00:03:25 · Kieran Flanagan
[00:03:25] S0: Yes. I'll break into three buckets because it's kinda complex in how... So we're a multi product, very large company. So we started with what we would call, like, marketing... We call them like auto prospecting emails. And so basically, when you come in through a high intent signal, you would get a triggered email from marketing. And we started to, like, really optimize their prompts based upon what the best sellers... The very best sellers based upon the language they would use and how they would do outreach. And one of the interesting things about that is I had had this rule forever that any new experiment... And this will change based on the size of the company. Any new experiment had to be proven within six weeks, and you had to demonstrate whether that was something that was worth doing or not worth pursuing. And so if you took that very first experiment with how do we enrich these emails with AI and personalize them for the individual, like forever, b to b marketing has been personalized based upon segments, and segments is, oh, you're a company size, you're a role, you're in this vertical, and we call that personalization. Whereas AI allows you to really personalize to the individual if you have good data. That took us really six months to see a huge improvement. And so if I had kept my six week time frame, we would have missed all of that upside, and so I would have been wrong. And the upside was huge. It was like three, five, 10 x improvements in conversion rate. We lost during that period of time 80% of our organic traffic because of changes with AI and Google, and we still grew revenue and demand because of these improvements we were getting from conversion rates. In terms of the prompts, there's a ton of iteration on... In terms of how you did prompts. Again, this was maybe a year ago, and so prompting has changed. But one of the main things they did was... Or we did was sit with the sales team, learn how they write, and then had individual prompts for each part of the email, the subject line, the hook, the kind of value offer, and then the call to action. And we iterated on them like module blocks. And so that was our first big win. I can pause there, see if have any questions. And then there's two other kind of building blocks that we went on to after kind of getting those initial upside.
00:05:38 · Guillaume Moubeche
[00:05:38] S1: Yeah. I'd love to get into, like... You you just said that you you lost, like, a huge chunk of traffic with Google. Like, can you maybe explain like how did that happen and what were the consequences? Was it just traffic or also like sign ups? And will the conversion like help grow because the sign up quality was higher or do you think it's a mix of both?
00:06:02 · Kieran Flanagan
[00:06:02] S0: So, yeah, we... So Kieran and I, my good friend, have actually just... We we launched the book last week called Loop. And Loop is a new marketing methodology, and it was created based on the back of us having to completely redo how we did marketing, rethink how we did marketing. And so most people know HubSpot for our our original, like, inbound marketing methodology. And inbound marketing was, hey. Like, outbound is kind of a sucky way to do marketing, and you as a marketer probably don't even enjoy doing that. Inbound, you can create a lot of valuable content. You can really help your prospects and customers and attract them to your website and convert them to customers. Now what happened was Google started this way before AI was like the thing. They started doing these things called like feature snippets and knowledge boxes, where they answered some questions, some small selection of questions on the homepage. And it was pretty small in terms of impact. We had done a bunch of analysis back then, and we could see when one of these featured boxes was on the... In the search results, it reduced the click through rate on that search result by about 30%. So 30% less people will click on any result. And this is when the kind of no click phenomenon started, which is, hey. I search. I get my answers, I never have to click on something. And then AI rapidly accelerated that. So Google moved into AI overview. So September... Twenty it twenty four, actually, it was, like, March, May 2024, Google had started to update its algorithm and started to integrate a lot more AI functionality. And we just... At that time, we were a huge content site. So you you have to imagine, even though we were a b to b company, we would get more traffic at TechCrunch. We would get more traffic than all of these large media websites. And within a single year, we lost a 140,000,000 visits. So 80% of all of our organic traffic. And we used to have this methodology, Kip and I, which is like every problem could be solved by volume, which means that we were tracking behind revenue targets or whatever was happening, we just had to figure out how to get more traffic sign ups and leads.
00:08:03 · Kieran Flanagan, Guillaume Moubeche
And we shifted to everything can be solved with value, which is we're going to get less over time, certainly less traffic that we can track, but we are gonna extract way more value because AI allows us to improve our conversion rates. And that led into us rethinking the entirety of how we did marketing, and then we published a book called Loop that came out last week about how we did that. [00:08:25] S1: Nice. Can you share some tips on on how you you did it?
00:08:29 · Kieran Flanagan
[00:08:29] S0: Yeah. So I can give you a little, like, introduction to the methodology, and then we can go into any part. And so the methodology is really built around AI. So key principles of how you do marketing in an AI world. Obviously, how you work with AI will change because the models and AI is accelerating at such an incredible pace. But there are four key stages, and it's really for how you work. The first stage is our favorite stage, certainly my favorite stage. It is express. And express is taste and judgment have never been so important. And so why is that? Like, why do we say that? Why is taste and judgment such an important part of marketing today? If you actually go back over the last a hundred years in marketing, the hardest part of any marketer's job was making things. Making things was really hard. I had to sit down and create a web page, a landing page, an email, social media post, whatever it may be. So the entirety of a marketer's focus was like, how do I get enough hours in the day to make some things that will attract my prospects and convert them into customers? And you fast forward to AI today, that is no longer the case. Right? The making is not the hard part. AI can make an infinite amount of things. So the problem has really changed to what should I make? And what should you make is really based upon the taste and judgment to understand what would resonate with your customers. And the way we teach that in the book and the express stage and what we did in HubSpot is too many people outsource their taste to AI. And so what do I mean by that? Outsource your taste means I am a marketer, and I go to Claude or ChatGPT, and I say, create me a a web page with this kind of look and this kind of feel. Well, the AI goes and looks at all of the things it's been trained on and gives you the average of the taste of those things. So it's not your taste. It's the average of everyone else's. And so that's where all of your competitors live. It's the sea of sameness. And we teach how to create a taste profile. So what you do is you in... You basically teach AI what is unique about your taste, unique about your customers, how they speak to you, their language. You teach it what resonates with your customers, what stops them scrolling on social, what actually causes them to convert on something.
00:10:53 · Kieran Flanagan
And so you teach it these things, and you create this artifact that is updated each and every month based upon the things you're learning about your prospects, your customers. And every time you work with AI, it reads that taste profile and then creates things that are uniquely you and only could be produced by you. So that's kind of the first part that we did is, like, we instrumented taste into the AI assistance and agents that we use to make sure that we were creating things that were differentiated from everything else. So the AI became a real useful and powerful tool because of that. And so that's the express stage. I can pause there, and then I can go into the three others and give you a quick introduction, but I wanna see if that... If you had any questions on that.