Power-Law User Acquisition in the Age of AI

1078
03:57:52.550 –> 03:57:57.240
Julia Nimchinski: Thanks again, and welcome to the show. Cindy Daio.

1079
03:57:59.130 –> 03:58:04.320
Julia Nimchinski: our next session is power law, user acquisition in the age of AI.

1080
03:58:05.220 –> 03:58:06.689
Julia Nimchinski: How have you been, Cindy?

1081
03:58:07.490 –> 03:58:10.320
Sandy Diao: I’ve been. Well, thanks so much for having me here, Julia.

1082
03:58:11.320 –> 03:58:19.099
Julia Nimchinski: Super excited. Got a lot of feedback from our folks, and can wait just to jump into it. Do you mind giving

1083
03:58:19.590 –> 03:58:22.240
Julia Nimchinski: quick introduction before we just get started.

1084
03:58:23.070 –> 03:58:47.490
Sandy Diao: Yeah, absolutely. So today, I’m going to spend the next 30 min with you all share a little bit more about some of my experience leading user acquisition at over 5 technology fast growth companies. And there’s really been a lot that’s changed particularly in the last few years. Here, however, there are a few universal truths around how these channels operate, and when it comes to finding really scalable, durable, repeatable channels for growth.

1085
03:58:47.490 –> 03:58:58.410
Sandy Diao: There are some fundamental frameworks that I’ve been able to develop in my career over the last 12 years and wanted to share a sneak peek of that with everyone today. So that’s what I’ll be diving into

1086
03:58:59.980 –> 03:59:00.820
Julia Nimchinski: Instead

1087
03:59:01.610 –> 03:59:05.200
Sandy Diao: Perfect. All right. Well, I’m gonna start by sharing my screen here

1088
03:59:06.400 –> 03:59:21.740
Sandy Diao: and let’s dive in. So yeah, welcome to probably the midpoint of your AI practice sessions or a little bit after that. And today we’re going to be talking all about this concept of power law user growth which I’ll define a bit more as we get into the presentation here.

1089
03:59:22.560 –> 03:59:32.580
Sandy Diao: Kind of quick context. I alluded to this a bit in the introduction here, but a lot is changing around the growth landscape. When you browse Linkedin, when you chat with your teammates, when you chat with

1090
03:59:32.800 –> 04:00:00.450
Sandy Diao: peers in the growth space, the number one thing you hear is this worry about moving faster, this fear of channels being saturated, and the outcome of that really has been that a lot of these AI. Native startups are seeing growth at a trajectory. That’s way faster than anything that we’ve seen before. These are just a handful of the examples of some companies that have gone from 0 to tens of millions, even hundreds of millions in arr within just the course of a few months.

1091
04:00:01.280 –> 04:00:24.670
Sandy Diao: And that’s actually a bit of a stark contrast from the last few years. So pre age of AI, we had startups really launch their hypothesis around their core products, and what we saw was that even some of the top b 2 b startups took an average of, let’s say, 2 years to hit their 1 million dollars in arr right? So the landscape is moving a lot faster and things are changing dramatically.

1092
04:00:24.670 –> 04:00:44.910
Sandy Diao: So the best way to really seize that opportunity is, we need to figure out how we can grow, not just with linear outcomes, but with power. Law outcomes right? Because if we know that this landscape of channels is getting saturated, and everybody’s vying for attention. And there are a lot more competitors out there, and the best thing that we can do is figure out how we can drive nonlinear growth.

1093
04:00:45.230 –> 04:01:07.720
Sandy Diao: And just to paint that picture a bit more. Here I have on screen a few examples of different traffic sources for some of the aforementioned fast growing AI native startups right? And there’s 1 pattern that you see consistently across all these traffic sources, which is essentially that there’s 1 or 2, maybe 3 channels that drive the majority of growth. These are what we call power law outcomes.

1094
04:01:08.600 –> 04:01:21.159
Sandy Diao: and this is the primary thesis that I have when I walk into any company when I join a company, and my mandate is to grow revenue for a company. My job is to figure out, how do we identify those power law outcomes.

1095
04:01:21.310 –> 04:01:27.230
Sandy Diao: and that’s to figure out the one or 2 channels that will help drive the majority of growth for a company.

1096
04:01:28.170 –> 04:01:54.130
Sandy Diao: And I’ve been able to do this at many companies where I’ve led growth. For instance, teams at Meta pinterest descript. Just name a few and also teach growth marketing, and this topic a little bit more broadly as well at Uc. Berkeley, Stanford, and run a growth advisory as well. But enough of that introduction. Let’s dive into the framework here. So I’ve done my best to condense that framework and those learnings into these 3 steps that I replicate every time I step into one of these new head of growth roles.

1097
04:01:54.880 –> 04:01:56.610
Sandy Diao: The 3 steps are the following.

1098
04:01:56.920 –> 04:02:04.989
Sandy Diao: first, st we want to identify for this company that we’re talking about. What are our products, natural advantages. We’ll define that in a second. Here.

1099
04:02:05.930 –> 04:02:11.969
Sandy Diao: The second step is, we need to figure out how we can match the best channels to fit those advantages.

1100
04:02:12.790 –> 04:02:23.630
Sandy Diao: And lastly, we want to use AI to maximize our positioning tests across those channels. So let’s dive into it. Starting with the 1st step here identifying our products, natural advantages.

1101
04:02:24.260 –> 04:02:28.419
Sandy Diao: Well, before we define that, let’s talk about the landscape of channels.

1102
04:02:28.740 –> 04:02:41.590
Sandy Diao: there’s actually a finite number of power law channels out there, meaning that these channels are repeatable, scalable cost, efficient to invest in, and we can count on them time and time again to drive continued growth

1103
04:02:41.780 –> 04:02:49.760
Sandy Diao: right. So if we can put all the channel categories on a list here, that means that we can actually limit the scope of what we’re testing.

1104
04:02:50.390 –> 04:03:12.219
Sandy Diao: But the reality is that when we’re leading growth, when we’re trying to drive revenue for a company, our main goal is not to come in here and just cross off the list like I don’t get hired to go through this entirety of the list and cross everything off and say I tested it. It didn’t work. This is not gonna work for us. No. Instead, my goal is to figure out how to be strategic and figure out which of the channels are the best fit for my product.

1105
04:03:13.680 –> 04:03:16.520
Sandy Diao: and that’s where natural advantages come in.

1106
04:03:16.880 –> 04:03:21.739
Sandy Diao: You have a product, and it naturally is going to have some kind of distribution advantages.

1107
04:03:22.120 –> 04:03:32.690
Sandy Diao: The 2 key criteria and the 2 factors that I found to be most important and advantageous are the following, the 1st is some kind of virality advantage.

1108
04:03:32.700 –> 04:03:55.270
Sandy Diao: This usually refers to product or social virality. So this answers the question of how likely are you to go viral? And something goes viral when it can hit an n plus one audience, meaning that the product is, for instance, used by multiple people. It’s easily shared socially on social platforms or with other people in conversations and threads and emails. So that’s virality potential.

1109
04:03:55.590 –> 04:04:14.749
Sandy Diao: The second category is whether or not the product is something people easily understand and are actively shopping for, and will buy. This refers to your conversion likelihood right? If somebody knows what they want, and they understand what it is that you’re offering, you’re much more likely to get them to buy it versus offering something that they don’t understand.

1110
04:04:15.120 –> 04:04:29.259
Sandy Diao: So the intersection of your virality potential and your conversion potential is where I believe, and I have learned over time. This gives you your best, unfair advantage when it comes to user acquisition and distribution more broadly.

1111
04:04:31.410 –> 04:04:45.690
Sandy Diao: If we break that down a bit further, I found that products generally have 3 types of natural advantages, discrete productized experiences. The 1st one is product sharing, the second is social sharing, and the 3rd is existing demand.

1112
04:04:45.850 –> 04:04:47.629
Sandy Diao: I’ll define each of these for you.

1113
04:04:47.890 –> 04:05:15.110
Sandy Diao: Starting with product, sharing product or sharing refers to the fact that the product just gets better as other people use it with you. So, for instance, if your product is a b 2 b tool, and it’s generally collaborative, people can comment in it. Edit with you, share workspaces. That’s a product that’s more useful the more people use it. Or it could also be a product that gets more valuable as other people join typical product network effects, you get more inventory in a marketplace that you can take advantage of.

1114
04:05:15.180 –> 04:05:31.860
Sandy Diao: And also these are products that are really meant for sharing with other people. So for example, if you film a video of yourself talking through meeting notes or talking through a concept, and you share with your teammate that’s really meant to be viewed by N plus one or somebody else in an organization.

1115
04:05:32.300 –> 04:05:51.350
Sandy Diao: So some product sharing examples here more visually as well. For instance, if you are using stripe invoices, and you, Bill, a customer they receive the bill, and they see the stripe logo as part of receiving that invoice to pay, and during that process, if they need an invoicing tool, they will naturally think of stripe.

1116
04:05:51.550 –> 04:06:03.569
Sandy Diao: Similarly with figma. If you’re a figma user, it’s really meant to be collaborative, right? You want to have somebody else on your team review. That file leave comments, edit it, etc. So it’s really meant to be shared.

1117
04:06:05.630 –> 04:06:19.659
Sandy Diao: The second category is social sharing. Social sharing advantages take place when sharing the product in some way gives the user who’s sharing it some kind of social utility, social benefit, recognition or rewards

1118
04:06:19.850 –> 04:06:37.230
Sandy Diao: right? And these are product experiences that sometimes are visual in nature, because visual items, like images or videos, are really easy to share with other people and post in different places. These could also be experiences that give you some kind of status or recognition.

1119
04:06:37.460 –> 04:06:54.849
Sandy Diao: For example, the Strava app, which is an app that helps you to track and document. How much physical activity that you’re engaging in. Well, if you just had a run, and you’re super proud of the stats of that run, you’re very likely to share it with your friends, or on social and in other places

1120
04:06:54.970 –> 04:07:12.320
Sandy Diao: or opus clip, which is a tool that helps you turn long form video into viral short clips. These are visually. This is very visually engaging content. So that video is going to naturally be shared on social platforms. So passive sharing experiences like watermarks. For instance, work really well here.

1121
04:07:13.930 –> 04:07:31.420
Sandy Diao: and the last category of advantages tend to be existing demand. This is where you are building a product that you know, a lot of people are already searching for where it’s targeting some kind of latent demand. So just a clear example here, if you are a company like, say, Asana.

1122
04:07:31.420 –> 04:07:42.510
Sandy Diao: a category that you probably care a lot about is say, project management software. Right? Well, it turns out that a lot of people are already searching for that in Google search. So you definitely want to show up in search engines.

1123
04:07:42.510 –> 04:08:11.619
Sandy Diao: But let’s say that you’re thinking about building a new product. Let’s say you want to think about like an AI girlfriend concept because you love the movie, her and you go on. You know, one of these keyword research tools. You find that it actually turns out there’s a lot of people searching for AI girlfriend and keyword difficulty is somewhere between easy and media, meaning that there’s not a lot of competition. So this is an example of you identifying that there’s people who are actively searching for maybe a category of product or tools that don’t exist yet. And you’re going to be able to target the latent demand that exists for it.

1124
04:08:14.020 –> 04:08:36.860
Sandy Diao: So once we’ve identified our products advantages. One of those 3 categories, product sharing social sharing or existing demand. Now, what we need to do is we need to go back to that list of 7 categories of channels, the one that fit the page, and we need to figure out what are the top channels that best fit my advantages. So I can really grow at the intersection of my virality, plus my high conversion.

1125
04:08:38.300 –> 04:08:56.579
Sandy Diao: And we want to start with the b 2 b framework here that helps us to match those 2 together on the X-axis. We have that virality potential combining the product and social sharing virality. And on the Y-axis we have how much existing demand we’re targeting whether it’s really high or really low.

1126
04:08:57.010 –> 04:09:06.340
Sandy Diao: If you’re a b 2 b company, and you’re trying to target high existing demand, depending on whether you’re targeting the low social virality versus high social virality.

1127
04:09:06.370 –> 04:09:27.129
Sandy Diao: There are different channels where your main goal is to figure out how you can capture an existing need. So search engines the ability to share the product within an organization and even outbound sales with a very, very targeted pitch targeting, very specific persona or Icp. These are great ways for you to be able to get power law effects for your acquisition.

1128
04:09:28.140 –> 04:09:53.430
Sandy Diao: and if you’re targeting low existing demand, your main job is convince people convince the people that are potentially buyers for your product of a need. So the best thing to do here is going to be focusing on channels that allow for you to reach your target audience, but use things like communities and content and passive shares to subtly and slowly influence these individuals over time, so that you build trust, credibility and ultimately conversion. In the longer run.

1129
04:09:54.900 –> 04:10:24.069
Sandy Diao: We’ll go through a few examples here of spanning the spectrum from, you know, low social sharing to high and starting with canva, which is a more individual focused tool to start with. You know, when Canva 1st launched, they’re in this category of digital design tools. And this is definitely already a very big category, meaning that it’s a lot of search demand for it. So SEO is a great channel for them, and because it’s something that people understand when they use organic, social as a strategy, they can create content that resonates with a lot of people.

1130
04:10:25.300 –> 04:10:49.989
Sandy Diao: Another example, a product that’s really made for small and medium sized teams. Vanta Vanta. When they 1st launched. They are a compliance solution for people like developers and people who are building products. And when they 1st launched, you really thought of compliance as a solution primarily for enterprises. So they really carve their own space, and they started out by creating a lot of content that helped developers and companies understand

1131
04:10:49.990 –> 04:11:02.379
Sandy Diao: why even a small and medium sized businesses needed to care about building compliance and security into their products. So this allowed for them to drive a lot of credibility and drive a lot of awareness amongst their target user base.

1132
04:11:03.730 –> 04:11:28.709
Sandy Diao: And lastly, we have an enterprise facing product, with high existing demand, a rippling and all in one Hr platform. This is definitely a space that has a lot of existing demand. You know a lot of companies and enterprises already use Hr solutions abundantly in their companies. So you know, rippling by innovating on a few feature sets and and their holistic all in one platform. They can be very targeted about who they reach out to through an outbound sales motion.

1133
04:11:28.710 –> 04:11:36.350
Sandy Diao: and, of course, as more employees join the company, they’re able to, you know, easily invite people and and onboard more people into the product itself.

1134
04:11:36.350 –> 04:11:47.889
Sandy Diao: So you know here the high existing demand channels are really those that allow for you to target as many of those high, intent customers as possible through search engines, through that cold outreach.

1135
04:11:47.980 –> 04:12:01.290
Sandy Diao: and then, secondarily, with a low existing demand. Your main goal is to use content or the influence of other people, whether that’s people in your target community, other thought leaders and influencers to convince people and convince companies of a need.

1136
04:12:01.460 –> 04:12:04.099
Sandy Diao: So these are some of the b 2 b examples. Here.

1137
04:12:04.910 –> 04:12:09.510
Sandy Diao: let’s transition. And look at what this might look like for consumer companies B to C,

1138
04:12:10.550 –> 04:12:39.600
Sandy Diao: we again look at the X-axis here the product and social sharing type on X-axis. And then Y axis is how much existing demands that we’re targeting well for consumer companies. If you’re targeting a lot of high existing demand. Your main goal is to target and appeal to personal utility, meaning that you’re trying to help individuals understand? Why is it going to be beneficial, that you use this product. And here, very similarly, your goal is to use channels that target, that intention through search through organic social referrals, etc.

1139
04:12:40.340 –> 04:12:59.120
Sandy Diao: If you’re a consumer company and you’re targeting low existing demand, you’ll want to actually appeal to some of that social utility. So your goal is to try to figure out how you can provide that status reward or recognition that gets people excited about engaging in this product, even if they’re not looking to use that product at this very moment.

1140
04:12:59.240 –> 04:13:06.560
Sandy Diao: So a few examples to go through here, going from, you know, individual to higher social and product sharing potential

1141
04:13:07.290 –> 04:13:21.239
Sandy Diao: individual product. Let’s start with aura ring. Aura ring is a sleep tracking wearable. You wear the physical wearable while you’re going about your day while you’re in bed, and it tracks how well you sleep, which gives you metrics so that you can actually improve your sleep.

1142
04:13:21.320 –> 04:13:37.339
Sandy Diao: And when Oura ring 1st launched, you know they were 1st in their category. People weren’t looking up sleep tracking wearables. People didn’t even know that there was such a concept. So Oura ring really leaned into paid advertising, working with influencers to drive up that credibility and awareness about the category.

1143
04:13:38.370 –> 04:13:47.649
Sandy Diao: Looking at products for small groups like venmo venmo is really designed for small groups of friends and and neighbors and people who knew each other to split bills.

1144
04:13:47.850 –> 04:13:58.810
Sandy Diao: And this is why it worked really well for them to launch within specific communities and benefit from referrals as a great way for them to drive parallel outcomes for user acquisition.

1145
04:13:59.800 –> 04:14:19.159
Sandy Diao: And lastly, large social fortnite fortnite is a game that’s played by a lot of people online. At the same time, when they 1st launched, they were category creating. So their main way of growing actually were twofold. Firstly, some of the game clips that users engaged with were shared on social media. Those clips are seen by a lot more other potential players.

1146
04:14:19.160 –> 04:14:29.939
Sandy Diao: and then, secondly, a lot of streamers through game, streaming platforms would stream their experiences and their massive audiences would follow along and want to be encouraged to play the games as well.

1147
04:14:29.940 –> 04:14:41.950
Sandy Diao: So you know similarly here for consumer products. If people already know that they have a need, your main job with these power law channels is to figure out how to find them across social search and these various communities.

1148
04:14:41.950 –> 04:15:00.030
Sandy Diao: And if you’re targeting a category with low existing demand. You really want to use channels that can put you in front of those target customers that you have. But again, you need to figure out how you can use a combination of content and influence to get them excited about potentially using the product at a later time.

1149
04:15:02.670 –> 04:15:22.620
Sandy Diao: Once we figured out what those top channels are that map to our advantages. This is where we want to use AI to maximize how well we test those channels. And the playbook here is simple. Now, in our in our age of AI, our main goal is to figure out how we can maximize positioning across these different channels and their audiences.

1150
04:15:23.420 –> 04:15:37.859
Sandy Diao: What is not going to work in today’s day and age is taking one singular message and positioning and spread it across all of those different tests that we have across different channels. And the reason is because every user that interacts with an ad or interacts with

1151
04:15:37.860 –> 04:15:55.960
Sandy Diao: some kind of inbound messaging from a brand or a company across all of these different platforms. They’re used to receiving information in different ways. So even something like, let’s say, paid ads, Meta ads versus Google ads. Even within Meta ads themselves. You want to make sure that you’re testing 5 to 10 different variants on your position

1152
04:15:55.960 –> 04:16:21.620
Sandy Diao: instead of just saying, well, my overall product positioning is X, and that’s the only tagline that I’m going to test. In fact, if you test it that way and it doesn’t work. It doesn’t necessarily mean that the Channel is not going to work for you. It means that you haven’t figured out how to talk to the audience yet. So this is where, really, in the age of AI, it’s really crucial and and really beneficial to be able to also use AI to generate that those different positioning tests on your behalf.

1153
04:16:22.240 –> 04:16:47.219
Sandy Diao: Here’s a quick little tutorial on how I’ve been able to do this. I have on the left side here an example of a prompt template that I use, no matter what channel it is that I’m creating different positioning variants for, and the process to go through an example with perhaps Ads copy is very simple. You choose your favorite Llm. Tool to generate copy. Imagine you can probably do this for visual assets

1154
04:16:47.220 –> 04:16:55.050
Sandy Diao: as well. You want to provide context and details around the Channel, the product, your brand style and tone, audience and conversion goals.

1155
04:16:55.290 –> 04:16:57.509
Sandy Diao: the most important part of this

1156
04:16:57.550 –> 04:17:21.390
Sandy Diao: example here. But all the other templates that you want to provide is providing examples. So definitely want to use tools like the Meta Ads library, the Google Ads library find competitor examples of ads that have been running for a really long time. You know, if it’s been running a really long time, it means it’s probably working for one of your competitors or adjacent companies that you are observing provide those examples as part of the prompt.

1157
04:17:21.420 –> 04:17:32.830
Sandy Diao: And you basically want to have an output format that gets the Llm. To generate 10 plus positioning variance. You pick the best ones, and then you run with it, and this will allow for you to

1158
04:17:32.830 –> 04:17:51.749
Sandy Diao: essentially identify whether or not it’s a positioning problem, or it’s just the channel that’s not working for you. So you know, you really want to maximally test as many positioning variances as possible. This is very different, by the way, than maybe doing this 5 years ago, where actually adding in new positioning variants, because.

1159
04:17:51.750 –> 04:18:15.740
Sandy Diao: you know, changing a positioning may actually require you to change, not just the headline and the description, but also the landing page and the copy on that landing page and the follow up emails. Now, all of that can actually be very streamlined and generated with AI tools. So you know the cost of testing these maximal positions is actually much, much lower, and should be a core part of you testing whether or not this is actually a channel that works for your product.

1160
04:18:18.580 –> 04:18:41.610
Sandy Diao: Those are the 3 quick steps. And you know, let’s say you’ve run through the steps. You ran all the positioning tests, and it. It didn’t work. So you know. Now, what do we start back at? Top number one? Well, actually, there’s another way to solve this which is actually thinking about your product strategy relative to this framework. What you want to think about is how you can actually build products for distribution. There’s a few ways to think about this.

1161
04:18:42.430 –> 04:19:12.089
Sandy Diao: If you’re building a product that maybe is emerging. Tech people don’t really quite understand or grasp it. Yet you want to figure out how you can get people excited about your company or your products. Well, one thing you can do is actually figure out how you can build core features into your product. Experience that target, that high existing demand right? A great example. Here is a company like Credit Karma. Now Credit Karma launched a Free Credit score poll, and before they did this, every time you wanted to check your credit score you have to pay one of these other providers.

1162
04:19:12.090 –> 04:19:22.409
Sandy Diao: So Credit Karma launched one of the 1st Free Personal Credit score checks. And when you search for a credit score check, you basically see Credit Karma offering the only free version of it right?

1163
04:19:22.410 –> 04:19:44.220
Sandy Diao: So users were inclined to convert. This also allowed for them to rank in search engines, and they could double down on that search presence as well by running those pages as paid search ads, because there’s so much demands there. So this is an example of a company that actually offers this vast personal financial data platform which people may not want to buy right away. But they care a lot about

1164
04:19:44.220 –> 04:19:51.829
Sandy Diao: free credit scores as the primary driving topic. So this allows for them to be able to target a really high demand category and acquire users.

1165
04:19:53.030 –> 04:20:13.800
Sandy Diao: The second way to build for distribution is, you know, build for even more demand that exists out there. For example, a company like Wix, they have an online website builder. There’s so many website builders out there. But we know that it’s a very big category. And we know people need these kinds of tools. But Wix actually built a logo maker on top of their core website builder. So this is actually a bit adjacent.

1166
04:20:13.800 –> 04:20:38.160
Sandy Diao: It’s not really part of that core website stack. But they launched a separate logo maker because they knew that their their target, small business audience would during their early stages of incorporating the business would eventually need to figure out how to get a logo before they launch the websites, so allow for them to target these users upstream, and they knew that there was a lot of existing demand for this category.

1167
04:20:40.080 –> 04:21:01.109
Sandy Diao: One of the other ways to build for distribution is trying to think about how you can make it easier for people to export the output of your product experience. For example, a company like 11 labs which actually also showed up on our list of some of the fastest growing and native startups out there. 11 labs is a voice Api for businesses and enterprises.

1168
04:21:01.340 –> 04:21:20.929
Sandy Diao: And they actually built a free AI voice generator Demo, that wasn’t just used by, you know, serious developers and enterprise customers who are seriously exploring the technology. But it’s actually used by a lot of consumers. And specifically, these consumers were content creators who really wanted to understand the bleeding edge of voice. AI, so they would use

1169
04:21:20.930 –> 04:21:41.689
Sandy Diao: the AI voice generator, generate these audio clips post about it on social. And this actually resulted in, you know, hundreds of millions of views across organic social media. So thinking about ways of making it super easy and valuable for people to get social value out of using the product by exporting the output in some way is a great way of thinking about building for distribution.

1170
04:21:43.970 –> 04:21:51.179
Sandy Diao: We’re nearing the end of our time here, and I wanted to provide a quick recap on the the past 25 min that we’ve spent together here.

1171
04:21:51.470 –> 04:22:06.850
Sandy Diao: When you think about power law acquisition, you really want to think about not just testing through every possible user acquisition channel. No. Instead, our goal is to figure out how we can focus most of our resources on our top, one to 2 power law channels because we want power law outcomes

1172
04:22:07.760 –> 04:22:23.090
Sandy Diao: when we’re targeting those channels. We don’t just arbitrarily pick them. We don’t just pick them based on what competitors are using. Instead, we want to mirror them to our products. Advantages. The 3 categories to consider are product sharing social sharing and existing demand advantages.

1173
04:22:23.930 –> 04:22:37.919
Sandy Diao: When we’re testing those channels, we really need to maximize how we test positioning. The framing on the product oftentimes can change the outcomes of the channel dramatically. So we definitely want to be using AI to maximize those positioning tests.

1174
04:22:38.380 –> 04:22:52.230
Sandy Diao: And finally, if all else fails, you can think about actually shaping your product strategy around distribution. How do you make it easy for people to help you grow the product. And this is what drives power, law, or viral growth rather than linear growth.

1175
04:22:54.050 –> 04:23:11.790
Sandy Diao: And I think we’re nearing the end of our time here really appreciate everyone’s time and attention. You’re welcome to connect with me on Linkedin. I teach these topics at Uc. Berkeley online, across many different platforms. If you’re interested in courses like that definitely. Shoot me a note and I’ll definitely send you the links. Thanks so much. Everyone

1176
04:23:12.220 –> 04:23:33.899
Julia Nimchinski: What a treat, Cindy! Thank you so much. Super insightful. I enjoy reading your newsletter and encouraging our folks. Please subscribe to Cindy’s Newsletter. I know you have a micro community which is excellent, too, Cindy. Just curious to get your take on, since you’re testing a lot of AI tools, especially for growth and marketing.

1177
04:23:34.210 –> 04:23:38.320
Julia Nimchinski: What technology are you most excited about in the market right now?

1178
04:23:40.560 –> 04:23:49.639
Sandy Diao: I’m really excited about tools that allow for us to upload as much context as possible. I know I was alluding a bit to this. There’s a lot of

1179
04:23:50.330 –> 04:24:08.549
Sandy Diao: unbundling that we’re seeing with different functions of growth. You know, for example, if we looked at just simply creative assets, there are tools that help you create. You know, long form video, like short social clips, clips. For, like Instagram versus Linkedin versus, you know, X, etcetera. So

1180
04:24:08.780 –> 04:24:31.230
Sandy Diao: there’s a lot of unbundling happening in the space. And I think one of the challenges with that is growth needs to growth. Decisions like this need to be made in the context of all the data that we have. And I think one of the challenges that I have with those tools right now is it kind of takes a really specialized operator to be able to make good decisions around whether or not the outputs of those are actually good, so to speak.

1181
04:24:31.230 –> 04:24:54.459
Sandy Diao: right? And so I know that these tools will improve over time, especially as they have more usage. And they have their own models and data around those specialties. But you know, personally, I think that it’s actually much more worthwhile to figure out how to build your own custom stack. Using your own like personalized model. You can pick any foundational model of your choice here, like I, you know, use clotted incredible amount for for writing.

1182
04:24:54.500 –> 04:25:22.529
Sandy Diao: but the more that Claude knows how I think about different channels, the types of copy that I consider to be good. And then, when I can feed back in the copy that actually worked. It does a much, much, much better job of producing net new content for me on the other side of it. So I think I’m much more excited about tools that allow for that context upload, and I’m curious whether or not that can be bundled in some way. So you know, kind of a question Mark for me, but that’s how I think about it.

1183
04:25:23.850 –> 04:25:29.435
Julia Nimchinski: I have a couple of questions here from our community. And 2 min to address some of those.

1184
04:25:29.860 –> 04:25:31.330
Julia Nimchinski: So here’s 1.

1185
04:25:31.560 –> 04:25:35.809
Julia Nimchinski: What’s the biggest mistake companies make when scaling acquisition channels

1186
04:25:37.760 –> 04:25:39.140
Sandy Diao: Biggest mistake.

1187
04:25:48.550 –> 04:26:08.559
Sandy Diao: I think it depends on the stage of company for earlier stage smaller teams. The biggest mistake, I find is not trying to double down on those power law channels. I think there’s this urgency in small companies to grow as fast as possible. So when you see something working early. Your 1st instinct is to like.

1188
04:26:08.720 –> 04:26:16.129
Sandy Diao: duplicate it right, and find us in a 3, rd or like 5 ways to do it rather than like, really make this channel scale up.

1189
04:26:16.220 –> 04:26:28.280
Sandy Diao: So I think that’s the mistake I see for early stage and then for more later stage companies. I think one of the challenges I see there is, you know, kind of letting the metrics get ahead of like intuition about customers.

1190
04:26:28.290 –> 04:26:49.210
Sandy Diao: You know, the larger a channel gets, the more it’s about figuring out the Ltv to Cac and balancing investments and and growth over time. And and then you really start to, you know, pick apart the data in every which way. But I think what that ends up driving sometimes is this overall lack of awareness around, like how customer intuition

1191
04:26:49.210 –> 04:27:17.000
Sandy Diao: should be paired into those channels, and that draws companies to basically pull back investments and channels that could actually be working better if they were to put in more resources. Oftentimes those resources come by the way, with that similar insight that we had in our presentation today around positioning insights, you know, over time, especially if a company’s invested in, let’s say, paid ads for 2 to 3 years, or even greater than that. Well, the market already knows the product. Your audience starts to become finite in who you’re able to reach

1192
04:27:17.000 –> 04:27:40.549
Sandy Diao: audience expansion only works. If you have positioning for a new adjacent group of users who can benefit from the product experience. So there’s saturation that happens. And it’s hard to build that into the strategy. If you’re just looking at Ltv to Cac, if you’re just looking at payback period for just measuring it, based on, you know, the existing experiments and and campaigns that have worked. So I think it depends on, you know, early stage Versus later stage

1193
04:27:41.990 –> 04:27:46.689
Julia Nimchinski: Thank you so much, and we are transitioning to our next session.