We like to think architecture starts with a problem. But often, by the time we're brought in, someone has already chosen the answer — and the real work becomes figuring out how to slow things down without making anyone feel foolish.

That's the situation Kim Kao, a solutions architect manager at AWS in Taiwan, walked into. A retail client — battered by the pandemic, competitors circling — had already been told by another vendor exactly what to do: containerise everything, deploy Kubernetes, and the operational problems would disappear. "Don't laugh," Kim recalls. The system underneath that confident prescription was twenty years old, and nobody left in the building fully understood how it worked. "Nobody knows what the content is."

Rather than argue about Kubernetes, Kim worked backwards. He ran impact mapping with decision-makers and functional leads, asking a deceptively simple question — who would support a goal of growing month-on-month revenue? — and watched the room fall silent. Then a two-day event storming workshop with forty to fifty people, many sitting together for the first time, surfaced merchant management as the place to actually begin. Somewhere in the middle of it, Kim had a realisation: "I found I was a businessman, not a technical guy."

This conversation explores what it takes to redirect a client who arrives with the answer already in hand — and why understanding the problem first is so often the faster route to solving it.

Key Discussion Points

• [00:04] The Answer Was Already Chosen: A retail GM, hit by the pandemic, arrives with a vendor's verdict — Kubernetes will fix everything
• [00:06] A System Nobody Understands: Twenty years of accumulated decisions, and "nobody knows what the content is"
• [00:09] Don't Decorate the Weakness: Why building on an unmapped system was "quite dangerous," and how merchant management emerged as the right starting point
• [00:11] Compared to What?: Making the true cost of Kubernetes visible — a million-plus active members, hours-long promotions, six-month hardware lead times
• [00:14] The Question Behind the Question: Separating the symptom from the cause before committing to any solution
• [00:16] The Silence in the Room: Running impact mapping, setting a revenue goal, and asking "who's your supporter?" to a team used only to taking orders
• [00:19] A Businessman, Not a Technical Guy: How connecting marketing, inventory, and logistics reframed Kim's sense of his own role
• [00:20] The Conjunction Role: Sending clients to DDD Taiwan and letting them discover the value of collaborative modelling for themselves

Guest: Kim Kao Hosts: Andrea Magnorsky, Andrew Harmel-Law, Kenny Schwegler

Hello, welcome to all of our listeners, at least

2

at least four of them right now.

3

my name is Andrea Mag Norski and

I'm here with my co-conspirators,

4

Andrew Harmer Lo and Kenny

5

Pass.

6

Uh, I think I'm completely destroyed that.

7

so, and we have a special guest.

8

Kim Ka.

9

We're here in stories of facilitating

software design and architecture.

10

Kim, you're gonna tell us a little bit

about what is it that we're gonna hear

11

about, a little bit about yourself,

and then we're hear the full story.

12

I'm so excited.

13

Kim Kao: Okay.

14

Yes.

15

Yeah.

16

Thanks for having me today.

17

Uh, hello everyone.

18

My name is Kim Go.

19

live in Taiwan and this is a, you

know, the great opportunities to have

20

a chance to link up the worldwide

DDD practitioners with you guys.

21

to, I think like to share a bit,

Personal experiences while I

22

was engaging with my customers.

23

before that I would like to put

on some background for myself.

24

I am a solution architect manager,

working with AWS for around

25

eight years to nine, nine years.

26

before that, prior to joining

a WSI am a software architect.

27

My job and the goal is to help people

to really understand technology

28

could support the business.

29

But in the past decades, opinions was

really learned from, you know, the

30

classic methodologists and the books,

you know, from the computer science.

31

So.

32

traditional, approaches like up

RUP, you know, the oriented stuff

33

that's we are really familiar with.

34

Right.

35

And during the decades.

36

I always think about the software

is everything software could deal

37

with any problem until I engaged and

learned from different perspectively,

38

so of the architects, for example.

39

People may have the different

knowledges from different perspective.

40

a, a really interesting point about

the architect, not, not sure everybody

41

really understand, what's of architect

if you took, if talk to different people.

42

They have different opinions.

43

So when I joined a WSI found

the name Solutions Architect.

44

really means to, you know, software

architect, it means everything.

45

You have to really understand the, like

for example, infrastructure, networking

46

databases, security, et cetera, et cetera.

47

They never realized wholly.

48

The world is too wide to learn.

49

And beyond this, I, have to manage the

conversation with different stakeholders,

50

like, you know, the decision makers,

product owner, and in different level

51

of the working level participant who

joined your meeting, you cannot even well

52

handle their desires or their challenges.

53

Then I failed.

54

The so-called solution architect, or

even architect couldn't be defined

55

by a single view of the points.

56

So that's brought me

stepping to another journey.

57

So today I want to share with you

a, a short story about how to engage

58

with my client tell a story the

vision to implement the solutions.

59

I would say the customer

is a retail customer.

60

I still remember the

date that the GM told me.

61

Kim.

62

you know, we, we were

under a certain situation.

63

We have the having impact from pandemic

and, you know, some of the competitors,

64

they are looking for chance to beat us.

65

We are almost down.

66

we want to have a different approach

to raise our, revenue growth

67

even to have better, performance.

68

So did you find out any way

to make some differences?

69

The GM told me, yeah.

70

We tried, we also, consult

with other called vendors.

71

but I don't, I can, I cannot mention

those, those vendors right name right now.

72

But the answer was like the, Jim told me

if the retail company would like to deal

73

with the operation issue and would like to

gain more business operation efficiency,

74

the only way is to use Kubernetes.

75

Don't laugh.

76

Don't laugh.

77

That's a crazy answer.

78

I had ever heard of it and I

was like, okay, you know the

79

answer, why you reach out to me?

80

I would say, okay, so definitely

containerize or the technology couldn't

81

drive you to the right, direction.

82

And then the team under the gm.

83

Nodding their heads

frequently and heavily.

84

Okay.

85

so then I started to

have more conversation.

86

I, I'm really new to your team, but

could you share with me how do you

87

figure out to the pinpoint, though

you, you know, under pandemic.

88

Several retail, customer se, several

retail, business getting, getting

89

gross because the online e-commerce

could be very bad, very good, right?

90

Because everybody just

broke something online.

91

However, they just shopping their,

their business was going down because

92

they infrastructure was not, good

enough to afford the last pressure.

93

But it is not the key point.

94

Because the complexity of the services

and, you know, the key wording, like

95

we always mentioned in the DDD world,

the, the, the, you know, the, the

96

model is or the big model, right?

97

Most of the team members told me.

98

The system they were built,

before, 20 years ago.

99

None of the employees really

understand how this system would

100

were built and nobody handle

and control the whole things.

101

So yeah, this is a huge challenge.

102

Nobody knows what's the content is so.

103

I, I try to engage with the clients and

tell them we don't really start from the

104

cloud journey, but maybe take a little

bit further to have an impact mapping.

105

At first, if you really want to bring

in the business value, then we have

106

to figure out who are the stakeholders

you are really want to take care.

107

So we try to identify the key passes,

identify the key concerns from

108

different stakeholders, After the impact

mapping, we have the deliverables.

109

Then just go through that, deliver

deliverables to think about, okay,

110

these items or the list might be

the potential enhancement point.

111

we can dip in deep dive

into, the potential system

112

design to see what we can do.

113

So eventually we started to have

more conversation like the code.

114

Code the series discussion.

115

They share with me that their code base,

and I mean to have a light, lightweight,

116

observation to see how the complexity is.

117

And then we found, oh, is this

impossible to resolve in a short time?

118

'cause, two many different

coding style, two, many different

119

architecture style mixing.

120

So.

121

We try to have more user story

conversation than bringing

122

some event solving practices.

123

Eventually, we figured out maybe in the

retail, selling journey, the better way

124

is to start from merchant management.

125

'cause once you decide to, to publish

your merchant on the store, then it's

126

just bringing the whole business process.

127

If the pinpoint is starting from.

128

The merchant with several different

selling policies, then we have to

129

start from the, the main point.

130

Otherwise you just try to decorating

everything, uh, on top of the, you

131

know, the weakness, the Peters,

that that was quite dangerous.

132

this is the very, uh,

interesting experiences.

133

We, we eventually have the first

draft of the, the, so code.

134

Moving out, merchant management system

from the legacy one, but move to,

135

cloud ready, microservices design.

136

But I mean, this is not a

fully integrated, but at

137

least we know how to fix it.

138

How to the conversation

with the different team.

139

still remember that day I, I held, I held

two days event filming workshop, plus one.

140

Impact mapping works up.

141

There were around 40 or 50 people

join the meeting, join the, you

142

know, the whole day workshop.

143

Several project manager, product

owner, and some sales assistants.

144

They just sit, sit in and

try to really understand.

145

How's the journey they were working

on day by day, I probably know, wow.

146

This is the first time they do have

the whole team sit down together and

147

figure out what's, what's going on

and how can they start the, journey.

148

Yeah, that's it.

149

Kenny Schwegler: Thanks.

150

So the first thing that pops to

mind is this was during Corona

151

and online, I guess, right?

152

Kim Kao: Yeah, the under pandemic

area, I try to do it offline.

153

Because, 'cause the time period,

is a little bit different.

154

In Taiwan, we, we didn't get impact

immediately, so we still, you know, people

155

wear masks and talk to each other in

156

Kenny Schwegler: Yeah.

157

because that's really

a challenge from the.

158

Kim Kao: yeah, yeah.

159

Kenny Schwegler: Another thing that came

to mind is, it started with a vendor

160

just telling, just go to Kubernetes

and it fixes everything, right?

161

Kim Kao: Yeah.

162

Kenny Schwegler: seen

that happening before.

163

Just put more hardware, to the point,

did you needed to convince your general

164

manager, or was your story enough to, to

help them understand like, oh, this is

165

someone new, this is someone different.

166

So how did you convince that?

167

Actually the, because this is

what we get a lot of questions.

168

You do two days event storming

and some impact mapping.

169

That is a, with 40 people.

170

That's a lot of investment.

171

Right.

172

And so could you explain a little

bit, how would you convince or tell

173

people to come and that it's valuable

instead of Yeah, it's easy, right?

174

Just deploy Kubernetes and let's go.

175

It fixes your problem.

176

Kim Kao: yeah.

177

That's quite interesting.

178

Um, I still remember

that GM told me, yeah.

179

They, they received the advices

from other cloud vendor.

180

They told them to.

181

The issue is try to add more hardware

like the de Bernet, but I told them, if

182

you found the lease is the only way to go

the the fastest way to go, then just go.

183

But you cannot afford it.

184

Day by day, week by-weekly.

185

'cause most of the e-commerce

company in Taiwan, they held, the

186

series of the promotion campaign.

187

And with the business growth, you

can mention that once you held the

188

company, you have more than 1 million

members active members day by day.

189

And then you have to held the

promotions maybe around the

190

three hours or four hours.

191

Wow.

192

How many activities there will be happen

there and you even cannot, have a clear

193

ma mathematical calculation to figure out

how many hardware you have to prepare.

194

That's incredible.

195

especially at that moment.

196

the retail customer, they held the

whole services, in our premises.

197

So, while you decided to buy the hardware,

you have to wait until maybe three months

198

or six months then that is impossible.

199

So.

200

Our is not the only way to go.

201

Even you can move to cloud, but still

a challenge because you have to pay

202

a lot until you start the business.

203

So we should have a deeper look on why

these solutions or why all of the existing

204

workloads, it's really suffered you.

205

What's the pinpoint?

206

If there is any complexity?

207

Combined it with some ambiguity,

you know, the ambiguity in here.

208

Not only, mentioned about the

system, but also the motivation.

209

Sometimes the marketer, they, they just

want to try to do some promotion, but

210

they may not really know what's the

return 'cause my leader told me to do.

211

Then I just do it.

212

Wow.

213

That that's incredible.

214

Kenny Schwegler: So you made it very

factual and I remember saying Yeah.

215

Compared to what and compared

to what you made very factional.

216

Right.

217

Uh, from what I hear you made very,

uh, that's a, that's a nice approach.

218

I.

219

Kim Kao: Hmm.

220

Andrew Harmel-Law: I just wanted to

ask him a little bit more about, you

221

mentioned impact mapping, and that's like,

so everyone talks about event storming.

222

There's a bunch of things we talk

about all the time, but impact mapping

223

is another powerful thing that I've

used it once, I think maybe, but

224

not as much as I should have done.

225

Can you.

226

Quickly, just kind of give us an idea

of, 'cause that's kind of the one of

227

the key points you're making, right?

228

It's like, we could do a lot of

stuff, but will it have an impact?

229

And the impact mapping is a way

to say, if we did this, what

230

do we expect the outcome to be?

231

And that helps people understand.

232

So could you maybe explain a little

bit more about, just about a little,

233

like, maybe not without the details

of what you did for this person, this

234

customer, but, um, but the, the, the

benefits that it gives to everybody.

235

'cause like Kenny says, you've had

40, you got, you did all of the hard

236

work of bringing everyone together.

237

So.

238

Kim Kao: Yeah.

239

Andrew Harmel-Law: Bringing them

together and helping them understand

240

what the impacts, what the options

were and what the impacts might

241

be, is kind of interesting.

242

So could you explain a

little bit about that?

243

Kim Kao: Yeah, thanks Andrew.

244

To raise the, the question, basically

I try to group the, the whole audience

245

into maybe two parts or three parts.

246

as for the impact mapping, I try

to invite the decision makers

247

and some functional leaders.

248

'cause impact mapping for me

is to really understand how

249

the trend really beneficial to.

250

For example, most of the time my

client reach out to me or my team.

251

They just want to resolve

the, obviously they have seen.

252

However, the symptom use, have

seen might not really be the cause.

253

So couple times I try, try to check okay.

254

If you, if you know the, the daily

transaction just turning down.

255

Is this really a issue the,

or the question behind the

256

question, what might be the really

beneficial one you want to resolve?

257

So I ask the GM to send out the, the, the

promotion team, the sales team, and the

258

engineering team leader to go with me.

259

If the goal is raise the daily transaction

amount, then what's the symptom?

260

You have figured out, and

then we have to define a goal.

261

We have the goal, if we want to raise,

for example, to bring in more revenue,

262

over amongst MOM or YOY, but, but during

pandemic, I think we can only look for

263

MOM because it may be feasible to do.

264

So we have to get a goal.

265

Okay.

266

Perhaps we want to, increase

the MOM revenue growth in

267

maybe 10 or 20 percentage.

268

Okay.

269

So who would be your supporter?

270

And things was very interesting.

271

Certainly they, they were, you know,

just like fallen into a silent space.

272

cannot figure out who are the supporters.

273

Wow.

274

That's surprise because they, they get

used to listen to the order, listen to

275

the commands to do something, however.

276

At a point, while we are using

the impact mapping approach, we

277

have to engage and encourage every

attendee to think about, you are the

278

boss, you are the decision maker.

279

You have to ask somebody to support

you because the go is company wise.

280

If everybody wants to survive,

then you have to jump in.

281

Eventually we, we identify some key

stakeholders like the, the marketing team.

282

They really understand the, the user

behaviors from their daily tracking.

283

Like even though like the

Google GA analytics could.

284

Well track the anonymous users behaviors

and then they can also, engage with

285

the, the inventory system manager.

286

They do really understand how's

the merchant, the consumption rate.

287

And then we also have to identify

that the, the key make, key decision

288

maker from inventory logistics.

289

They do understand how long it will

take to shift the per merchant.

290

So eventually write understand how is the.

291

Transition rate and, cash turn, cash,

show up rate and et cetera, et cetera.

292

So this is very interesting.

293

I would say this is not a technical

conversation, but also lead us to

294

think about how eager you are if

you really want the business growth.

295

We help jumping and, then I found I

was a businessman, not a technical guy.

296

Andrew Harmel-Law: That's really

interesting because I think the two

297

are really closely related, right?

298

You can't solve all these problems.

299

Like you said at the start, by just

throwing more hardware at it if

300

you don't understand the problem.

301

So

302

Kim Kao: Yeah.

303

Andrew Harmel-Law: go.

304

Andrea.

305

I think you were gonna.

306

Andrea Magnorsky: Yeah, so my

question was exactly that, Alex.

307

So something made you realize

that you need to understand.

308

It's a stakeholder and, understand

the domain and, and help people

309

make kind of probably quite

difficult, strategic decisions.

310

So how is this the point at which you,

for you was like, oh, I really need

311

to understand the domain, and how was

that process, if this was the time that

312

for you, everything started clicking

313

Kim Kao: Hmm.

314

Andrea Magnorsky: how was

that process of Oh, now.

315

I am not just technical,

but what does it mean?

316

Because A, I think a lot of people

that are in that position that

317

are either of close to technical

leadership or in technical leadership,

318

they find that extremely hard

to accept just knowing a lot of

319

technical stuff is not gonna cut it.

320

And it's like, it's, it's, you

need to actually do everything.

321

You need to the technical chop and.

322

You need to be able to talk to

the business in their language.

323

Kim Kao: Yeah, exactly.

324

To be honest, until today I am still

not an expertise in retail, be honest.

325

but approach I took in in the

past years in the case was

326

like, pay a lot of attention.

327

To really engage the audiences.

328

For example, a few years ago, I, I

held the DDE Taiwan annual conference

329

for years, and Kenny and Speakers had

also joined the online conference.

330

in that year I sent out

five tickets to the client.

331

I asked them to join the annual conference

to really understand how technology.

332

It'll be well utilized, along

with the business understanding.

333

'cause domain driven design

might be one of the approach.

334

I won't say this is the only

one, but I want to let them know.

335

There are several different roles, who

are really participated in DDD community.

336

So, some of the members from that

company, they just joined the, the

337

annual conference and they found wow.

338

There are several project manager,

product owner, and even UI UX designer.

339

They joined the annual community

found the topics quite interesting

340

and some of them feedback.

341

That's okay.

342

Seems layer seems like Layer should

be an approach could help us to

343

connect each one of us to have.

344

Maybe to come out the, the

language with that we could really

345

understand on the same page.

346

But at that moment, they couldn't

speak out the ubiquitous language.

347

But I know what they are, what

they, they were mentioned.

348

So I was like, okay.

349

They were fallen in the trap.

350

In the trap.

351

I said, I want to make sure

they were, they were very

352

interesting this, so eventually.

353

I have more conversation

to talk to a team.

354

I'm not the experts in retail that

I would like to share with you

355

systematical approach figure out

the connections, to figure out

356

the between the decision you made.

357

And targeted the to be

solution that you want.

358

So the conjunction role is just the

role that I played in that case.

359

Andrea Magnorsky: That makes sense.

360

I think we can kind of

relate with your journey.

361

I love engaging the audience and managing.

362

And directing people towards, their own

discovery of what can actually help them.

363

So that's, an excellent point to finish

this session of stories of, facilitating

364

software design and architecture,

which is, we keep changing the name.

365

Yay, Kim.

366

It is been wonderful to have you

and, as usual, thank you to, my

367

co-conspirators, Kenny and Andrew.

368

We'll see you next time at virtual TDD.

369

Kim Kao: Okay.

370

Thank you.

371

Andrea Magnorsky: Thank you.