What to do with extra marketing budget

What is Startup to Last?

Two founders talk about how to build software businesses that are meant to last. Each episode includes a deep dive into a different topic related to starting, growing, and sustaining a healthy business.

Tyler (00:01)
In this episode, we talk about some newly found money and less annoying CRM's marketing budget. Well let's go.

Rick (00:25)
What's up this week, Tyler?

Tyler (00:26)
Yo, Rick. I see you've got actual stuff to talk about this week. Y you you no normally we

Rick (00:31)
what is that supposed to mean?

Tyler (00:35)
open up the notion board, you show up, and you spend fifteen seconds typing some things into the notion board before I hit record. You actually you like really prepped some topics today.

Rick (00:46)
It's what does that mean? That means my life is settling, I guess. it's well, I'm actually sick this week, which is surprising. So the whole like

Tyler (00:50)
Yeah. What's what's going on? What's wrong?

Rick (00:56)
Sable and I are both sick at the same time, and anytime we're both sick at the same time, it's like, who's sicker? They get to lay down, you know?

Tyler (01:02)
Yeah. yeah.

Rick (01:04)
And I've not this I think Sable's sicker this time, so I'm

Tyler (01:07)
Mm.

Rick (01:07)
I'm the one like powering through. but yeah, it's I'm not a napper. I wish I was when I was

Tyler (01:11)
I think you said you're not a napper, right? In general. Shelly and

I, our whole relationship is based on trading off naps. Like in a on a Saturday, there's a good chance we're both taking two naps alternating while the other watches Sydney.

Rick (01:26)
That's hilarious. No, I don't

take naps, so Sable take but Sable du it's usually

Tyler (01:31)
She does actually.

Rick (01:32)
I my aut my alterate alternate alternating key sort of sequence with Sable is I go exercise, she and then she goes naps, I go exercise, she goes nap.

Tyler (01:42)
Man,

you are really getting the worst end of that deal, but okay.

Rick (01:45)
I don't know.

Tyler (01:47)
anyway, yeah, what's going on?

Rick (01:50)
well, Leg Up had a record month in August. We hit 30k in revenue, which is pretty cool for the month. and then s another thing that I wanted to share was I don't think you're gonna be able to join us, but JD and I booked a trip. I'm gonna go visit him in Texas in the beginning of October in Austin. And so it was kind of cool. this is the first time where we booked an annual trip where I didn't even think about

How much it cost.

Tyler (02:20)
Mm-hmm.

Rick (02:21)
You know, I didn't like usually I'm like, gosh, should I fly coach? You know, like, you know, like exact exactly.

Tyler (02:25)
Yeah, are we sharing a hotel room?

Rick (02:28)
And this time I was just like, it's not even a worry. I'm just gonna

Tyler (02:31)
Yeah.

Rick (02:32)
book book a nice ho resort and you know, not worry about like pick the right flight and pick a flight based on what times are convenient to maximize the time, not necessarily like saving a couple hundred bucks. And so I don't know, that's kind of a a milestone worth worth recognizing, I think.

Tyler (02:48)
Definitely.

It's good to have revenue.

Rick (02:50)
Yeah.

Tyler (02:53)
cool. What do do you do know what you guys are doing there? You said a resort, so you so this is like outside the the city?

Rick (02:58)
It's yeah, so

we were we looked like four or five different places, all about the same, you know, sort of nightly rate. And this one looked the best because it was outside of town and had like everything you need, twenty-five minutes outside of town, like golf course, bunch of pools, resort more of a res resort feel. and we figured we'd probably go into the city for, you know, the evening, you know, one one night. So that that way we can just hang out and take advantage and and relax.

I don't know. Maybe we'll maybe we'll wish we had we had stayed downtown.

Tyler (03:31)
Yeah, well, no, that sounds great. I mean my friends and I, like my high school friends and I would joke back like in our twenties, we would travel somewhere cool together, friend friend trip, and we would just hang out in the hotel room drinking beer the whole time. And so our joke became we might as well just go to Sioux Falls every single time. I don't know why we're picking on Sioux Falls, but it doesn't matter

Rick (03:52)
Sue Fa Sioux

sounds amazing. Like that like that's a ma like I don't even know what it is. Is that in like the Dakotas

Tyler (03:57)
That's in

Rick (03:58)
somewhere?

Tyler (03:58)
yeah, North North Dakota, I think. I I think one of my friends just picked out of like, you know, picked out of a hat. What's the most random, like nondescript town you could go visit? That probably no one's like a tourist there, you know.

Rick (04:09)
Yeah, I see.

Tyler (04:12)
sorry. Sorry, to our one Sioux Falls list Yeah, yeah.

Rick (04:12)
I mean, you should just stay in St. Louis. Just get a hotel. Like get

the four the four seasons in St. Louis has great rates.

In nice rooms.

Tyler (04:23)
these were not

these were not really four seasons types of trips, but yeah. These were like can we pull together a couple can we steal some tables from the Holiday and Express Business Center and play beer pong on them? That type of thing.

Rick (04:34)
Yeah. That

reminds me of this one time I went to visit some college buddies in San Francisco when we were in our twenties and one guy was responsible for getting a hotel and I didn't even know these types of hotels existed, but he got us for Ford guys, he got us a hotel room, one hotel room to share, and it had one twin bed.

Tyler (04:56)
Whoa.

Rick (04:56)
Yes. So we had like there were two people in the bathtub, two people like, you know, in the twin bed. it was bad.

Tyler (05:05)
Hmm. I've never I've never slummed it quite like that, but yeah, okay. Yeah, yeah.

Rick (05:09)
Yeah, me w that was a one time experience for me. He d he he

lost he he his his booking responsibilities were were revoked.

Tyler (05:17)
Yes. cool. Okay, so you're going to Austin, had a record month. Is a record month I know sometimes with leg up the the money comes in in kind of lopsided ways, just like one month you make more and one month you make less. Is that what's going on here? Or is this like s slow, steady growth?

Rick (05:29)
No. Yeah, it's a it's a surge.

it's it's probably it's indicative of like sort of like a a a jagged trend line up. but you know most money we've ever collected in a month, for sure.

Tyler (05:43)
Yeah.

But it is still growing. I my impression is not a lot of new sales, but the the like the expansion revenue is great and so it the business keeps growing. Yeah.

Rick (05:50)
Yeah. Exactly.

Yeah. We still gotta figure out how to bring on more new employers. and that's that's the main sort of proactive focus that we have right now. And I I imagine that's there are two things we'll probably talk about during the the trip. One is, you know, that how do we get more customers? And the second is you know, how do we taking a look at JD and how he's spending his time.

And seeing where there's opportunities for him to get more leverage ahead of our busy season in open.

Tyler (06:25)
Yeah, cool. that's

Rick (06:28)
That's that's what what's going on with you?

Tyler (06:30)
like a pill. Yeah, I I don't have any like super meaty topics, but one thing I thought I'd do, I I was doing a little looking through some numbers. I feel like I'm doing this more and more these days because AI makes it so much easier to like write a super complicated database query that gives me the exact answer that I want. but so I figured I'd just share some of the things I uncovered. Th well, actually this one wasn't even digging through the database.

I have an embarrassing thing to admit here. so a couple years ago, paid advertising kind of shifted from one person on the team to another. And in that transition, some of our terminology got mixed up. And when we were talking about price per user, it went from price per paying user to price per free trial user.

So, like, what do we have to pay in ad spend to get a paying user? Is what I thought we were talking about. And the actual metric I was getting data on was what do we pay to get a free trial user? Which converts

Rick (07:31)
Mm.

Tyler (07:31)
at like much worse rate than our like an organic, like someone who comes to our site deliberately converts at about 25%. An advertising user converts more like 10 to 15%. So that means like the price per free trial user is something like 10x.

Or rather, the price per paying user is something like 10x the price per free trial user, meaning I was 10x off what I thought our ROI was for our ad spend. very embarrassing. That's a big problem.

Rick (08:05)
So but is this how much money are you spending a year?

Tyler (08:08)
Well now, none. We just killed it all. But yeah, I

Rick (08:10)
Okay. So wait, so should

Tyler (08:12)
just cut $150,000 a year of marketing spend.

Rick (08:16)
So you're sp you were spending $150,000 a year on Google search and like what like what what what were you spending?

Tyler (08:20)
Primarily Captera.

Rick (08:22)
Okay.

Tyler (08:23)
We we didn't actually go to zero, but we we we took our Captera spend, which are their was our biggest channel, to zero. For people who don't know, Captera is like a kind of like an app review for business software. G2 is they just got acquired by G2, actually. They were previously

Rick (08:38)
Mm.

Tyler (08:38)
owned by Gartner. but Captera is a fucking scam to begin with. You go there and they're like, here's our list of CRMs. That list is literally just who pays us the most is at the top.

There's they have they have all these reviews. We have hundreds of reviews on Captera. That that does not factor into the algorithm at all. It's purely just who pays the most. there was a time when Captera was actually providing pretty good ROI for us, but it's just gotten worse and worse and worse over the years. I mean every channel has.

Rick (09:07)
Mm-hmm.

Tyler (09:08)
But yeah, I the the numbers are when I finally got, you know, dove into and did the math myself, it was like, this is atrocious. So

The silver lining is now we have a hundred and fifty thousand extra dollars, either to play with to spend on something else or just to save, but yeah.

Rick (09:26)
Can we put this in context

of like your overall like ad budget? Like, is this a subset of your total ad budget or is this the ad budget you're talking about?

Tyler (09:35)
I would say the total ad budget is maybe twenty K a year and a hundred and fifty of it or or two hundred, sorry. this is about seventy five percent. I I'm I'm just estimating, but

Rick (09:44)
All right.

Tyler (09:44)
something like seventy five percent of the total ad budget.

Rick (09:47)
So you have like Google search. I'm so assuming if someone types in less annoying CRM, you you pay to make sure you're number one on that.

Tyler (09:51)
Yeah, we're d well this we're examining

this too right now. Yes. We are we we're doing Google AdWords only for branded keywords, which is what you're talking about. we

Rick (09:57)
Okay. Yep.

Tyler (09:58)
used to do it for all kinds of keywords and that stopped working years ago.

Rick (10:03)
Have

you tried like simple CRM and that kind of stuff?

Tyler (10:05)
Yeah. I mean that's our whole business is built on AdWords for the simple CRM is the keyword that we always had success with. But now, I mean we have to pay three thousand dollars to get a paying fifteen dollar a month user. I mean, the the math just does not work out.

Rick (10:23)
Crazy.

Tyler (10:24)
It's rough out there. I it's a combination. Everything is more competitive, but when you also

Rick (10:27)
You're already number one for simple

CRM on an organic basis.

Tyler (10:32)
I don't think that's always true, but yeah, we we do get some organic traffic for simple CR.

Rick (10:35)
no, okay, no,

you're paying for simple CRM. I see.

Tyler (10:38)
we are?

Rick (10:38)
Yeah, you are. You come up number one sponsored result as no, you're number one for no, no, no,

Tyler (10:43)
Well now I gotta look into that, huh?

Rick (10:45)
you are if I just typed in simple CRM and you were the number one result on simple CRM and Google organically. I'm doing it in I incognito

Tyler (10:51)
Huh. It's

Rick (10:53)
now and and you still are number one if I hide all the sponsored results. Yep.

Tyler (10:56)
Really? Man,

you'd think that would be more valuable than it is, huh?

Rick (11:00)
Maybe it is. Like do you do you have a Hub Spot set up to track the Okay.

Tyler (11:05)
I don't we use refs.

I I don't think we are actually number one most of the time. You know, these things fluctuate. I'm I'm excited to hear that

Rick (11:12)
Mm.

Tyler (11:13)
but I I was just I was just in Google Webmaster whatever the Search Console, whatever they call it now. we do get some some organic traffic from that keyword, but it's like a hundred clicks a month or something. It's not it's not hyper.

Rick (11:28)
All right, sorry, I I got us down a rabbit hole. your

Tyler (11:30)
Ha ha.

Rick (11:30)
ad budget, you sp you spend on branded keywords, you've been paying for ads in sort of these directories, you're cutting the directory ads.

Tyler (11:39)
Yeah. we still pay G two. That's not like for ads exactly. It's it's kind of a package of we're eligible for certain awards and we get to use their logo on things. so it's it's not like a performance marketing thing exactly.

Rick (11:55)
I like that kind of market. Like that's where I was going with this budget. Like how do we like what I'm trying to set up is like the conversation around how do you spend 150K? I

Tyler (12:02)
Yeah.

Rick (12:03)
I like the idea of like more brand recognition, more brand sort of affinity, that type of spend with the right sort

Tyler (12:09)
Yeah.

Rick (12:10)
of audience.

Tyler (12:11)
Right. So let's brainstorm this for a second. Like what does that look like? And let me just give the context. So most of our ad spend historically has been classic PPC online ads. So Google AdWords or in a directory or a website, like CRM.org is a website about CRMs we advertise on there. That's one category. We've that's not the brand stuff you're talking about. More recently I've been talking about the health insurance stuff a lot, which is

With that one, obviously we're spending on going to conferences, we're trying to get webinars going with FMOs, but we're also doing some retargeting ads where basically once we get someone's information, whether it's through a conference or however, then we put them on a re then they'll start seeing our display ads. But we're not saying we want every health insurance agent in America to see our ad on Instagram every time they log in. Like what do you have in mind when you say more branding?

Rick (13:08)
man.

well, obviously someone who's on G2 looking at CRMs is so so I guess like when I say brand branding, it's paying more for to have your brand present versus like a paying for a lead. Like that's the I guess the fun

Tyler (13:27)
Yeah.

Rick (13:28)
the the definition difference of what I'm thinking. And examples of this are billboard advertisements, magazine advertisements, I'm like display advertising that is more like

Tyler (13:37)
Yeah, yeah. Display ads.

Rick (13:40)
Yeah, it's more like where is the ad showing up is more of the question, not like what's the conversion rate on the ad? Who whose eyeballs it's like CPM.

Tyler (13:49)
Yeah, absolutely.

Rick (13:50)
It's like who what what are the eyeballs, number of eyeballs seeing the ad and who are what confidence do we have it if they're that they are the right eyeballs? More so than like, is it passing a lead?

Tyler (14:00)
Yeah, that's the hard thing, of course.

Yeah, so I'm interested in this, and we're doing display ads for those retargeted health insurance people, because that's a case where you can like, we know the right people are seeing them. My understanding of display ads in general is good for like Coca Cola and Nike because everyone drinks coke, everyone wears shoes. It's harder to make it work for like the vast, vast, vast majority of people are not CRM buyers, and even those that are we're only a small, small part of the CRM market.

How do you get display ads in front of the right people? At the right time, yeah. Or or maybe

Rick (14:32)
At the right time. Yeah, the right t I think timing

Tyler (14:35)
I maybe not. Like this is one nice thing about display ads versus like PPC type ads. Is a display ad is saying if they just see our name seven times and then a year later they need a CRM, that that that might work out. Rather than like when you see an ad in Google, if you don't click it right that moment, it's very unlikely you even noticed what the name of the company was or something.

Rick (14:57)
Yeah, yeah.

Yeah, I see what you're saying. So you're you're you're like saying timing is sort of a secondary optimization, not the primary optimization. yeah.

Tyler (15:03)
I think so. Or yeah, you'd you'd rather have both for sure, but

Rick (15:07)
I don't know. Yeah. So

Yeah, it seems like if you had a good aud like sort of master audience that you could say, hey, I don't w you could use for like almost suppression against all your visitors that you don't want to retarget. so like if if you say like there's all these people that are coming to your website, some subset of them meet your ICP and you want to target with display advertising. and there's probably various like some sort of engagement score, like okay, that what

Check one is do they meet the ICP? And then the second is like, how likely are they to buy based on where they are in the journey? Are they just checking out a blog post or are they active in a free trial? And then having some sort of like campaign sort of prioritization where you exhaust the ability to retarget to the people in the free trial until they, you know, and then you work down in terms of engagement. That's probably how I would think about it.

Tyler (16:02)
Yeah. Okay. And almost all

of our website traffic, I mean, your your your intuition is correct that the the vast majority of our website traffic is current customers just trying to log in. So yeah, you gotta you

Rick (16:12)
You'd suppress all you suppress all them.

Tyler (16:14)
gotta suppress that. But of the remaining traffic, it's almost all people just hitting our homepage. we don't have a lot of like like we have a lot of content, but that content's not getting direct traffic for the most part. so pretty much everyone, it's like

You you hit our homepage and then maybe you signed up for a free trial. There's kind of two levels, I feel like, versus a HubSpot type model where it's like, well, how many of our articles have they read and did they ever download a white paper and stuff like that? but so basically what I'm hearing is retarget people who hit our homepage, excluding the people who logged in.

Rick (16:53)
I think so. assuming that you have an audien but but I would say through the lens of an audience eligibility that you've have confidence in. So like, meaning like if there's probably a single I don't know, if there's a way for you to say this is likely to be a three user account you know, versus a one user account,

Tyler (17:15)
yeah, we don't we don't have anything like that.

Rick (17:18)
yeah, like I don't know, I just don't know what audience capabilities there are out there in terms of

Tyler (17:22)
yeah. Do you do you have any experience with look alike audiences? Do you

Rick (17:27)
Yeah, this is what Windfall does.

Tyler (17:29)
do you trust that?

Rick (17:31)
Well, Winfall's concept is you shouldn't trust it. it's it's good enough in the platforms to do

Tyler (17:36)
Just measure it.

Rick (17:37)
the but no, build your own model. And so what what what kind of Winfall's idea is like build your own predictive model to identify people in your database or in the in the universe of people, and then pick your build build the audience based on your model and then deploy that audience.

Tyler (17:56)
Okay, that's fine if you're a company like Windfall that you're in the business of having a knowing who every person on Earth is.

Rick (18:03)
There are companies that do this. So like Windfall is the consumer version of this, but there is a a business version of Windfall out there that you could go build an audience with and probably build a predictive model with and have them deploy that audience for you. And then you use that audience

Tyler (18:15)
Okay.

Rick (18:17)
to to build off of within the platforms.

Tyler (18:19)
Gotcha. Cause a normal I mean, there are basically two ad platforms in the world now. It's Google and Meta. And you know, what most companies are doing is just using their tools. Like they they both have look alike products where you can say, anyone like I want to retarget anyone who hits my website, but I also want to send ads to anyone that is similar to people hitting our website but that haven't hit our website yet. Yeah, I I I'm I don't think I trust that those are

That that works, but I guess you can. Yeah, they both like golf. That's not yeah, exactly. Like,

Rick (18:50)
Yeah, j similar in what way, right? Like, they're a male at forty years old lives in Ohio.

Tyler (18:57)
is it smart enough to be like, no, it the of the hundred characteristics we know about this person, them being a health insurance agent is the one that matters, not the other ninety nine.

Rick (19:03)
I think I think there are professional

look-alike sort of audience capabilities within the platforms. One, this isn't a platform question. This is an audience question. And you really want a unified audience across all the platforms. So the the you know, you want to have like this audience that that sort of you you feel confident in, and then pick your platforms that you want to display against. like if they're on if if your ideal customer is watching golf.

On TV, you want a less annoying CRM ad to show up for them, right? For 30 seconds. and so

Tyler (19:39)
That's that's

a big company play, like

Rick (19:42)
I don't know. You're I like it's it's it's much more digital marketing is much more capable now. and you might be willing to spend that that might be a better use than paying on Captera if for the right person. So

Tyler (19:56)
Yeah.

Rick (19:58)
the the second thing that that Winfall solves, and I'm assuming that there is a another company out there that would solve this for you, is more granular targeting around the characteristics that our customers care about. And

Wealth is I the obvious one.

Tyler (20:11)
So how does this work? They

they give you basically a list of email addresses and then you upload it to Google and Meta and no. How does it work?

Rick (20:19)
It's there's so

okay. So this is what I've okay. So there's a company called LiveRamp. Have you heard of LiveRamp? it is a audience platform. Basically, they what they do is they help if you if you have just say a list of people, offline data, like let's just say you just download the white the white pages and you upload that into LiveRamp, they're going they have like mechanisms to match that offline data.

to online digital identities through the the platforms. And so there are basically you can take a list of any it doesn't even need to be an email address, it could be a street address, P I I. P I I. So basically you

Tyler (20:56)
Yeah okay, yeah. Any identifier, sure. Yeah. But how does that

get into the ad platform?

Rick (21:02)
live ramp or direct direct there there are live ramp is the main player but and then there are lots of like direct sort of deployment integrations out there now like windfall can direct deployment audience for our customers to Google.

Tyler (21:16)
Okay. So I'm still

logging into Meta and saying I want to run an ad campaign and saying who do you want to advertise to? And then I say connect it to

Rick (21:21)
You pull a th

con I wanna pull a third party audience from

Tyler (21:25)
Gotcha. Okay. It's interesting.

Rick (21:29)
Yes, and I I think ad role plays in this for like a B2B perspective. Like I think I think one option would be yeah. So

Tyler (21:34)
Okay, we're using them for for retargeting right now, so

Rick (21:37)
I think I think you can th I think they have an A B product. I don't know if it's called I don't know it used to be called next role, but I think you can go you can build there should be some sort of audience capabilities in there that you can deploy across multiple platforms.

Tyler (21:51)
Okay. Hmm. I'll look into that a bit. I'm also seriously considering just banking the money. you know, we I we don't need to get too deep into this because I feel like we've already talked about it quite a bit. But like the the main dilemma I'm facing as a business is is our current strategy gonna work, but it just needs more time? Or is our current strategy not gonna work? And like if it's not gonna work, then we should be

A doing different things, but especially we should be like a little more willing to spend money, because we have a lot of money in the bank. Like invest in whatever would work. But if it's more like what we're doing is good, but it's a very long-term play and it takes time, then the right thing to do is put the money in the bank, give ourselves more time, to just execute on what we're doing right now.

Rick (22:41)
Yeah. When do you think you'll

Tyler (22:42)
So those are that

Rick (22:44)
know if what you're doing is gonna like what how what how long do you need before you know?

Tyler (22:53)
I think two years is what I want to say here. And there's two reasons for that. One, this health insurance thing that we're doing, I think that will take two years to see anything. But bigger than that, like our we're not betting the whole business on health insurance, right? We're that's just like a channel to boost our growth a little above what it would be otherwise. But the big thing is all of our other channels, I've said this before too. All of our other channels are dying. Even more now that we're killing Captera, right? Like

We have gone from a business that's getting half of its customers through various channels and half through brand recognition, word of mouth, whatever, like like direct searches for our name, to now a business that's getting 90% of our of our traffic, of our leads from branded search. Well on track for that to be 95 or whatever. So the two-year answer is basically saying the question is not do any of our ad channels work?

Yada yada. The question is, does the organic side of it keep growing gradually? Because it has been growing gradually for a couple years. if it keeps growing gradually for a a long enough period of time that we have confidence that it's a real trend and not just like a fluke, I need about two years before I think I'm I'm just making up a number. That's when I think I would have confidence of like this is long enough of a trend that I think we can just base our business around it.

Rick (24:21)
Yeah, I guess regardless of that that timeline, you still, you know, if you don't know how to spend the money, don't burn it, you know? Like save it.

Tyler (24:29)
Well,

yeah, and but that this is the big question is do we need to know how to spend the money? Like I don't want to. I want to run a business where we focus on customer service, price, which I know price sounds boring because like it's just a number you make up, but there are actual hard things about offering a good service for a low price, and like being able to do that is a skill that we have. so like customer service price and just like a kick-ass simple, easy to use product.

And I just want people realize the value, they like the way we do business, they tell other people, we don't need rocket ship growth, we need 5% per year growth. If we can get that, I don't want to do anything else. I don't want to chase stuff that's outside of our wheelhouse and all that. But of course, like any business, you don't get to you don't just get to make up some fantasy business that works the way you want. Like it has to face reality. And like, is that realistic? Is is what I think two years from now we might have a better sense of.

Rick (25:28)
Yeah, the the only other variable that that would factor into thinking is is AI and its threat to your business. But I guess your your kind of mindset there is like if that's a threat, like I'm there's nothing I can do.

Tyler (25:42)
More or less Yeah, yes. We need to stay on top of like if if the threat of AI is a new CRM comes out that has an AI feature and we don't have it and we lose customers, that's the same as the CarPlay thing I've been talking about for years. It's just AI instead of a mobile app. if the AI threat is people don't need C RMs anymore, yeah, there's nothing I can do about that one.

Shoot, I had something. yeah. When speaking of AI, you know, all this stuff you you naturally talk to Claude or ChatGPT or someone about these days, just to be like, H am I thinking about this right? Not even that I trust any of its answers, but like it's just a brainstorming partner. When I bring this up with them, they're like, there's no fucking way less annoying serum's gonna be able to have any success with any form of advertising. Stop trying. That's what both Claude and ChatGPT say.

They're just like you're you're in too competitive a market with customer with competitors that have a hundred X the LTV, a thousand X the LTV per lead. They're like anything that is price competitive, whether it's PPC ads, display ads, sponsoring people, whatever, you're just never gonna be able to win any of those games.

Rick (27:04)
Yeah, I I'm just wondering like that's not helpful. Thank you.

Tyler (27:08)
Ha ha.

Rick (27:09)
so so so I have w word of mouth brand recognition business. How can I spend money to in accelerate word of mouth is like the next question. What does it say when you ask that?

Tyler (27:19)
Yeah, I mean that's where we

we landed on the health insurance thing there. Now we now have more money than we had when I had that conversation, but that that was the the logic led us there, right? It's too big of an ocean with too many big fish.

Rick (27:34)
pick pick small pockets ponds and and own those those ponds.

Tyler (27:40)
And specifically things that are not purely digital, like a big part of the health insurance push is going to be digital online stuff. It'll be retargeting ads, it'll be webinars, but having that, there's a gatekeeper, the FMO, that's a physical or like a non that's a real thing that you, you know, HubSpot can't buy their way around that. And then also having

Rick (28:01)
Yeah, distribution channel.

Tyler (28:02)
the conferences, like that's the stuff that there's there's not nearly as much competition because of that.

Rick (28:08)
Yeah, it makes sense. But then you run into separate issues, which are like pond specific requirements.

Tyler (28:13)
Yes. Yeah,

like are we even competitive in in that small pond? Because yeah, exactly. You know, it's hard running business. That's life. But I'm still like crossing

Rick (28:23)
Yeah, we'll find problems.

Tyler (28:25)
my fingers and thinking the the organic trend is still it's very slow growing, but it is growing. All the other channels go to zero. That one keeps growing. If that's true, we've got a perfectly fine business here indefinite.

Rick (28:40)
Bye bye.

Tyler (28:41)
We'll see. okay. Thanks for one other just number update. Also in the in the the theme of things are bad things happening, but it's all it's stuff that I don't like doing going away, but that was good. our largest our our one time largest account, which at their peak I think was close to six hundred users, is now down to a hundred.

Our other biggest account, which is now our biggest, they're at 350 right now, told us a year ago that they were gonna cancel. I don't know if they're actually going to cancel. That was like literally more than a year ago, but basically take those two out of the equation, and we don't have any accounts over a hundred users. We're back we're we're kind of getting back to a much more pure play small business CRM with less distraction than we used to have, but the revenue's nice.

Rick (29:38)
Yeah, you should yeah,

that's a good point. You should look at like your data through different segments. Like and if you it's kind of like that we were talking about with the thirty K this month, like there's some noise in there. If you if you mute the noise on the outside of of your core business, is it it are the trends looking good?

Tyler (29:54)
Yeah, good question. I have been looking at this, and the the simple way I'm doing that is I'm not looking at revenue anymore and I'm not looking at users anymore. I'm just looking at what we call user companies, which is like an account. our accounts going up or down. And like, yeah, all the yeah, maybe a s like a a big customer removes a hundred users, and so it looks like a bad month, but our user companies go up. looking at it that way, we are roughly plateaued with a slight drop over the last two years.

But then if you break it down by channel, what you see is our organic is is doing well enough. Everything else is doing worse. I will say we've had a slight uptick in churn recently. is that because of economic conditions? Is that because of AI? Is that just normal old-fashioned competition? I don't know. But I would say overall, organic user company signups looking pretty good with the main.

recent worry being that churn is increasing.

Yeah. I don't know. I I've I am my brain loves a recession story. Anytime I see a news article that's like, here are the reasons we're about to have a recession, I'm just like, I believe it a hundred percent. Like and so I'm my my first instinct when you see turn like that is like, man, small businesses are suffering right now, but that might be total bullshit. I don't know.

You don't think it's right?

Rick (31:26)
I have no idea. I have no idea. I

have I like there's so much going on right now from a I mean we're we're in a war, literally in a war and the economy like right now, like interest rates are high errors than we're used to, I should say.

Tyler (31:43)
Yeah.

Rick (31:44)
we also like I I I think unemployment is going to go up because of AI. Like I don't I don't see that not happening.

Tyler (31:52)
I mean it is g it

is going up for w for some reason. Like it it is going

Rick (31:56)
Yeah.

Tyler (31:57)
up.

Rick (31:57)
And then, you know.

The the the speed at which AI is getting useful, is scary. So I don't know, I if I think about it too much I get I get like my brain starts to hurt.

Tyler (32:12)
Yeah.

Well, and it's it's so funny that on the one hand, a big fear is AI's too good too fast, and on the other hand, the concern is that AI is the most likely reason for a recession 'cause it pops. Like th there's a very narrow band in between those two where it's like it's not devastating in either direction, you know? Cool, cool, cool, cool. what else from you?

Rick (32:36)
on the A AI front, I've been experimenting with AI agents. So how how familiar are you with like clawed scheduled tasks, clawed routines, in clawed code, that kind of stuff?

Tyler (32:46)
not a heavy user of them. I'm familiar with the features existing.

Rick (32:49)
Okay. So and how familiar would like Claude Tag the Slack bot that you can invite to channels? Have you installed that in Less Annoying?

Tyler (32:57)
no, I've read discussion about it. I have not used it. It that does not solve a problem for us at all.

Rick (32:59)
Okay. All right.

Okay. So so Claude Tag is interesting. it's like a it's basically an agent that you can add to Slack channels and give channel specific memory instructions, scheduled scheduled sort of actions, that kind of thing. but it it acts autonomously. So a good example of of a of a Claude sort of tag workflow would be let's say you have a channel in Slack that

Is an alerting channel and it alerts a human to investigate some sort of issue. you could actually have Claude Tag monitor that channel and provision Claude Tag access to the accounts necessary to handle the action and have Claude Tag handle the action. So like this is this is happening in real time. now what what what that leaves is like, okay, well, I don't really want Claude to have access to some of these tools. I really need a custom agent that I manage.

that is in the channel doing what Claude does, but I have more control over it than just Claude, you know, out of the box.

Tyler (34:03)
Okay. Why do you w

what do you mean more control? Like it's L L Ms are kind of a black box no matter what, right?

Rick (34:09)
yeah,

I I don't know. Like maybe you want to configure something more specific than the Claude interface allows you to configure. Because it's basically you can just give general instructions and memory. That's it. So I I don't know a good example, but like let's just say let's just like that's that's these case. Let's say you want to rebrand it. Let's say you don't want it to be called Claude. You want it to called something independent of Claude so you can switch out the LLM underneath. Right? Like maybe you don't want to use Opus, you want to use chat GTP, chat.

Tyler (34:36)
Yeah, you

you want a third party harness, like yeah. The coding world

Rick (34:39)
Yeah. Exactly.

Tyler (34:40)
is is there's a million of these, right? You can use

Rick (34:42)
Yeah.

Tyler (34:43)
AMP, you can use open code, yeah.

Rick (34:44)
Yep. So so what I've started so the evolution has been Claude Tag is like a a a good sort of like for like experimental ground for like what an autonomous agent can do. Eventually you're like, okay, I want to build a custom agent. And that's what I'm starting to exp hit hit right now. And it and it's very, very interesting. I it kind of is forcing me to get uncomfortable with like coding because you have to host the thing.

Tyler (35:12)
Well and presumably you have to write code. Like an agent is not giving a good prompt to a model. An agent is deterministic code that then calls the model. Yeah.

Rick (35:21)
Exactly. And so then you end up with like I I'll just simplify it into the three options that I'm I'm considering. So let's say you you don't want to use Cloud Tag. That's not what I'm talking about here. That is an agent, but like you want something that is an agent that you built. You can Claude has something called Claud Managed Agents where you build custom agents and you skin them however you want to and you can you know call it whatever you want want to and you deploy you deploy it to I think they call it Claude Console.

And they do all the managed the infrastructure management. You basically just set everything up and it runs. On the other end of the spectrum, you build a custom app and you hosted an AWS, like like the like the I I will I'll say the good old days that I'm not familiar with. like you tried to get me to do years ago. And then there's this like emerging middle path, which is companies like Render. Have you heard of this?

Tyler (36:15)
I know them pre AI, I don't think I know Railway I know, yeah, yeah.

Rick (36:15)
rent or like railway, I think is another one. Yeah.

And I and I don't fully understand the the the the the details, but it's some sort of middle path where you they'll host you still have to configure stuff, but they'll handle a lot of the infrastructure management for you. So that's basically you just push push your your code.

Tyler (36:34)
Yeah, it's like the Heroku

Heroku for the AI era sort of.

Rick (36:38)
Can you

can you in in non tech speak explain to me what what's a company like Render is doing or Railway?

Tyler (36:45)
well let me explain what Heroku did, because I'm

Rick (36:47)
Okay.

Tyler (36:47)
I'm not talking out of my ass with that. So like

Rick (36:50)
Okay.

Tyler (36:51)
you could go to AWS and say, okay, I need to host an app. What does that mean? Well, I need an EC2 instance, which is where I'm gonna host my app server, and I need to use RDS for my database, and I'm gonna use the Elastic Load Balancer to handle incoming traffic, and I'm gonna need to set up firewall rules, and I'm gonna need a blah blah blah blah blah blah blah blah blah, right? And then I also need my own custom code to orchestrate all of that to say.

When demand gets higher, I need to spin up extra EC2

Rick (37:16)
I see.

Tyler (37:16)
instances and all that. Heroku, I believe, is literally built on top of AWS, or at least it was. It is, it's all the same stuff, but rather than saying I have to understand all of those pieces and write my own code to orchestrate it all together, I'm gonna pay Heroku a premium. Like I'm I'm paying for all the AWS costs plus whatever Heroku's taken off the top. And it's just a very simplified package that manages all of that stuff for you.

so I wanna think, I want a database. And it's like, here's the database. It'll we'll scale it up, we'll scale it down, and but you know, but you have to pay a lot more for it. I think Railway is more or less that, but really for vibe coders.

Rick (37:56)
Yep. I think render's the same thing for for for this. So I just sent you the the link.

Tyler (38:02)
okay.

Rick (38:02)
okay, so that helps. So so where I'm kind of going now is the reason to go to render is I think you can sub out the model. Like it's a third party. The the reason to go with like a cloud managed agent is it's simpler, right? Like I don't have to like I can literally just like vibe code and claude code and push it to to it can handle everything.

And then I I go I I guess I don't see the benefit of going to the third option, self hosting on AWS. Like, is that just extinct? Like, why would you do that?

Tyler (38:36)
I think you do it because the Herokus of the world charge a premium, or you need more control, neither of which apply. And probably render

Rick (38:42)
Okay. So anyway, go ahead.

Tyler (38:45)
I mean, I'm looking at this, probably render and railway aren't charging a premium yet. These are, I think, VC backed, like still in the subsidizing everything era. So it's probably more a promise that in the future they're gonna have to charge a premium.

Rick (39:01)
Okay, so this kind of sets up the dilemma that I'm having. We are, so I have a clawed license, right? And I get usage with that license, and it is a fraction of what the actual usage costs if you go get build by use on usage. So when I set up a let's just call it, I'll call it a routine, but what I basically mean is like an autonomous, like an a a triggered autonomous task, meaning it happens on a schedule, but the thing that's happening is like,

has LLM calls in it. it is going it is it is billing against my license, meaning there's no cost to it. Like it's free. and it but if I if I set up a third party, if I set up an agent through cloud managed agents, it's 100% usage based. So

Tyler (39:47)
Yes. I

mean right now there's a solution to this. And that the

Rick (39:51)
Okay, say more.

Tyler (39:52)
solution is to use open AI instead of cloud, instead of anthropology.

Rick (39:57)
So open AI can has subscriptions for hosted agents.

Tyler (40:04)
Yes. So or I don't I I I know this more with c in the coding world, but there is a whole history here where for a while everybody you you had your subscription and then if you wanted to use a coding agent, like let's say AMP, or you'd connect your Claude account to AMP and it would use it would use your usage against your subscription. Then Anthropic says, Hey, the what they reali I mean, I'm I'm kind of inferring what was going on internally there, they're saying

This is very bad for us. We're just a commodity here. We're never gonna have any pricing power if people are using the interface of a third-party tool and they can point it at whatever model they want. We need them to use our tools. And so they lock it down and do exactly what you just said. They say, if you want, you can use our models through AMP, but you're paying API pricing. They're heavily subsidizing the subscription specifically to get people locked into their ecosystem. But they've also got this competition with ChatGPT.

And ChatGPT d did the arms race thing of being like, Well, you know what, we're not gonna have that restriction. So you can use the subscrip you can use the subscription with these third party harnesses.

Rick (41:12)
Nice. Well, Anthropic's not there because you because in order to basically anything that the agent does, it's basically consuming API credits, which are are built usage-based. and so anyway, that's where like I'm like, well, that kind of sucks. So now now I'm in this situation where I'm like, okay, I've got this thing working in my Claude license instance as a routine. So my workaround for this is like basically having a shared account.

That everyone logs into. So we need to cla so so the thing I'm running into is people need to collaborate on what this agent can do. and if we each have our own routines and our own logins, it's very difficult to sort of coordinate and manage those routines across let's just call them twenty to thirty routines. But if we have a shared account, we can all log into the same cloud account. Does that make sense? So that's probably what we're gonna do.

Tyler (42:03)
Yeah, yeah, yeah, sure. This sorry, this is a windfall you're talking.

Rick (42:06)
I can't say who well who it's for, but like yeah.

Tyler (42:10)
yeah, okay.

Yeah, I yeah. The reason I asked that question is if it was like up health, I'd say like just switch to just switch to

Rick (42:21)
Open.

Tyler (42:21)
chat GPT. That's what we're doing. We're we're moving all our devs off cloud right now.

Rick (42:27)
Fascinating. Okay.

Tyler (42:29)
It's so weird the whole the whole thing. It's like on the one hand, you want it like this is all really powerful. It has the ability to transform the business. And so you want to invest a lot of time and energy figuring it out, getting really good, being AI native, whatever that means. And then on the other hand, you're like, everything is changing so fast that it almost it feels like a waste to do any of it because you're gonna spend a month automating away something that took way less than a you're not you're not one to one.

Time invested in it to automation in the short term. So it's like, can I just wait a f six months? And by then there will be some startup that that does exactly what you're talking about.

Rick (43:06)
That's exactly right. So so where I that's why I kind of like the shortcut of like just creating a shared account and building all this into routines, which is working for us. It's just the problem is collaboration right now. Like if someone gets hit by a bus, like how do get access to the routine? Like this these actually routines matter.

Tyler (43:20)
Yeah. Yeah.

Rick (43:21)
Like they actually drive value. So that's probably the the the short solution. But then it's like this age at space is more experimental in the short term. but anyway, how do you get like

Tyler (43:30)
The long

term though is you will be paying the API pricing. Like these subscriptions

Rick (43:34)
You have to.

Tyler (43:34)
are yeah. So maybe bite the bullet and do like that would solve all your problems if you could have a shared place that's you it's a shared account in the sense it's the API is one thing, but everyone has access to the the front end or whatever.

Rick (43:47)
the agent.

Yeah. So my question is like when we start talking about collaborating on an agent, like I built a startup to last bot this morning that every other Sunday it goes and looks at startup to last dot com to see if there's any episode new episodes. It reads the transcript and then it suggests derivative content ideas. that's that's what the bot does. it costs like 30 cents per per run. Yeah.

Tyler (44:14)
Per run. Okay.

Rick (44:17)
The the like say I wanted to like, okay, like that's the first capability of the startup to last bot. We'd like startup to last bot to do more, like share the share these derivative content ideas as a tweet or a LinkedIn post or whatever. what like how would I collaborate with you on developing the agent? Like this is where I'm like, I don't know how to collaborate in this

Tyler (44:43)
Yeah.

Rick (44:44)
world.

Tyler (44:45)
I mean, so A, everything you're saying sounds to me like a startup idea. I'll just say that. That probably means there is a startup doing this, but it also probably means there's an opportunity to start a startup doing this. But if you just think of an agent as code, how do you collaborate on writing code the good old fashioned way? We both have GitHub, we pull down we clone the repository, we pull down the code to our computers, we

We both have the API key on our computers. When I make a change, I push it up to GitHub, you pull it down. Now, is it realistic to expect a marketing team at a Fortune five hundred company to do that? Absolutely not. But I think that's the answer for you and me probably.

Rick (45:24)
Well that's the so so let's let's talk about the non technical people. Like I could probably bridge to what where you are easier than the marketing team. So we marketing team like the the future is marketing teams are gonna need collaborate on managing agents. How do they do it?

Tyler (45:40)
Yeah.

Rick (45:40)
Is it GitHub?

Tyler (45:42)
I bet I bet it's a standalone product that it's an agent builder. Like how do you how do you collaborate on design? You use Figma. How do you collaborate on text? I mean there's a lot, but Google Docs or whatever. How do you collaborate on agents? I bet there's a tool that.

Rick (45:58)
Well, now you're validating the fact that like this is the reason for non-tech the short-term non-technical solution is shared claud account with like a shared login and managing it through an inner the routine interface because that is very like it's an interface. It's not code. and yeah, you use claud code to build the prompt, but the prompt sits inside of a a WYSIWYG.

Tyler (46:21)
Yeah.

Slightly more short term, a probably if you want to do the code thing, but make it as easy and accessible as possible. I just became aware of AMP orbs. So AMP is a coding harness. I think AMP orbs are basically like hosted virtual machines. So like rather than coding on your computer, you're it they're running the code in the cloud. I'm sure there's other versions of this, but this is just the one I recently heard about where

I think you could so the hardest thing about a non-technical person collaborating on GitHub now, it's not the code. It used to be, well, they don't know how to code. That's not a problem anymore. AI does that for you. The hardest part is the environment. It's like, how do I even run the code? How do I store my anthropic API key securely, etc.? I think with something like Orbs, where it's all hosted, you, Rick, or someone who's knowledgeable about it, could go set up an environment, and then anyone can log into it from anywhere. So one of the keys

This is a pattern for all collaborative tools everywhere, is it had to be moved online. Local work is very fucking hard to collaborate on. That's why Claude lets lets you solve this is because it's hosted.

Rick (47:35)
Yeah, I see. There's no deployment. There is a deployment, but like it's it handles it all for you.

Tyler (47:43)
Well you edit a file and you don't have to with I don't have to commit it to GitHub and have someone else pull it down to their computer. I edit the file, that's the same file you're looking at because we're both accessing the same server.

Rick (47:52)
I see, yeah. So how do they handle like us not right overwriting each other's changes?

Tyler (48:00)
I mean, how does Google Docs do it? I don't know. And I don't even

Rick (48:03)
Doesn't matter.

Tyler (48:03)
know if orbs orbs might not they're not like I don't think meant for like collaborative coding per se, but some something like that. It would still be an improvement to just be like

Rick (48:10)
There's some sort of version

history. And like if if we have different API keys, like it'll tell us like this person used this API key, like Rick's API key

Tyler (48:17)
Well

Rick (48:17)
made this edit.

Tyler (48:18)
yeah, it well if you're doing AMP specifically, it's still connecting to GitHub and all that. But my point is like

Rick (48:22)
Okay.

Tyler (48:23)
the invi someone doesn't have to be like, I have to install GitHub on my computer and clone the repository and run install Docker and run the right command to set it all up. You just say, Hey, here are your login credentials to your orb. Yeah. Like, yes, the code

Rick (48:35)
And it just handles it all for you in the cloud. that's cool.

Tyler (48:38)
still needs to get committed to Git, but that other stuff is solved, I think, is the idea.

Rick (48:43)
That's very cool. This was a helpful conversation.

I I I'll I'll share more about this in the future.

Tyler (48:52)
Yeah, I'm I need to go in a s minute here, but I'm really interested in what you come up. I I feel like I have more coding knowledge and I probably l like follow more coders, so I know more about even though I'm not doing it myself, I know a lot more that about the coding side of AI than you do, I think. But you are so far ahead of me on agents, which is actually the more interesting thing right now, I think. And like, how do you how does a non technical person use AI to do whatever the hell they want? I think you're gonna be

a good resource for me and for all the listeners of this podcast to to learn about that stuff.

Rick (49:26)
I I I think so. The the the thing that I'm scared about is like I think it's really easy to just spin up a bunch of shit in a silo, but like how do you

Tyler (49:33)
Mm-hmm.

Rick (49:34)
collaborate on these things across an organization? That's where I'm getting stumped. And it sounds like you know how to do that from your background.

Tyler (49:40)
I mean better better than the ba the I have some ideas, but I I

Rick (49:46)
Mm-hmm.

Tyler (49:46)
again I think there is a very legitimate startup idea here of of doing exactly what you're doing and making it so seamless that it's like a Google Doc. It's like just go in and start typing, don't worry about where it's hosted, don't worry about how code versioning's working, etcetera.

Rick (50:01)
Supposedly Claude, ChatGPT, Jim and I should be the ones to solve this, right?

Tyler (50:06)
No, but but no one wants to get locked in. Those the model providers are all fucked long term. That's my I feel strongly about that. That people are gonna want third-party harnesses that can use any model.

Rick (50:20)
I see what you're saying. Interesting.

Tyler (50:22)
We'll see. We'll see. But that's that's my bet. So anyway, I gotta run. Good talking to you. See you later.

Rick (50:26)
All right. See ya next week or two weeks from now.

Tyler (50:29)
Two weeks, yeah.

Rick (50:30)
See ya.