Building a Next-Gen Composable Martech Stack - Niels Fogt

Justin: 00:00:14

For most of us in B2B, the marketing automation platform,

2

: 00:00:17

AKA map is the anchoring piece of

technology in the Martech stack.

3

: 00:00:22

When I was first buying a map, there

were way more options than today,

4

: 00:00:25

but generally they all bundled

a few core features together.

5

: 00:00:28

They provided a database for

your lead and activity data.

6

: 00:00:31

They provided a way to segment audiences.

7

: 00:00:33

They send emails, they give

you forms and landing pages.

8

: 00:00:36

And they provide workflow functionality,

the ability to execute if this, then

9

: 00:00:41

that style logic for various things

like your scoring, your data management,

10

: 00:00:46

your pipeline management, and so on.

11

: 00:00:48

So whether you're using Marketo or Pardot

or Eloqua or HubSpot, you're kind of used

12

: 00:00:53

to thinking of all those functions bundled

together, but does it have to be that way?

13

: 00:00:57

Is there a reason why you need

to do your lifecycle processing

14

: 00:01:00

in the same tool that you send?

15

: 00:01:03

And of course there isn't, but what

does a Martek stack look like where

16

: 00:01:07

these rules are distributed differently?

17

: 00:01:09

Today's guest has answered that

question and put it to the test.

18

: 00:01:13

Neils Fogt is senior director

of automation solutions at tray.

19

: 00:01:17

io, where he's been for almost five

years and he's built a composable stack

20

: 00:01:21

and documented exactly how it works

in a really interesting series of blog

21

: 00:01:25

posts on the And one thing I'll say

Niels is I've been thinking about these

22

: 00:01:29

problems for a lot of years, and you

really have put something together.

23

: 00:01:32

That I've thought of a few times,

but hadn't yet seen or done,

24

: 00:01:35

which is a pure workflow layer

that's abstracted from the map.

25

: 00:01:38

And you've done it in a really

thoughtful and well architected way.

26

: 00:01:41

So I'm really excited to dive

into the details with you.

27

: 00:01:44

Welcome to the show.

28

: 00:01:46

niels-fogt_1_02-20-2024_140545: Awesome.

29

: 00:01:46

Thanks for having me, Justin.

30

: 00:01:47

And just want to say a big thank

you for having me on the podcast.

31

: 00:01:50

did have a chance to listen to your

episode on the humans of Martech this

32

: 00:01:54

morning and just really appreciated

the level of depth and of insight that

33

: 00:01:59

you had as you were talking through

how we're all thinking about AI.

34

: 00:02:03

these are all really interesting

technical challenges and speak to me

35

: 00:02:06

and my technical itch when it comes to

working on these kinds of problems and

36

: 00:02:11

understanding I think that affinity

towards looking at these sorts of

37

: 00:02:14

technical problems innovating on

technology is kind of what drove me to

38

: 00:02:18

rethink how we approach the Martech stack

and Anyway, I'm happy to go wherever we

39

: 00:02:23

want to go and talk through that journey.

40

: 00:02:25

But this is definitely

a privilege to be here.

41

: 00:02:28

So thank you.

42

: 00:02:28

Track 1: Oh, thank you so much.

43

: 00:02:29

I really appreciate that.

44

: 00:02:30

And why don't we start with just

like a quick replay of your journey

45

: 00:02:34

as a martech architect, because

I think all of us make decisions.

46

: 00:02:37

In large part, based on our past

experiences, what we've seen

47

: 00:02:40

that's worked and didn't work.

48

: 00:02:41

So, before you came to Trey, were the

sort of tools that you had worked with?

49

: 00:02:45

niels-fogt_1_02-20-2024_140545:

Good question.

50

: 00:02:46

. I actually didn't start super heavily on.

51

: 00:02:50

Ops and rev ops until I got to Trey.

52

: 00:02:53

was always in and around it and always

working closely with ops people, but

53

: 00:02:57

it was not my primary profession.

54

: 00:02:59

So I actually came up in the tech

space through a user experience agency.

55

: 00:03:05

So what we would do is we would,

this was in the lean startup days.

56

: 00:03:10

If you're familiar with.

57

: 00:03:11

When the Eric Ries craze was going

on, we were working with a lot of

58

: 00:03:16

high growth startups who needed

help with user experience design on

59

: 00:03:20

their either through their product

or through their marketing website.

60

: 00:03:22

Going crazy and their needs

are constantly evolving.

61

: 00:03:25

That's the type of work I came up doing

and what I always gravitated towards that.

62

: 00:03:31

Is the technical side of it.

63

: 00:03:33

And so for example, I would

basically be like a product or

64

: 00:03:37

project manager for a UX engagement.

65

: 00:03:39

And we would have inevitably

some kind of technical piece that

66

: 00:03:42

needed to be delivered with that.

67

: 00:03:44

Let's say front end code or

website CMS or something like that.

68

: 00:03:48

And on the background, what I would

be doing is actually can I help

69

: 00:03:51

with Devin, some of that WordPress

side, or can I like, figure out

70

: 00:03:53

how to hack in tag manager here?

71

: 00:03:55

I would, always just find ways to insert

myself into the technical side and

72

: 00:03:59

self taught myself a lot of that stuff.

73

: 00:04:01

So, ended up doing was I went

in house to a company called New

74

: 00:04:04

Relic, basically doing exactly

what I was doing in an agency.

75

: 00:04:07

Setting and inside new relic, which

was a high growth dev tool startup.

76

: 00:04:11

And I got exposed to a lot

of interesting technology.

77

: 00:04:15

So at the time segment came along

and we were using analytics dot JS

78

: 00:04:20

as our data layer, so to speak for

collecting all this clickstream data.

79

: 00:04:23

And we're like, okay, well now we've got

to put it in a data warehouse, right?

80

: 00:04:26

So we got red shift, right.

81

: 00:04:28

And we started putting all that data

into Redshift and they're like, Oh, crap.

82

: 00:04:31

Now what do we do?

83

: 00:04:32

We've got to get it back out.

84

: 00:04:33

So we were doing like reverse

ETL before reverse ETL is cool.

85

: 00:04:36

I like to say and so, Redshift, analytics.

86

: 00:04:39

js, tag manager.

87

: 00:04:41

Those were like where I came up on the

data side and just kind of like, how

88

: 00:04:44

do I pull this all together and use it?

89

: 00:04:47

And then as I.

90

: 00:04:48

Come into more of a

operations role here at Trey.

91

: 00:04:51

I've gotten exposed to the classic MarTech

stack, which when I first got here was

92

: 00:04:55

Marketo and Salesforce and outreach and,

all the sales enablement side, sales

93

: 00:05:00

navigator and sales OS and all this

kind of enrichment vendors, all of it.

94

: 00:05:04

, I'm also, I'm familiar with a lot of

the traditional sort of Martech stack

95

: 00:05:08

and have a really close familiarity

with, I would say like the whole

96

: 00:05:12

data stack that's been evolving

over the last, , six or seven years.

97

: 00:05:15

Track 1: I could really relate to what you

were describing on those projects, like

98

: 00:05:19

gravitating towards the technical, because

I felt very much the same when I first

99

: 00:05:23

started getting into MarTech more broadly.

100

: 00:05:26

And I don't know what it is, but

there's just something about the.

101

: 00:05:30

Oh, this is cool.

102

: 00:05:31

Like I could, stitch these things

together and build something cool,

103

: 00:05:35

almost like playing with toys, but not to

trivialize it and describing it that way.

104

: 00:05:39

But there's something very fun

about making something work

105

: 00:05:41

and stitching it together and

having an elegant architecture.

106

: 00:05:44

So there's a certain scrappiness to

your approach that you described.

107

: 00:05:47

And even though you dent inherited

a conventional stack at Trey, you

108

: 00:05:50

were starting out, it seems like at

new relic with a lot of different.

109

: 00:05:54

So it wasn't like this scary

undiscovered territory, you were

110

: 00:05:56

already exposed to, you know, what these

alternative paradigms would look like.

111

: 00:06:00

So you came to Trey, you've got, your

standard Marketo Salesforce outreach,

112

: 00:06:04

which a lot of us can relate to.

113

: 00:06:06

What were the pain points and what

were the challenges that you were

114

: 00:06:09

experiencing with that design?

115

: 00:06:11

niels-fogt_1_02-20-2024_140545: Yeah.

116

: 00:06:11

So when I first came in, my job was

actually nothing to do with this, right?

117

: 00:06:17

So what they actually wanted me to

do was a lot of what I was doing in

118

: 00:06:21

Heroic, which was like, hey, help

us stand up a PLG motion, right?

119

: 00:06:25

We've never had a trial.

120

: 00:06:26

How do we do that?

121

: 00:06:27

And help us get our web presence up

to snuff and like that sort of stuff.

122

: 00:06:31

But again, inevitably I found

myself gravitating towards

123

: 00:06:35

those technical challenges.

124

: 00:06:36

And so I was talking to our.

125

: 00:06:38

CMO one day.

126

: 00:06:39

And he's look, like we are having all

kinds of data reliability problems, right?

127

: 00:06:43

Whether it's with not understanding what

stage our leads are at in the funnel or

128

: 00:06:49

not being able to attribute all these

programs that we're working on, right?

129

: 00:06:54

We're spending all this money.

130

: 00:06:55

And doing ungodly amounts of paid

advertising events, all the things right.

131

: 00:07:00

And I just don't feel like I can

trust what this data is telling me.

132

: 00:07:06

And yeah, we've got all this great

tooling but you still have to be

133

: 00:07:10

able to orchestrate that tooling and

make everything work and make it work

134

: 00:07:13

consistently in a reliable fashion.

135

: 00:07:15

And so I think it was like a lot

of the classic problems that young

136

: 00:07:19

marketing teams have and young

evolving organizations have, and they

137

: 00:07:23

just needed help perfecting that.

138

: 00:07:24

Now the interesting part was I didn't come

into it with a preordained mindset, right?

139

: 00:07:29

Like I didn't come into it with.

140

: 00:07:31

Okay.

141

: 00:07:31

Well, this is how you solve

that problem on Marketo.

142

: 00:07:35

I just came at it from a more, I don't

want to be snobbish about it, but

143

: 00:07:38

I a first principles like mindset.

144

: 00:07:40

It was like, okay, well, what

are the jobs to be done here?

145

: 00:07:42

What do we actually need to do?

146

: 00:07:44

And with that different perspective,

I started trying chipping away at it.

147

: 00:07:47

And I worked pretty closely with a,

died in the world marketo champion

148

: 00:07:51

who we had employed there who was

also struggling with these problems.

149

: 00:07:55

Like she too was.

150

: 00:07:56

Going I don't know what's happening when

and where, and things are overwriting, and

151

: 00:08:00

I can't seem to quite choreograph things

in the way I need it to choreograph.

152

: 00:08:05

And so we paired together and slowly, but

surely we started building capabilities.

153

: 00:08:10

And so a simple first

version was email validation.

154

: 00:08:14

Like it's really simple and silly,

but every time we get a new lead,

155

: 00:08:18

we need to conduct some email

validation post processing, right?

156

: 00:08:21

And so I built her a little service that

called clear bit risk and never bounce.

157

: 00:08:25

He did some basic syntactical

validation and a workflow.

158

: 00:08:28

And I'm like, here, just call

this through your Marketo program.

159

: 00:08:30

It'll tell you if it's good or not.

160

: 00:08:31

And we started just standing

up these little capabilities.

161

: 00:08:34

And that's how it started.

162

: 00:08:36

Track 1: What I was thinking in some

ways as you were talking was like,

163

: 00:08:39

Oh, this is really interesting.

164

: 00:08:40

Cause like you can solve a lot of those

problems with call it Martech, or map

165

: 00:08:47

native approaches or architectures.

166

: 00:08:51

Maybe the way that you're doing it is

probably better for various reasons.

167

: 00:08:54

Like actually have come around

to that conclusion, or at least

168

: 00:08:57

as a hypothesis, probably too

early to say it's a conclusion.

169

: 00:09:00

But you, Gravitated naturally.

170

: 00:09:03

I think it helps that you're

working at a, at an iPass company.

171

: 00:09:06

So you had those tools available

to you and there's a lot of

172

: 00:09:08

confidence in the tools, but

is a less conventional approach.

173

: 00:09:12

And was there anyone or any company

out there that you were inspired by,

174

: 00:09:16

or you really were just well, like

we need to put this nail into this

175

: 00:09:19

wood and this looks like a hammer.

176

: 00:09:20

And I think, you know, it looks

like a good way of doing it.

177

: 00:09:23

Let's do it that way.

178

: 00:09:24

niels-fogt_1_02-20-2024_140545: I would

say it was probably more the latter.

179

: 00:09:27

If anything, what I've been, was

inspired by was actually New Relic.

180

: 00:09:32

It mostly, I was really

exposed to developer mindsets.

181

: 00:09:36

And how you architect systems.

182

: 00:09:38

At the time we were real big into service

oriented architectures and that was right.

183

: 00:09:43

Break down the monoliths and to more

narrowly scoped services, which have

184

: 00:09:48

a very narrow function and purpose.

185

: 00:09:50

And then by doing that, you can

use them more interchangeably.

186

: 00:09:54

So just that idea was sitting there

in my head of okay, services, right?

187

: 00:09:59

Narrowly scoped.

188

: 00:10:00

I could build services on tray so

I can build a validation service.

189

: 00:10:04

I can build a MQL service.

190

: 00:10:06

I can build a routing service.

191

: 00:10:08

And so all these again, jobs to be done.

192

: 00:10:11

I was like, well, yeah, all

you need is some inputs.

193

: 00:10:14

Right.

194

: 00:10:14

And if you don't bloat it up

it'll be real easy to use.

195

: 00:10:17

. And then now they're like little

pieces that I can then choreograph.

196

: 00:10:21

And so that was really a big.

197

: 00:10:23

influence for how I just thought

about solving the problem.

198

: 00:10:26

We just kept chipping away at it, chipping

away a little bit more, a little bit

199

: 00:10:30

more, and to the point where it's wait a

minute, we just look back and we're doing

200

: 00:10:34

all this work that was previously done

elsewhere here and man does it scream

201

: 00:10:39

like it's like fast, like really fast.

202

: 00:10:42

Like from the time I see a lead until

the time that person's coming out of

203

: 00:10:45

outreach, it's like 45 seconds sometimes.

204

: 00:10:49

No more than 90 seconds

and that's pretty cool.

205

: 00:10:53

And

206

: 00:10:53

not only is it screaming fast,

but like I can, if anybody ever

207

: 00:10:57

has any complaint anywhere, like

I just go straight to the service.

208

: 00:11:00

And I can see exactly what happened when

and where and how many times it's happened

209

: 00:11:05

then go back to the process from there.

210

: 00:11:07

It just really felt

elegant and easy to manage.

211

: 00:11:11

Track 1: I love what you said

about the developer mindset.

212

: 00:11:13

And this is coming from, an English

major who fell into marketing and

213

: 00:11:18

then somehow fell into technology.

214

: 00:11:20

But I picked up a lot of that from

folks like Sanford Whiteman or Jep,

215

: 00:11:27

you also know who have kind of just

modeled a lot of those ways of thinking

216

: 00:11:31

and things like You know, don't repeat

yourself or having like loosely coupled

217

: 00:11:35

services, exactly what you said.

218

: 00:11:38

So it's kind of used to building those

within Marketo, which you can do in

219

: 00:11:41

a way like you can modularize things,

but it's not built to have the same

220

: 00:11:47

level of, let's call it like ability to

orchestrate in a defined way, or probably

221

: 00:11:53

the level of logging that you get.

222

: 00:11:55

Certainly you don't get to see

like a full JSON input and output

223

: 00:11:58

for everything the way you can.

224

: 00:11:59

I

225

: 00:11:59

think in a lot of I pass tools.

226

: 00:12:01

So where you went with it

seems really logical to me and

227

: 00:12:05

I've kind of had the thought.

228

: 00:12:06

You know, you're used to doing it one way.

229

: 00:12:08

And you don't go there.

230

: 00:12:09

niels-fogt_1_02-20-2024_140545: I'm

sure there are a lot of fantastic

231

: 00:12:12

Marketo power users out there, right?

232

: 00:12:13

You just have to know how to use the tool.

233

: 00:12:15

But one of the things I would maybe argue

is that, While things like use cases

234

: 00:12:22

around the bicycle are important and

probably, a first class challenge for

235

: 00:12:27

ops teams to solve, they're not the only

challenge where automation and a service

236

: 00:12:33

oriented architecture can be applied.

237

: 00:12:35

And as you look across there

are use cases in which this

238

: 00:12:39

core, fundamental skillset.

239

: 00:12:41

Is applicable meaning relational

systems, data models, service

240

: 00:12:47

oriented design, understanding

business logic, all that kind of stuff.

241

: 00:12:50

There are many processes which you can

apply that to that once you have that core

242

: 00:12:55

foundational skill set, make you a much.

243

: 00:12:59

Different player in the ops space, right?

244

: 00:13:02

You can impact many more parts of

the business that where they're

245

: 00:13:05

just not getting the love of

those kinds of capabilities.

246

: 00:13:08

Like a simple example is your

community experience, right?

247

: 00:13:11

So everybody's using a Slack based.

248

: 00:13:13

Community experience now where you

have this ability to have this very

249

: 00:13:17

interactive experience for your customers.

250

: 00:13:20

And the capabilities to, connect that

experience to your rev ops systems

251

: 00:13:25

or your product systems can create

some really cool things, right?

252

: 00:13:28

So like we have.

253

: 00:13:29

A very interactive slack community where

we built these various slash commands and

254

: 00:13:34

things that allow our customers to like,

check your user score progress, check your

255

: 00:13:37

education progress, apply a badge so that

all your best power users have a special

256

: 00:13:43

little annotation on their name, right?

257

: 00:13:44

And everybody knows they're the

best one at your tool, right?

258

: 00:13:47

Or , people have a lot

of workflows on trade.

259

: 00:13:49

So look up a workflow by keyword

through slash command, right?

260

: 00:13:52

Like use AI, it doesn't really matter.

261

: 00:13:54

What I'm trying to say is

there are many places in which

262

: 00:13:57

this capability can manifest.

263

: 00:13:59

And when you house it within this really

rigid frame of marketing automation,

264

: 00:14:04

then I think it limits your worldview.

265

: 00:14:07

can do a lot more.

266

: 00:14:09

Track 1: I want to zoom out for a second

and look at your stack as a whole tray

267

: 00:14:13

obviously being a critical part of it.

268

: 00:14:14

But what are the other main components

of your martech stack today?

269

: 00:14:19

niels-fogt_1_02-20-2024_140545: Like

traditional demand gen stuff, right?

270

: 00:14:21

So we have pay all the paid

social lead sources, right?

271

: 00:14:24

You have those we have all

your traditional enrichment,

272

: 00:14:28

like vendors, right?

273

: 00:14:29

So zoom info, clear bit.

274

: 00:14:30

We have event platforms.

275

: 00:14:33

We use zoom for our webinars.

276

: 00:14:35

We use a little bit of event here

and there for some bigger events.

277

: 00:14:38

We have, of course, Trey, right?

278

: 00:14:40

Acting as this orchestration layer.

279

: 00:14:42

We use iterable.

280

: 00:14:43

We use them for journey orchestration

and all the email campaigns

281

: 00:14:47

and all that kind of stuff.

282

: 00:14:48

We use iterable for that.

283

: 00:14:50

And then we have sales OS,

sales navigator, right.

284

: 00:14:54

For all the contact sourcing, that kind of

stuff and outreach for sales engagement.

285

: 00:15:00

Salesforce.

286

: 00:15:01

Track 1: I'm broadly familiar with like

kind of the iterable embraces and in my

287

: 00:15:05

mind, in this like customer journey.

288

: 00:15:08

engagement type of platforms, but in terms

of what parts of what we might think of

289

: 00:15:14

as Marketo functionality that iterable

is taking over and what parts are in

290

: 00:15:19

tray, how would you divide up those kind

of duties between the two platforms?

291

: 00:15:24

niels-fogt_1_02-20-2024_140545: centric.

292

: 00:15:25

Iterable is essentially what I call

journey orchestration, which is.

293

: 00:15:29

Most of your customer facing

communications, ? Usually there's a

294

: 00:15:33

variety of use cases for that, right?

295

: 00:15:34

So you might have a nurture journey

that you take somebody through once

296

: 00:15:37

they hit a certain status within your

CRM, or at least for us CRM you'd have

297

: 00:15:43

like fulfillment type stuff, right?

298

: 00:15:45

So you had some piece of content

that you need to fulfill.

299

: 00:15:48

Right.

300

: 00:15:48

And then you need to deliver that.

301

: 00:15:50

You might have just basic

database marketing stuff, right?

302

: 00:15:53

You have an event coming up.

303

: 00:15:54

You need to segment your

database by a certain, through a

304

: 00:15:57

certain group of people, right.

305

: 00:15:58

And get those communications out and

webinar operations, follow up, all of the

306

: 00:16:02

traditional customer facing journeys is

happening through iterable and then all

307

: 00:16:07

of the lead processing data management

is happening through Trey, meaning,

308

: 00:16:12

okay, from the moment you capture that

lead from any given source could be your

309

: 00:16:15

website, could be your chat platform,

could be your, ad vendor, whatever it is.

310

: 00:16:20

We ingest that through a Trey workflow.

311

: 00:16:22

We normalize the data.

312

: 00:16:24

So we need to then normalize it so

that it fits a common data model.

313

: 00:16:29

Apply things like lead source,

all that kind of stuff.

314

: 00:16:31

And then we run it through

what we call our lead pipeline.

315

: 00:16:34

And our lead pipeline is this sort of

amalgamation of lead processing jobs.

316

: 00:16:38

So the first thing we always do is

run it through email validation.

317

: 00:16:41

Then we run it through a lookup

process to see if we have that

318

: 00:16:44

record in our CRM already.

319

: 00:16:46

If we don't, we enrich it, create

it, lead to account match it.

320

: 00:16:50

Run our attribution process, which is

basically just stamping the dates and

321

: 00:16:54

stuff on the life cycle objects and

putting campaign memberships in place.

322

: 00:16:58

And we run it through an MQL process.

323

: 00:17:00

And the MQL process is

like, Hey, are you an MQL?

324

: 00:17:02

Yes, you are.

325

: 00:17:02

Okay, then route it.

326

: 00:17:03

Right?

327

: 00:17:04

All that sort of end to end

journey is happening on Trey.

328

: 00:17:07

And then there's always the ongoing

stuff of are we constantly making

329

: 00:17:10

sure we understand where somebody is

at in the funnel and progressing them

330

: 00:17:14

through automation so that we all, run

our reports and intelligence on, how

331

: 00:17:19

everything's working and where people

are standing, that's always up to date.

332

: 00:17:22

Track 1: I'm thinking of MQL and I'm

thinking as you're talking, all right,

333

: 00:17:25

there's that initial processing flow

and then potentially somebody comes in.

334

: 00:17:30

They're not an MQL yet.

335

: 00:17:31

Maybe you have a concept of an MQL

threshold, which is based on points or

336

: 00:17:34

an accumulation of activity, and it made

me think about where's that activity

337

: 00:17:39

layer like the piece that's tracking you

visited these web pages and then maybe Me.

338

: 00:17:44

You know, assigning some points to that.

339

: 00:17:45

Is Iterable responsible for

that, or do you have a separate

340

: 00:17:48

event collecting system?

341

: 00:17:49

niels-fogt_1_02-20-2024_140545:

Yeah, so we use analytics.

342

: 00:17:51

js.

343

: 00:17:52

You can think of it like munchkin.

344

: 00:17:53

So,

345

: 00:17:54

Segment provides analytic.

346

: 00:17:56

js.

347

: 00:17:57

So we collect all of our sort of

clickstream behavioral data through

348

: 00:18:00

segment, and that includes the

page view data, the form submit

349

: 00:18:03

events, all those sorts of things.

350

: 00:18:05

And so we can essentially listen for

those events through Trey, and then we

351

: 00:18:09

can action upon those events, whether

that be for changing data values or

352

: 00:18:15

trigger orchestrating some kind of like

internal action or external action,

353

: 00:18:20

we can just listen to those events.

354

: 00:18:22

Track 1: Are you streaming those

events also into a warehouse or

355

: 00:18:26

something, and simultaneously into Trey?

356

: 00:18:28

And

357

: 00:18:29

niels-fogt_1_02-20-2024_140545:

So segment has a native Redshift,

358

: 00:18:31

Track 1: get

359

: 00:18:32

niels-fogt_1_02-20-2024_140545: pipeline.

360

: 00:18:33

So they're constantly keeping up our

different segment sources that are

361

: 00:18:36

collecting all this behavioral data

and pushing it into Redshift and then.

362

: 00:18:40

They're all the data team will have like

jobs that sit on top of that to basically

363

: 00:18:43

amalgamated, you know, data and it's for

all form isn't always the most useful.

364

: 00:18:47

Sometimes you need to aggregate it and

create views on top of it to make it more

365

: 00:18:51

consumable for some system or process.

366

: 00:19:01

Track 1: Independent event collection

and storage, and then making those

367

: 00:19:04

events accessible to other tools.

368

: 00:19:07

There's a lot of debate around what's

that order of operations look like

369

: 00:19:12

and potential latency and, so you've

made the decision obviously to stream

370

: 00:19:16

it to, both places, which is the one

that makes the most sense to my mind.

371

: 00:19:21

Presumably you considered at least

the option perhaps of sending it to

372

: 00:19:24

Redshift and then pulling from Redshift.

373

: 00:19:26

What made you not go in that direction?

374

: 00:19:29

I guess, what were the disadvantages

if you did consider that?

375

: 00:19:32

niels-fogt_1_02-20-2024_140545: Well, for

the most part most of our needs are met

376

: 00:19:37

through this event driven architecture.

377

: 00:19:40

And we don't have like crazy

complex scoring and point systems

378

: 00:19:45

and those sorts of things.

379

: 00:19:46

Actually, we tried it and

we basically decided it was

380

: 00:19:49

this sort of like subjective.

381

: 00:19:51

Assignment of value that like, drinking

a lot of Chris Walker Kool Aid these days

382

: 00:19:57

where like he has this point of view on

your declared intent or high intent, like

383

: 00:20:04

leads being your most valuable leads

and this idea that like we have this,

384

: 00:20:09

weird subjective view of Oh, well, if you

viewed two pages and you like downloaded

385

: 00:20:15

an ebook, like that makes you qualify.

386

: 00:20:17

That's the ticket.

387

: 00:20:18

Right.

388

: 00:20:18

Maybe we're over complicating this idea

of this whole MQL in sort of a stick your

389

: 00:20:25

finger in the air kind of methodology.

390

: 00:20:28

And what we really need to do is like

optimize for a great experience on your

391

: 00:20:32

high intent leads, which can be basically

done through an event driven system.

392

: 00:20:36

Right.

393

: 00:20:36

In my opinion.

394

: 00:20:37

And then there are better ways

in which to surface the right

395

: 00:20:42

people to talk to sales and the

experience that you should give them.

396

: 00:20:45

So, in that regard, we're okay with more

latency essentially is what I'm saying.

397

: 00:20:49

Like you don't need this real time scoring

routing kind of thing for the people

398

: 00:20:54

who are not declaring their intent.

399

: 00:20:57

And so that's why we're okay

with essentially pushing

400

: 00:21:00

things to the warehouse.

401

: 00:21:01

And by the way, we're not perfect here.

402

: 00:21:03

Like we're still in this process of how

do you basically take this first party

403

: 00:21:07

data set that you have and provide sales

with a good subset of people in which

404

: 00:21:12

to prioritize and use the correct motion

with those people, because they're

405

: 00:21:19

not saying to you, I want to talk to

you right now, right here, right now.

406

: 00:21:22

And so as long as we can get the

high value leads through real quick.

407

: 00:21:27

We're okay with a little bit

of latency on the backside.

408

: 00:21:30

Track 1: So those events, it sounds like

are mainly like form fills of specific,

409

: 00:21:34

like what I would call hand raised forms.

410

: 00:21:36

And then for iterable, does it sit

on top of the warehouse or does it

411

: 00:21:41

sit on top of CRM from an integration

point of view or a combo of the two?

412

: 00:21:45

niels-fogt_1_02-20-2024_140545:

It mostly CRM, right?

413

: 00:21:46

So most of the data that we need for

segmentation purposes is written to CRM.

414

: 00:21:51

And then I built a custom sync

between, cause they don't offer an

415

: 00:21:54

out of the box Salesforce integration.

416

: 00:21:57

So I built a custom sync on tray to

basically push that data between them.

417

: 00:22:01

And then we have a handful of custom

events that basically act like program

418

: 00:22:05

member triggers So like anytime we

apply a person a campaign membership

419

: 00:22:09

to a person in salesforce we listen for

that through tray and then we emit an

420

: 00:22:14

iterable event which gives you program

membership triggers Like that's the

421

: 00:22:18

campaign membership triggering process.

422

: 00:22:21

Track 1: Let's talk about

integrations a little bit.

423

: 00:22:22

Cause one of the.

424

: 00:22:23

And I think one of the pieces that

you focused on a lot in your blog

425

: 00:22:27

post series was moving away from

native integrations and using Trey

426

: 00:22:32

as this primary integration hub.

427

: 00:22:35

And I was curious what the benefits

of this are from your point of view.

428

: 00:22:39

We all have seen some limitations to

native integrations on the plus side,

429

: 00:22:42

like they are a lot more out of box and

presumably there are a lot of things

430

: 00:22:46

like error handling and that sort of

logic kind of built in to some of these

431

: 00:22:50

integrations, at least ostensibly.

432

: 00:22:52

That, , otherwise, you have

to rule your own, which can

433

: 00:22:54

be daunting for some people.

434

: 00:22:56

What influenced that decision,

as you were building this out?

435

: 00:23:00

niels-fogt_1_02-20-2024_140545: It

mostly came from Necessity, right?

436

: 00:23:03

So we had a belief that best in

breed tooling is how we want to go.

437

: 00:23:11

We have a belief that we can own

our own process in a way and make

438

: 00:23:20

it work in a way that we want to

work by taking this approach, but

439

: 00:23:24

the truth is do have a connectivity

issue at that challenge, right?

440

: 00:23:29

Just like we were saying

with Bitterable a minute ago.

441

: 00:23:31

Like they don't offer a

native Salesforce integration.

442

: 00:23:34

So that's a problem that we have to solve.

443

: 00:23:37

So I'm not against native

integration by any means, right?

444

: 00:23:40

if it's going to do the job that it

needs to do and save me the cost of

445

: 00:23:45

ownership, please, I will take it.

446

: 00:23:48

Because we have the capacity to look at

these problems a little bit differently.

447

: 00:23:54

I don't have to put up with the

shortcomings of those integrations.

448

: 00:23:58

And so that's what generally drives it.

449

: 00:24:00

It's sure.

450

: 00:24:01

We could use that, but then we

compromise that last 10 to 15

451

: 00:24:06

percent where it's falling short.

452

: 00:24:08

Right.

453

: 00:24:08

And so rather than build around

those problems and, you know, try to

454

: 00:24:14

explain those problems to somebody

who just doesn't get it, right.

455

: 00:24:18

Like we just simply have decided

that we don't have to do that.

456

: 00:24:21

And nine times out of 10, it's not

this crazy complex integration problem.

457

: 00:24:25

It's usually just a few

small things that are.

458

: 00:24:28

Causing it to be non ideal, right?

459

: 00:24:30

When you have those hard skills around,

well, okay, I understand webhooks.

460

: 00:24:35

I understand the APIs.

461

: 00:24:36

I understand basic data

mapping and transformations.

462

: 00:24:39

You could actually close

those gaps really easy.

463

: 00:24:42

You just have to have those

hard skills to be able to do it.

464

: 00:24:45

So it's just come out of, okay,

we have those hard skills.

465

: 00:24:49

We know what the gaps we need to close.

466

: 00:24:51

Let's just close them and move on.

467

: 00:24:53

Track 1: One of the experiences

I've had building Kind of a native

468

: 00:24:58

integration equivalent process.

469

: 00:24:59

And I pass was with Trey.

470

: 00:25:01

It was a different one, but similar

sort of logic and workflows.

471

: 00:25:04

It's definitely achievable

was my observation.

472

: 00:25:07

You're a lot more experienced than me

in this domain, but you quickly realize

473

: 00:25:10

like some of the things that you take

for granted, for example, like preventing

474

: 00:25:14

infinity loops, where you have something

like this recipe or this workflow triggers

475

: 00:25:19

off an update in this system and makes it

up and that makes an update, but then this

476

: 00:25:23

workflow triggers off an update in this

system and then they start ping ponging.

477

: 00:25:26

And so then now I have

to solve that problem.

478

: 00:25:28

And you're like, oh, but what

if something is updated in

479

: 00:25:31

both places at the same time?

480

: 00:25:33

Now I have to do like field level

comparison and resolve conflicts and

481

: 00:25:37

just things that you don't really realize

are happening under the hood in a normal

482

: 00:25:40

integration, when you actually have to go

and build it out, you kind of appreciate

483

: 00:25:43

that, oh, it's doing a lot for me.

484

: 00:25:45

How have you thought about solving those

things within your architecture or have

485

: 00:25:49

you built it out in such a way that you

could sidestep some of those problems?

486

: 00:25:54

niels-fogt_1_02-20-2024_140545:

Yeah, there's not an easy answer.

487

: 00:25:55

So first of all, I would argue that

those problems you're articulating are

488

: 00:26:01

actually probably happening already

in your native integrations, and you

489

: 00:26:06

probably don't realize it, or at least

you only hear about it episodically.

490

: 00:26:12

And a lot of things are

black boxy, so to speak.

491

: 00:26:15

So, just because a native integration

does solve some of those problems,

492

: 00:26:19

doesn't mean they've solved all the

problems and it's perfect, right?

493

: 00:26:21

These are all highly relational

systems and the more things that

494

: 00:26:26

are interfacing with those systems,

be it native integrations or custom

495

: 00:26:32

integrations or whatever it is, there

is this ping pong effect happening.

496

: 00:26:35

Toe one extent or another.

497

: 00:26:37

So almost certainly everybody's dealing

with this problem, it's just more that

498

: 00:26:42

you're less aware of it when you're

not on the front lines dealing with it.

499

: 00:26:46

Track 1: We're all just sitting with

our head in the sand, blissfully unaware

500

: 00:26:49

niels-fogt_1_02-20-2024_140545: yes,

501

: 00:26:49

Track 1: of all these issues.

502

: 00:26:50

niels-fogt_1_02-20-2024_140545:

or aware, not

503

: 00:26:51

having much bliss at

504

: 00:26:52

all.

505

: 00:26:53

Track 1: Yeah, or the

opposite of blissfully.

506

: 00:26:56

Painfully unaware,

507

: 00:26:57

niels-fogt_1_02-20-2024_140545:

What I would argue is like when you

508

: 00:26:59

own some of this and you literally

have the process level visibility

509

: 00:27:05

when somebody goes, Hey, I see this

problem why did I get this link?

510

: 00:27:09

You go specifically to the metal

where you see exactly where it

511

: 00:27:12

happened, exactly when it happened.

512

: 00:27:13

And you can to that specific

situation, then you can solve it.

513

: 00:27:18

Otherwise you're beholden to

the level of observability

514

: 00:27:22

that you have on that system.

515

: 00:27:24

So I'd rather at least know where my

problems lie and that I can fix them.

516

: 00:27:30

Then not know,

517

: 00:27:33

Track 1: Do you believe in keeping a

kind of intermediary source of truth

518

: 00:27:38

between two integrated systems so that

like you kind of dump changes from one

519

: 00:27:41

system in here and dump the other ones

and then you kind of have this neutral

520

: 00:27:45

place where you can sort them out?

521

: 00:27:47

Or have you found that just like two

point to point integrations, you're able

522

: 00:27:51

to build the logic in such a way that

works an acceptable amount of the time?

523

: 00:27:55

I'm going to use the marketo integration

as an example, but my understanding

524

: 00:27:59

of how it works is it's taking all

the changes from the marketo side

525

: 00:28:02

during any particular sync interval,

and then taking all the changes from.

526

: 00:28:05

The Salesforce side and

then it's comparing them.

527

: 00:28:08

Cause if you just think from point

to point, it's kind of like blind.

528

: 00:28:10

It's I have this, I go over there and

you have this and you go over here.

529

: 00:28:12

And there's other systems out there

that I've seen that less pure I

530

: 00:28:16

pass systems, but maybe like Synchry,

for example, that maintains this

531

: 00:28:19

sort of independent, neutral data

store in the middle for integration

532

: 00:28:23

purposes, which I think makes sense.

533

: 00:28:26

Like it does answer a lot of those

challenges, or you could just use

534

: 00:28:28

a Snowflake database or a Redshift

database of your own in the middle.

535

: 00:28:32

niels-fogt_1_02-20-2024_140545: I'm sure

there are circumstances where the level

536

: 00:28:35

of complexity between two systems like

MAP and CRM and the sort of throughput.

537

: 00:28:42

Let's say going through those two systems

is sufficient that you need a much

538

: 00:28:46

more sophisticated layer of triaging

conflict and resolving some of the

539

: 00:28:51

issues that you've articulated, but in

our particular case, it hasn't become a

540

: 00:28:55

big enough of a problem and that we've

needed to resolve those conflicts.

541

: 00:28:59

And that's not to say that

means It's not a perfectly valid

542

: 00:29:03

challenge that people have, right?

543

: 00:29:05

I think is given that we are able to

kind of standard on Salesforce as our

544

: 00:29:09

source of truth and move out from there.

545

: 00:29:11

We've not hit some of

those traditional snags.

546

: 00:29:14

Track 1: That makes perfect sense.

547

: 00:29:16

I want to dive back into your

lead processing workflow,

548

: 00:29:20

and you've given a high level

overview of how it works so far.

549

: 00:29:23

Most people as we've talked about, handle

those functions in the map and in like

550

: 00:29:28

my Mariketo instance right now, I have.

551

: 00:29:30

A lot of those services that you

described built out, but they're

552

: 00:29:32

just like little brown program

icons, like enrichment, MQL, scoring,

553

: 00:29:37

works pretty well.

554

: 00:29:38

I can orchestrate them.

555

: 00:29:39

Our needs are not like

hugely complex at this stage.

556

: 00:29:42

But you've moved all

those functions into Trey.

557

: 00:29:44

You don't need to convince me of the

benefit of doing that because I've

558

: 00:29:46

seen the power that having a dedicated

workflow layer brings, but From your

559

: 00:29:51

point of view, what are you gaining

from going there that you couldn't

560

: 00:29:54

have just built into Marketo perhaps?

561

: 00:29:57

niels-fogt_1_02-20-2024_140545: Probably

a few things like I gained confidence.

562

: 00:30:01

I feel very confident using Trey, right?

563

: 00:30:03

And I can like.

564

: 00:30:05

I can absolutely tell you two

decimal places, how often any one

565

: 00:30:09

of those processes ever go awry.

566

: 00:30:12

I get notified if anything's

going wrong, instantly, basically.

567

: 00:30:16

So, to me, having that level of

comfort and confidence to know

568

: 00:30:21

exactly what happens when and where.

569

: 00:30:23

And where to go to fix it is

huge for me just from pure nuts

570

: 00:30:27

and bolts, ops perspective.

571

: 00:30:29

I'd say the other thing, if I just kind

of take a step back and think about

572

: 00:30:33

us as an organization we have a lot of

leverage in our marketing stack, right?

573

: 00:30:39

= Because.

574

: 00:30:40

I have such an ability to

evolve my process quite easily.

575

: 00:30:45

I don't have to have as much technology

in our stack, I have a lot of leverage

576

: 00:30:50

and vendor conversations, right?

577

: 00:30:51

Because it's a job to be done,

which is the process orchestrates.

578

: 00:30:55

And if I find that one vendor happens

to be better than the other, I'm not

579

: 00:30:59

constrained by all of these more rigid.

580

: 00:31:02

connections between those systems.

581

: 00:31:03

So that ability to have leverage

and flexibility and how we approach

582

: 00:31:08

our technology stack is quite nice.

583

: 00:31:10

And the other thing that

we gain is speed, right?

584

: 00:31:15

I mean, this thing is just fast, right?

585

: 00:31:17

And I love.

586

: 00:31:19

Knowing exactly how long something

takes and if I need to deal with a

587

: 00:31:22

huge amount of throughput, suddenly

I can do that and it's not a problem.

588

: 00:31:25

We can do it really fast.

589

: 00:31:26

So let's say those are

some of the main things.

590

: 00:31:28

Track 1: Make me jealous while

you're talking because I don't

591

: 00:31:31

have, what I would consider a really

robust workflow tool right now.

592

: 00:31:36

We have Zap here, but, it's

certainly not as powerful, let's say,

593

: 00:31:38

as some of the other options that

I've used in the past or for Trey.

594

: 00:31:42

And the freedom that it brings to know

when this thing happens over here,

595

: 00:31:46

that I can go over here and update this

data value, and there's this power and

596

: 00:31:50

control, as you're saying, that you have.

597

: 00:31:52

It's so liberating.

598

: 00:31:53

It's one of the reasons why I

really fell in love with that

599

: 00:31:55

category, for what it can do.

600

: 00:31:58

People that don't have that,

I think, don't know what

601

: 00:32:00

they're missing, number one.

602

: 00:32:02

But it also made me curious, as

you were describing what you can

603

: 00:32:05

do, You, I'm assuming, have a

fairly unmetered version of Tray.

604

: 00:32:10

And you mentioned at one point, I

remember when you first brought this up in

605

: 00:32:13

conversation, this design of your stack.

606

: 00:32:16

There's cost savings involved.

607

: 00:32:18

From a real world economic perspective,

if I was to become a Tray customer

608

: 00:32:22

and how did the economics shake out?

609

: 00:32:24

Because different vendors

price in different ways.

610

: 00:32:26

Some price per task,

some price per workflow.

611

: 00:32:29

Am I saving money at the end of the day,

do you think, or am I going to be forced

612

: 00:32:33

to constrain myself in a different way?

613

: 00:32:35

. niels-fogt_1_02-20-2024_140545:

So we're a consumption model,

614

: 00:32:37

meaning we build off of tasks.

615

: 00:32:39

We don't believe that a workflow

based model is the right mentality

616

: 00:32:43

because it doesn't fit with our

philosophy that you want this

617

: 00:32:47

service oriented architecture, right?

618

: 00:32:48

Like you want to be able to build

a process, whether or not that

619

: 00:32:52

process runs a couple of times a

day or a million times an hour,

620

: 00:32:57

should build.

621

: 00:32:58

A smart architecture that handles

your processes in the most efficient

622

: 00:33:03

and manageable way possible.

623

: 00:33:04

And we think that when you put like

billable constraints on things like

624

: 00:33:08

workflows, essentially people don't

solve problems, so we want to tie the

625

: 00:33:13

consumption of the platform to the cost.

626

: 00:33:16

I do think that there's true

savings that you can have,

627

: 00:33:19

cause I don't own a lean data.

628

: 00:33:20

I don't own a marketo.

629

: 00:33:21

I don't own a lot of tools that you

may have to pay a lot of money for.

630

: 00:33:26

So, what that exact number is I

don't want to pretend that I can give

631

: 00:33:31

everybody a really clean answer on it.

632

: 00:33:32

And you do have a cost of ownership

with this model, to your point, right?

633

: 00:33:35

Where you're building certain things.

634

: 00:33:37

So to me, do think there are cost

savings in there, but probably the

635

: 00:33:43

better mindset to have about it is the

level of flexibility and capabilities

636

: 00:33:49

that you're able to bring to your

organization and solve problems in

637

: 00:33:52

a much faster and efficient way,

638

: 00:33:54

so you don't have to have this

straight jacket around you where

639

: 00:33:58

you're trying to operate within the

constraints of that straight jacket.

640

: 00:34:01

Like I can build my stack in the way

I need to build it and make it work

641

: 00:34:06

the way I need it to work, which

allows me to help our organization

642

: 00:34:10

facilitate, rapid change faster,

643

: 00:34:13

we're all talking about AI now, how are

we going to get AI into our process?

644

: 00:34:17

And we're worried about governance

and we're worried about security

645

: 00:34:19

and we're worried about, how do we

use our first party data and these

646

: 00:34:22

models and all this kind of stuff.

647

: 00:34:24

Right.

648

: 00:34:24

Well, I would argue that when you have

a capability like this to build your

649

: 00:34:29

processes in a way, then you can take

new capabilities like AI and you can bring

650

: 00:34:35

them into your process much more easily.

651

: 00:34:37

And so to me, indexing on that agility and

the ability to do what have your stack, do

652

: 00:34:43

what you need to do for your organization

is just as important as saving, you know,

653

: 00:34:47

five figures on your vendors this quarter.

654

: 00:34:50

Track 1: You mentioned the humans

of Martek podcast and my friends,

655

: 00:34:53

Phil and John over there, they've

talked a lot in some of their

656

: 00:34:56

episodes about, composability

and creating a composable CDP.

657

: 00:35:00

And it's one of those

words that's very much.

658

: 00:35:03

In Vogue right now.

659

: 00:35:05

Were you thinking about composability

when you were designing this?

660

: 00:35:08

Or do you see your stack as

falling under this label?

661

: 00:35:11

Or so happened that you've

approached it in that way?

662

: 00:35:14

niels-fogt_1_02-20-2024_140545:

It's absolutely a composable stack.

663

: 00:35:16

If it doesn't meet the definition, then

I'm not sure what does . The whole idea.

664

: 00:35:21

Is that you can take these core services

or jobs to be done and orchestrate

665

: 00:35:27

and use them in the way that you

need to this idea of abstractions.

666

: 00:35:30

It's nothing new per se, we were

talking at the start of the call about.

667

: 00:35:35

Analytics.

668

: 00:35:35

js or segment, right?

669

: 00:35:37

Like they came in with an abstraction

layer to instrumentation so that we can

670

: 00:35:41

move event data to either point solutions

or your data warehouse quite easily and

671

: 00:35:47

not have to go through that process for

every new tool we bring into the stack,

672

: 00:35:50

This idea of abstraction has been.

673

: 00:35:53

Going on for quite a while.

674

: 00:35:55

To me, what we're now getting into

is kind of had this initial sort

675

: 00:35:59

of data layer abstraction happening.

676

: 00:36:01

And you have in the sort of

cloud data warehouse, right?

677

: 00:36:03

Consuming that and acting like its

own abstraction layer your workflow.

678

: 00:36:07

Version of this is just another version

of abstraction and composability right

679

: 00:36:12

where we're now you're creating these

capabilities that can be just like

680

: 00:36:17

those other tools they can be used in

a fashion that gives you The flexibility

681

: 00:36:22

to decide when and how they're used.

682

: 00:36:24

Track 1: It seems that there

is an increasing trend towards

683

: 00:36:27

companies designing these sorts of

point tools microservices and they

684

: 00:36:32

all kind of come up in that way.

685

: 00:36:33

And at the same time, though, there

is an increasing economic pressure

686

: 00:36:37

in the marketplace for companies

to consolidate functionality.

687

: 00:36:41

And I think we see this in lot

of the sales engagement space

688

: 00:36:44

where we have the gong and the.

689

: 00:36:46

Sales loft and, you know, various

tools like that starting to try

690

: 00:36:49

to eat each other's lunch from

various angles and acquiring

691

: 00:36:54

tools and bringing them together.

692

: 00:36:57

And, , there's reasons that it's

easier for budgetary purposes

693

: 00:37:00

for companies to do that.

694

: 00:37:01

Or you have the HubSpot is maybe

the perfect example where they're

695

: 00:37:04

like, we'll just build one of

everything, we're just going to build

696

: 00:37:06

it and it's all going to be one.

697

: 00:37:08

Huge monolithic stack.

698

: 00:37:09

Do you see both of

these models coexisting?

699

: 00:37:11

you see the future as one in where there's

just like this endless diversity of point

700

: 00:37:15

services and then orchestration tools that

bring them all in these wonderful ways?

701

: 00:37:19

What does the future look

like from your point of view?

702

: 00:37:22

niels-fogt_1_02-20-2024_140545:

I don't think it's a endless sort

703

: 00:37:24

of amalgamation of point services.

704

: 00:37:26

At some point, there is going to be

some benefits to taking certain more,

705

: 00:37:33

let's call it more broad platforms

and Leveraging whatever capabilities

706

: 00:37:38

those platforms have if it makes a

lot of sense for your business and

707

: 00:37:41

it does what it needs to do, and

there's less overhead for you, right.

708

: 00:37:45

And running that model, then

run that model that is okay.

709

: 00:37:49

There's nothing wrong with using a

really smart set of tools to do the

710

: 00:37:55

core jobs to be done that need to

be done, like we don't need to just.

711

: 00:37:59

For the sake of composability, right?

712

: 00:38:02

To develop a composable stack.

713

: 00:38:03

That's definitely not

the mindset I would take.

714

: 00:38:06

I think it's just more so that

having the ability within your

715

: 00:38:12

four walls of your organization to

cross that line when you need to.

716

: 00:38:16

Is a capability that we should be

prioritizing for our ops functions

717

: 00:38:20

. Meaning this is a critical skill set

for operations teams to have within

718

: 00:38:27

house, so that when that platform

doesn't do what you need it to do.

719

: 00:38:33

You can make the changes that you

need to make and are not beholden

720

: 00:38:38

to the constraints of that system.

721

: 00:38:40

And, That is where you use it, right?

722

: 00:38:42

And that's a always evolving thing.

723

: 00:38:45

Our stacks are constantly evolving.

724

: 00:38:47

The vendors of today are going to get

beat by the vendors of tomorrow, right?

725

: 00:38:52

And we want the ability to be

able to move with that change.

726

: 00:38:56

And so in order to do that, you have to

have these competencies in house, because

727

: 00:39:00

otherwise don't control your destiny.

728

: 00:39:03

Track 1: So speaking of the ever

evolving state of things, . What are

729

: 00:39:07

the, to do list that you're like,

yeah, I got to fix that someday.

730

: 00:39:11

Are there still some rough

edges here and there?

731

: 00:39:14

niels-fogt_1_02-20-2024_140545: One of

the harder things is what we've been

732

: 00:39:15

talking about is like, do we Make sure

that we can continue to run this practice.

733

: 00:39:22

So one of the harder things is just

teaching people the hard skills, right?

734

: 00:39:24

Like I've learned that there's

a certain profile of operator.

735

: 00:39:27

I need to have underneath our roof who

has the right sort of skill sets to

736

: 00:39:33

be able to run this sort of practice.

737

: 00:39:35

So early on.

738

: 00:39:37

For example I was probably not looking

at the right types of individuals

739

: 00:39:41

when the right types of skill sets

because is a more technical challenge.

740

: 00:39:44

Whereas maybe you may have brought

in more junior practitioners, for

741

: 00:39:49

example, as part of an ops function,

you can't execute this model with a

742

: 00:39:52

junior practitioner who doesn't have

a good systems architecture mindset.

743

: 00:39:56

So that's been a challenge that I've

learned just like, how do you find the

744

: 00:39:59

right skill sets to run the approach.

745

: 00:40:01

And then as far as areas for evolving

the system I don't have this major

746

: 00:40:07

thing that stands out where it's

just Oh, we absolutely have to fix.

747

: 00:40:12

I don't know, our lead scoring

approach or something like that.

748

: 00:40:15

I've been running it for

probably three and a half years

749

: 00:40:17

now . And so it's pretty mature.

750

: 00:40:19

So what we're actually more so

thinking about is, what is next in our

751

: 00:40:25

Inside of our house of capabilities.

752

: 00:40:28

And so we're doing the

trendy stuff right now.

753

: 00:40:29

Like, how do we actually understand

where we can apply AI in, inside of

754

: 00:40:35

our organization to do like the things

you guys were talking about on humans

755

: 00:40:39

and MarTech, like how do we take

less structured data and what are the

756

: 00:40:43

technical underpinnings that we need

to have in place to be able to consume

757

: 00:40:46

that and make it operationally useful.

758

: 00:40:49

For our organization, kind of skill

sets in services do we need to have in

759

: 00:40:52

place to do that in a way that security

is also going to be like, that's cool.

760

: 00:40:55

You're allowed to do that.

761

: 00:40:56

And it's also useful and

functional for the business.

762

: 00:41:00

So one thing we've been doing right now

is building a service around our all

763

: 00:41:04

of our sort of knowledge content of how

we consume that Put it into a vector

764

: 00:41:08

database, create a service that, we can

then use with our knowledge content and

765

: 00:41:14

then use that for both internal enablement

use cases and customer facing use cases.

766

: 00:41:20

Track 1: Chat GPT that's

trained on your own docs or

767

: 00:41:23

your own data kind of scenario.

768

: 00:41:24

Maybe last question I'll ask

is in terms of the evolution of

769

: 00:41:29

this and how you evolve it day

to day, what's your relationship

770

: 00:41:33

to your business stakeholders?

771

: 00:41:34

Is it you coming up with

ideas or presenting them?

772

: 00:41:37

Are they coming to you with requests?

773

: 00:41:39

Is it more of a product management style?

774

: 00:41:41

How do you triage and

roadmap these things?

775

: 00:41:45

niels-fogt_1_02-20-2024_140545: Pretty

much like any classic ops team, right?

776

: 00:41:47

So like I am a business partner to

other parts of the organization.

777

: 00:41:51

I report to our CMO.

778

: 00:41:53

And so, a lot of what we're doing

right now is just ensuring that we're

779

: 00:41:58

facilitating the go to market strategy

that he's laid out for the organization.

780

: 00:42:02

And we were in a as an organization, a

bit of a transformational period, where

781

: 00:42:06

we're shifting our go to market strategy

a fair amount, which requires that you

782

: 00:42:10

got to reevaluate all the things that were

tied to the old go to market strategy.

783

: 00:42:15

And so just like any other ops team,

like we'll hear the challenges that

784

: 00:42:20

he is putting forth for us to change,

and then we create a roadmap and we

785

: 00:42:24

prioritize that we work with our same

thing with our sales leaders, right.

786

: 00:42:27

When I go to market strategy evolves, the

sales organization evolves, you get things

787

: 00:42:31

like new territories and new segmentations

your customer base and all that sort

788

: 00:42:36

of stuff, I don't think we're doing

anything all that revolutionary there.

789

: 00:42:41

We just have to be, tied to what the

business is trying to do and creating a

790

: 00:42:45

set of priorities, determining what we

can do easy and as impactful and making

791

: 00:42:50

those changes as quickly as we can.

792

: 00:42:52

Track 1: Well, they're lucky to have you.

793

: 00:42:53

And I love how you think about this

stuff it sounds like in many ways

794

: 00:42:56

you've built something really awesome.

795

: 00:42:58

So I'm glad.

796

: 00:43:00

That we could just talk about it and

expose more people to the way that you've

797

: 00:43:03

thought about your stack I will include

links to your blog post series and the

798

: 00:43:08

show notes because you go into Really

high level of detail on what you've done

799

: 00:43:12

You have videos if you're listening to

all this To those of you out there and

800

: 00:43:16

thinking like oh, I just really should

really see and understand more What

801

: 00:43:20

Niels is talking about you can because

he's gone over it with screen share.

802

: 00:43:22

So we'll share all that But yeah, thank

you so much for coming on the show.

803

: 00:43:26

It's been a ton of fun learning about

804

: 00:43:28

niels-fogt_1_02-20-2024_140545: Awesome.

805

: 00:43:28

Yeah.

806

: 00:43:28

Thanks for having me.