gonna share anymore,

or they're just gonna do what they want.

278

I remember, people saying, yeah,

I'm just here to do my job and then

279

I get back home, which is fine.

280

But at least.

281

Make it explicit how we're gonna do it.

282

And I think consent is there.

283

Good.

284

How are we gonna work together, as well.

285

I think that's the important

part is, okay, how much decision

286

making power do you have?

287

And this is what I talk a lot and we

tell in the book as well, I've been in

288

this situation where, I was going to a

CTO and the CTO came in and sat me down.

289

He says, Hey, you've been working for six

months now on this thing, and we're going

290

to Azure, and I have this idea maybe,

291

looking at the market, we should

move to Kotlin, for instance.

292

I don't think that's a good idea

because everyone here does C and

293

it's a whole different ecosystem

and yeah, we're gonna do it anyway.

294

And I'm like, no, wait,

what happened here?

295

You already took the decision

for yourself, and that's fine.

296

You're in leadership position, but at

least tell me that you took the decision.

297

Even though in your book, Henry, you

talk about decentralized decision and

298

the advice process, you might not be in

that situation, but you can start writing

299

ADRs and say, this is the agreement.

300

This is how we are gonna take the, there's

several other options of taking decisions.

301

And even though you don't follow

the advice process, you can

302

at least make an a DR and say,

we're gonna follow this decision.

303

Andrew Harmel-Law: 100%

304

Andrea Magnorsky: I

have an add-on to that.

305

what if that actually everything

is being documented, but

306

listen to what I just said.

307

Everything is being documented.

308

So actually finding

something, finding coherence.

309

It is, let's imagine that, not just that

it's being documented, but something else.

310

It's kind of what if, the drs and other

similar documentation is being weaponized.

311

I am adding to the, so how

would you decentralize there?

312

'cause everyone's being

transparent enough.

313

But that doesn't remove the

inherent problems this broken

314

and hard place situation.

315

So

316

Andrew Harmel-Law: Hundred

317

Andrea Magnorsky: actually

what do you do then?

318

Andrew Harmel-Law: I've

got opinions on this.

319

Andrea Magnorsky: Okay, I'm all ears.

320

Andrew Harmel-Law: so before it became

popular, I quoted, 'cause I think you

321

know this, the famous CIA, manual for like

disrupting, some people seem to think it's

322

for disrupt, like it's for disrupting,

like, kind of, socialist bureaucracies

323

and like communist bureaucracies, but

like the patterns they lay out is like

324

just basically how to make everything

go really slowly or to like just confuse

325

everything or just create loads of noise

and, and it's basically follow the rules

326

to the letter and constantly complain

about things and write everything down

327

and all these kind of things, right?

328

So kind of.

329

Take autonomy outta the people

and put them into the system and

330

then you weaponize the system.

331

Right?

332

what I do to do that?

333

'cause people do this, right?

334

And people figure out, I mean

like I've done it, right?

335

Like if I'm like, this decision is

implicit, so let's weaponize an a d

336

but making it explicit so that we can

say, right, we know we don't agree,

337

but here it is, it's written down.

338

the other thing I use though, and

I need to write a blog post about

339

this, 'cause this isn't in the book.

340

If you do the four key metrics for

decisions, you can see the broader systems

341

like the, the flow of decision making.

342

So in the four, four key

metrics of decision making would

343

be the number of decisions.

344

Per month or whatever time

it takes to make a decision.

345

The number of, so instead of deployment

failures, it would be like decisions

346

which are superseded, and then the, the,

the time to restore service would be

347

the time it takes to come back, right?

348

So then you would see how many

decisions where they're coming from.

349

You could also look at like the

number of sources of decisions, right?

350

How long it takes to go.

351

So if it's very bureaucratic, they'll all

take ages and you'll see the same decision

352

being revisited over and over again.

353

And then you can step back and

go, this feels a little bit.

354

Weird, right, that's, this feels

like it's being weaponized, or

355

at least we're paralyzed even

if we're not weaponizing it.

356

We're just, we're weaponizing

it against ourselves.

357

'cause we're scared to

decide or something.

358

Kenny (Baas) Schwegler: I think you

can add that we use in the book,

359

so you can actually do a sense

making where you ask people how

360

ready are you to make the decision?

361

Right?

362

And then you can sense, we sense

make that, and we teach that.

363

People just, just start that.

364

Andrea Magnorsky: The quantitative

and the more qualitative,

365

Andrew Harmel-Law: Yeah.

366

Andrea Magnorsky: Way to kind

of actually, look at decisions.

367

So that's a heuristic, Kenny,

since you're apparently today

368

in a heuristic collecting mood.

369

Kenny (Baas) Schwegler: Yeah.

370

Andrea Magnorsky: when you're doing

decisions, make sure to, check

371

qualitatively and quantitatively

how your decisions are doing

372

Andrew Harmel-Law: totally.

373

Like Rebecca is also like,

you should add an extra thing.

374

I don't think it comes from Rebecca,

but there's some studies on this.

375

You should add an extra field

to your ADRs, which is how

376

I feel making this decision.

377

So again, things right,

378

Kenny (Baas) Schwegler: Yeah,

that's the sense making.

379

I usually do.

380

How do you feel about it?

381

Are you ready for the,

382

Andrew Harmel-Law: frustrated.

383

Kenny (Baas) Schwegler: power.

384

I did it in a previous training where

I just teach them this media meteor

385

developers, me and Evelyn teach that.

386

And their discussion was like, huh.

387

Oh, interesting.

388

And then the decision got much more

powerful and they were like, ah, this

389

is one thing that I can do tomorrow.

390

Right.

391

So adding these sensemaking to it,

and Rebecca talks about it as well.

392

Andrea Magnorsky: So there is

a few Rebeccas in the world.

393

And can we qualify

394

just

395

Kenny (Baas) Schwegler:

Rebecca Wirfs-Brock.

396

Andrea Magnorsky: Other people in this,

in this area, but I figured it was her.

397

But just.

398

Kenny (Baas) Schwegler: Yeah.

399

And people who, so that's

another heuristic start

400

following Rebecca Wesbrook.

401

Now, she's got some good content,

and she started the dd meme, right,

402

with a responsibility driven design.

403

She started, it's all Rebecca's

fault with the RDD, but,

404

thanks again for this episode.

405

So this was the last episode

with the three of us.

406

Next time we're gonna invite other

people to come up with stories,

407

so, we have a whole list already

that we talked about in these last

408

episodes that we're gonna invite.

409

So I hope, everyone enjoyed again,

watching this, reading this,

410

listening to this, whatever.

411

Alright, we're trying to bring

the content the way you like it.

412

If you like it, please follow us on the

podcast or on the YouTube and, give us

413

feedback, join our Discord community.

414

So you can go to virtual dd.com

415

and there's a big button that you

can join the Discord community and

416

you can keep the conversation alive.

417

Hope to see you there.

418

And if you have a story to

tell, come there as well.

419

Start chatting with us and, we

can invite you to one of the.

420

Other episode.

421

Thanks, Andrea.

422

Thanks Andrew, and see you next time.

423

Bye.