00:16:06 · Sandy Diao
[00:16:06] S0: Yeah. It's a great question. So I would actually share this. Our outbound worked with two different segments. And I think the first segment is unique to us because at the time, Pinterest owned distribution, in the sense that outbound was actually a house channel, meaning that we had this list of consumers that were already somewhat prospected in the sense that if you were a business, you know, you either sign up for a business account and or you maybe sign up for a consumer account because, you know, you made a mistake or you signed up too early, and then you have a website listed, and you have, you know, a lot of repin content. Right? So there's a lot of prospecting we can do within our within our own database. So that was more like product led sales. Right? And that one was a little bit more straightforward. We own first party signals around who was actually performing well, and we can enrich that with information. At the time, the more traditional way of enriching is you buy lists and you buy data, and you enrich it a little bit manually, although there's some, you know, ways of coding that in and automating it. But you could get things like, what is their total marketing budget? Or, you know, the things of that sort that kinda give you more signal. The other type of outbound that we did was realizing that social media advertising, especially for that discovery stage, the brands out there, think about, like, you know, the Coca-Cola of the world or the Burt's Bees, maybe products and categories that are more of a fit for Pinterest. They were actually spending on other social platforms, maybe doing the same thing that they would be doing on Pinterest. So like Facebook and Twitter at the time. So one of the things that we did was we actually took a list of some of those top target accounts, and we would actually look at the source code on their sites to see if they had a Facebook pixel at the time, to see if they had spend allocated on those platforms. And then we would add that as an additional high intent layer of signal to say, let's also reach out to these individuals and have our sales team reach out to them. So it was a combination of that product led sales as well as that outbound based on people who are actually spending money on other competitive platforms.
00:18:02 · Guillaume Moubeche
[00:18:02] S1: Nice. Super interesting. I think... I'm not sure if you wrote this or if you said this in an interview, but it was... You you said that you were data inspired and not data driven. Can you elaborate on this?
00:18:17 · Sandy Diao
[00:18:17] S0: Yeah. So Pinterest was a great example of it, but I can also point to a lot of the work that I did at Descript in the early early days. When we're building growth teams, at least in my experience, when I've been building growth teams, oftentimes, there's kind of this push, especially if you're working with very data minded people. And, you know, for example, like data scientists, data engineers, data specialists, etcetera. There's kind of this push to be very precise about things like attribution, or measurement, or experimentation. And don't get me wrong, those things are all super, super important. The thing I've learned best in all these early stage growth scenarios is that when you try to push too hard for accuracy in the early days, you add a lot of bottleneck and time cost to what it is that you're doing, when in reality, what you're looking for is directional data. So a lot of what I mentioned with Pinterest, for example, going back to the insight around, hey, maybe businesses are using Pinterest in addition to consumers. If we spend a lot more... We probably could have spent a lot more time validating that insight. Right? Probably shouldn't just trust me looking at, you know, 500 tickets and saying, oh, there's businesses writing in. Let's fork the sign up flow. Because you could have major impact on the total percentage of consumers that sign up there. And so, you know, I I think one of the most important things I learned was that we should, you know, test all of that early on and use the data as an, you know, inspirational road map for figuring out what we should be doing instead of trying to get precision on every single insight and essentially placing this, like, really, really heavy investment to validate all of it before testing everything. And it worked really well at Pinterest. It continued to work well in my, you know, other growth roles at Indiegogo, One Smart Piano, Descript, and a lot of the teams that I work with as well. So when I say data inspired, I mean that we're looking for directional insights, especially when we're just building the growth engine. We're not looking for precision. We're not trying to solve attribution. You know, companies will continue to have evolving attribution models.
00:20:17 · Sandy Diao, Guillaume Moubeche
So that was really, I think, the genesis of how I was able to help a lot of these growth efforts and growth teams move relatively quickly in the early stages. [00:20:26] S1: And have you seen... Or do you have an example where, like, the data you got was actually, like, misleading?
00:20:35 · Sandy Diao
[00:20:35] S0: Yeah. Well, actually, a great example that I'll share is when I was at Descript, one of the decisions that we made was around trying to figure out how we could revamp our pricing. So Descript, for those who aren't familiar, is a audio video editing platform. And its innovation is that it turns content into a text based transcript. So you could edit videos the same way that you edit a Google Doc. And the way that we traditionally priced was by essentially words. Right? Like transcription of that content. So you get a certain number of hours of transcription. And we had this idea of, well, what if people don't understand transcription as a value metric? Why don't we just, you know, essentially give the thing for free to everybody, you know, essentially take those premium features away at some point so that they get a taste of it, and they're gonna really, like, feel the pain of missing out on it? Essentially, what is called a reverse trial. Right? And we thought it was a great idea because we looked at everybody in the space, every prosumer, consumer SaaS company, and everybody was doing reverse trials. So we thought, this thing must be working because everyone's doing this. So a lot of the, I think, the external benchmarking data firstly told us that this was going to work. The second thing that we did was we surveyed our users, and we asked them, like, do you understand what it is that you're paying for? Or, you know, we asked the better version of that, which is what are you paying for? And nobody said transcription. They just said they were paying for editing. So we thought, the data is pointing to basically all the data, external and internal, is pointing to we gotta get rid of this paywall. So we ran an experiment where we shifted our pricing model to a reverse trial. Lo and behold, I wake up the next day after the experiment, and every single user in the exposure group, not every single user, but the conversion rates, you know, basically halved slash worse, you know, over time. And, you know, I was just totally panicking. It's like we spent all this time, you know, optimizing all these different aspects of our pricing and packaging and our free to paid motion, and then all of a sudden, you know, this exposure group is like really, really suffering. What do we do now?
00:22:39 · Sandy Diao
I I think a lot of that, you know, early insight data really pointed us in this direction of saying this is going to work. But I think the lived reality is that when you actually run the experiment, users behave very, very differently. When I kind of reflect on that experience, our biggest learning here is that different types of users have different expectations about the value of these different pricing models. What we found when we segmented the data is if is that businesses or more of that b to b customer had value from that reverse trial because they could essentially kind of leisurely go back to it and test it out over time. And they didn't really have this, like, day to day DAU, WAU distinct need for it. So it felt more aligned with their value model. But for a lot of the consumer, prosumer creator type users using the platform, it didn't make sense to them. Right? Because they just didn't have the two week bandwidth or timeline in their minds for making a decision around a video editing tool. They're like, I need to cut a TikTok short right now. I need to cut a YouTube video right now. It's like, okay. I'll get this thing done, and then I'll disappear, I'll never come back. It doesn't... It didn't really help reinforce the idea of the value of the product. So, you know, we ultimately ended up reverting back to our original pricing model, but that's an example of how that early data really pointed us in the wrong direction.