WEBVTT 00:00:00.000 --> 00:00:07.000 And he is going to be speaking about users, pain points, and intent. 00:00:07.000 --> 00:00:12.000 And he does AI search and SEO at AT&T. 00:00:12.000 --> 00:00:14.000 So I'm excited to hear what they're up to. 00:00:14.000 --> 00:00:19.000 [Applause] 00:00:19.000 --> 00:00:22.000 Thank you. Thank you, everyone. 00:00:22.000 --> 00:00:28.000 I will make sure this 20 minutes is valid. 00:00:28.000 --> 00:00:32.000 So, these are the things I'm going to talk today. 00:00:32.000 --> 00:00:37.000 I'm going to talk about volume and talk about users, 00:00:37.000 --> 00:00:40.000 pain points, and intent. 00:00:40.000 --> 00:00:44.000 Why it's ten times more important than blindly following volume. 00:00:44.000 --> 00:00:48.000 And of course, because my co-speakers did a great job explaining that, 00:00:48.000 --> 00:00:52.000 I'm going to share a little bit more on that. 00:00:52.000 --> 00:00:55.733 So user first, algorithm second. 00:00:55.733 --> 00:01:00.433 Again, I don't want to chase algorithms because no one really 00:01:00.433 --> 00:01:01.667 knows how [inaudible] 00:01:04.667 --> 00:01:06.667 No one knows how it works. 00:01:06.667 --> 00:01:07.667 [inaudible] 00:01:07.667 --> 00:01:13.633 OpenAI, Claude, whatever, and they launch every [inaudible] 00:01:13.633 --> 00:01:16.867 So the retrieval method is different. 00:01:16.867 --> 00:01:20.567 So do I want to chase algorithms or models or [inaudible], 00:01:20.567 --> 00:01:25.567 or, for example, [inaudible], 00:01:25.567 --> 00:01:27.567 so do I want to care about that? 00:01:27.567 --> 00:01:30.567 Or do I want to care about users? 00:01:32.567 --> 00:01:36.567 So simple, I think I'm going to talk about the thesis, 00:01:36.567 --> 00:01:40.567 two case studies I implemented with a great team, 00:01:40.567 --> 00:01:43.567 and these are real case studies, 00:01:43.567 --> 00:01:46.567 and third about how I bought a new car, 00:01:46.567 --> 00:01:52.633 and what was my journey from point A to purchasing. 00:01:52.633 --> 00:01:56.733 And let's talk about AI and the method. 00:01:56.733 --> 00:02:01.067 And the key important thing here is building something 00:02:01.067 --> 00:02:03.100 for myself is easy. 00:02:03.100 --> 00:02:04.867 When you're working for an enterprise, 00:02:04.867 --> 00:02:08.367 you need to build a system at scale that can 00:02:08.367 --> 00:02:11.767 be used by multiple users. 00:02:11.767 --> 00:02:14.167 Thank you. 00:02:14.167 --> 00:02:15.167 [inaudible] 00:02:15.167 --> 00:02:20.833 So, zero volume is not a zero demand. 00:02:20.833 --> 00:02:22.900 So, trust ahead of sales. 00:02:22.900 --> 00:02:25.000 So, this is a case study. 00:02:25.000 --> 00:02:28.900 This is an industry leader site, and we want to run an experiment 00:02:28.900 --> 00:02:33.733 on using, we ignored search volume. 00:02:33.733 --> 00:02:38.133 We found that users want information that is evergreen, 00:02:38.133 --> 00:02:39.767 that's going to stay long. 00:02:39.767 --> 00:02:42.100 So example, let's take an example of area codes. 00:02:42.100 --> 00:02:45.900 People search for which number is coming from this location. 00:02:46.800 --> 00:02:49.033 That doesn't have any commercial value. 00:02:49.033 --> 00:02:50.600 And we know that for a fact. 00:02:50.600 --> 00:02:54.433 But we want to be, we want to build a content 00:02:54.433 --> 00:02:56.633 that is relevant for industry 00:02:56.633 --> 00:02:59.833 at scale using programmatic SEO strategies, 00:02:59.833 --> 00:03:03.600 that is going to be staying forever and earn trust. 00:03:03.600 --> 00:03:07.033 And we did an experiment exactly [inaudible] years ago. 00:03:07.033 --> 00:03:11.533 And again, I can't use the first-party data sources to show, 00:03:11.533 --> 00:03:15.100 but this is an estimate of the impact. 00:03:15.100 --> 00:03:18.500 We almost drove about half a million impressions, 00:03:18.500 --> 00:03:20.500 clicks, I don't really care about that. 00:03:20.500 --> 00:03:24.300 But the key thing is, people genuinely love the source, 00:03:24.300 --> 00:03:27.700 and we didn't follow the SEO strategy or AI strategy. 00:03:27.700 --> 00:03:31.867 Every AI platform, every search platform 00:03:31.867 --> 00:03:33.133 are citing our sources. 00:03:33.133 --> 00:03:36.033 We never tried to do that, but it happened. 00:03:36.033 --> 00:03:39.067 That's the impact when you focus on users 00:03:39.067 --> 00:03:41.100 and create a meaningful content 00:03:41.100 --> 00:03:44.633 and the AI citations, search rankings will follow those. 00:03:47.900 --> 00:03:55.900 Let's talk about case study two and when I joined the company, they have a decent amount of pages. 00:03:55.900 --> 00:04:01.900 We analyzed the user behavior and the bounce rate was significantly bad. 00:04:01.900 --> 00:04:10.900 So, rather than searching volumes, keyword data, all these things, I decided to think like a user. 00:04:10.900 --> 00:04:14.900 Okay, so I'm reading this content, I hate this content. 00:04:14.900 --> 00:04:19.900 Why? I don't want to read 3,000 words of a guide. 00:04:19.900 --> 00:04:25.900 Okay, in my feedback, I asked a couple of my team members and did a small test. 00:04:25.900 --> 00:04:27.900 And everyone said the exact same thing. 00:04:27.900 --> 00:04:30.900 Who wants to read 3,000 words of a guide? 00:04:30.900 --> 00:04:32.900 They want the videos. 00:04:32.900 --> 00:04:36.900 So we created the videos and injected into these pages. 00:04:36.900 --> 00:04:41.600 You can see the video snippets went up to 18,000. 00:04:41.600 --> 00:04:43.900 Again, that's for the metrics. 00:04:43.900 --> 00:04:44.667 I am happy. 00:04:44.667 --> 00:04:48.500 The metric I am happy about is users are happy because we 00:04:48.500 --> 00:04:51.333 are producing what they need. 00:04:51.333 --> 00:04:55.500 So, I'm a new dad, so I was trying to upgrade my car, 00:04:56.667 --> 00:04:59.733 [inaudible] 00:04:59.733 --> 00:05:03.733 and then also my family and safety. 00:05:03.733 --> 00:05:06.600 I decided to buy a Toyota Grand Highlander Hybrid. 00:05:07.633 --> 00:05:12.633 I was reading pages, web pages, and I hated it. 00:05:13.467 --> 00:05:15.633 Everyone is like, oh, this is the best, 00:05:15.633 --> 00:05:16.967 this is the best, this is the best. 00:05:16.967 --> 00:05:20.700 No, I don't want to buy a $60,000 product 00:05:20.700 --> 00:05:22.567 based on sales page. 00:05:22.567 --> 00:05:25.867 I genuinely want the users who bought the product 00:05:25.867 --> 00:05:27.133 and share their reviews. 00:05:27.133 --> 00:05:29.500 Then I landed on a Facebook group dedicated 00:05:29.500 --> 00:05:32.333 to people who bought Grand Highlanders. 00:05:32.333 --> 00:05:33.767 And they're sharing the perspective 00:05:33.767 --> 00:05:37.133 of which model to buy, why they like specific things, which 00:05:37.133 --> 00:05:39.167 dealership is best to buy from. 00:05:39.167 --> 00:05:43.333 And these are the queries we need to answer to our users. 00:05:43.333 --> 00:05:47.267 It's not about search volume, it's not about anything else. 00:05:47.267 --> 00:05:49.233 you need to find the real queries 00:05:49.233 --> 00:05:53.067 that people are actually typing in different platforms 00:05:53.067 --> 00:05:55.233 based on their persona or whatever it may be 00:05:55.233 --> 00:05:56.433 and try to answer them. 00:05:56.433 --> 00:05:58.800 If you can answer this query, 00:05:58.800 --> 00:06:01.267 many people lost my purchase 00:06:01.267 --> 00:06:04.000 because they were not answering the specific questions. 00:06:04.000 --> 00:06:06.833 So, again, if I use traditional tools, 00:06:06.833 --> 00:06:09.033 I'm not saying traditional tools are bad, 00:06:09.033 --> 00:06:11.267 but they're only good for direction. 00:06:11.267 --> 00:06:13.133 The volume is not a strategy. 00:06:14.033 --> 00:06:15.667 You understand the users 00:06:15.667 --> 00:06:17.767 and trying to build a content around that. 00:06:17.767 --> 00:06:19.833 And content doesn't mean that textual content. 00:06:19.833 --> 00:06:24.567 It can be video, audio, podcast, newsletters, whatever it may be. 00:06:24.567 --> 00:06:28.933 That's where real success happens. 00:06:28.933 --> 00:06:33.000 Again, I mainly, I don't have all the answers 00:06:33.000 --> 00:06:37.967 which I'm asking here, because keywords tracking was great. 00:06:37.967 --> 00:06:41.400 And Google used to be not personalizing results. 00:06:41.400 --> 00:06:45.400 And also, very short, one or two keywords. 00:06:45.400 --> 00:06:48.533 But no one is trying to search, like, Grand Highlander. 00:06:48.533 --> 00:06:53.600 People are searching, like, I want SUV with highest mileage 00:06:53.600 --> 00:06:56.167 and most relevant brand. 00:06:56.167 --> 00:07:01.133 So keywords are transforming into conversations. 00:07:01.133 --> 00:07:05.033 So I was trying to search this query in different platforms 00:07:05.033 --> 00:07:09.433 to understand the reviewers. 00:07:09.433 --> 00:07:14.633 [inaudible] is trying to pull the data from Facebook, Facebook 00:07:14.633 --> 00:07:21.233 groups, and then [inaudible] is trying to pull from Reddit. 00:07:21.233 --> 00:07:23.500 And Gemini, similar. 00:07:23.500 --> 00:07:26.633 Google [inaudible] from multiple sources. 00:07:26.633 --> 00:07:29.067 So here I'm not talking about how 00:07:29.067 --> 00:07:32.033 Reddit is citing my [inaudible] 00:07:32.033 --> 00:07:35.067 Who cares if you have a meaningful content? 00:07:35.067 --> 00:07:36.733 It's not about Reddit. 00:07:36.733 --> 00:07:37.733 It's not about YouTube. 00:07:37.733 --> 00:07:39.633 It's a moving target. 00:07:39.633 --> 00:07:43.900 You need to find where your users are spending the most time. 00:07:43.900 --> 00:07:48.200 So the method, I'm going to share the method. Here is how I do it. 00:07:48.200 --> 00:07:54.200 So these four slides will be the most important slides in this entire presentation. 00:07:54.200 --> 00:08:00.200 So we heavily focus on search data, but Google My Business reviews. 00:08:00.200 --> 00:08:09.200 There is a roofing company in New York and how I use the reviews to understand the pain points. 00:08:09.200 --> 00:08:12.200 If you analyze all the roofers in New York, 00:08:12.200 --> 00:08:16.200 analyze, classify the reviews and response rates and variance, 00:08:16.200 --> 00:08:19.200 then you will understand the pain point of the roofers. 00:08:19.200 --> 00:08:24.200 It can be expensive pricing, they are doing a bad job, or they don't serve [inaudible]. 00:08:24.200 --> 00:08:28.200 So if you can solve the pain point via content and SEO strategies, 00:08:28.200 --> 00:08:32.200 That's where the commercial traffic follows. 00:08:32.200 --> 00:08:36.200 Again, the biggest search engine of Google is probably YouTube. 00:08:36.200 --> 00:08:40.200 People talk, comment, respond, engage 00:08:40.200 --> 00:08:44.200 in YouTube, in YouTube platform. But are we pulling 00:08:44.200 --> 00:08:48.200 the content, classifying it to understand users, 00:08:48.200 --> 00:08:52.200 how they're behaving in YouTube. Again, Twitter 00:08:52.200 --> 00:08:56.200 competitors, Reddit. People know that TikTok 00:08:56.200 --> 00:09:01.700 specific age group of audience who love to spend more on TikTok. 00:09:01.700 --> 00:09:03.833 So they like short-form content. 00:09:03.833 --> 00:09:05.133 They want instant answers. 00:09:05.133 --> 00:09:10.833 Example, are we serving the product if the product is going to target them? 00:09:10.833 --> 00:09:17.233 Which is, we keep it as a classic example to understand what are the topics we need to cover. 00:09:17.233 --> 00:09:19.067 So, conference insights. 00:09:19.067 --> 00:09:24.900 I went to multiple conferences to understand missing topics. 00:09:24.900 --> 00:09:27.133 topics I don't know like this. 00:09:27.133 --> 00:09:29.733 I learned a lot from different speakers. 00:09:29.733 --> 00:09:32.800 So similarly, if you go to industry relevant conferences 00:09:32.800 --> 00:09:35.833 like [inaudible] or telecom or something like that, 00:09:35.833 --> 00:09:39.100 you'll know a lot from that. 00:09:39.100 --> 00:09:43.400 So this is pretty much public data influence in the podcast. 00:09:43.400 --> 00:09:46.333 The key thing is the earnings calls. 00:09:46.333 --> 00:09:49.300 I collect the data from earnings calls. 00:09:49.300 --> 00:09:50.700 Why? 00:09:50.700 --> 00:09:51.267 Am I mad? 00:09:51.267 --> 00:09:52.733 No. 00:09:52.733 --> 00:09:55.533 I know what's coming about my competitors. 00:09:55.533 --> 00:09:58.700 My competitor says, "We allocated this much on budget 00:09:58.700 --> 00:10:00.667 on X, Y, Z product." 00:10:00.667 --> 00:10:04.400 Then I know for a fact they're going to take our market share. 00:10:04.400 --> 00:10:07.867 Do I need to react or do I need to react 00:10:07.867 --> 00:10:09.700 after they took my market share? 00:10:09.700 --> 00:10:13.100 Or I create a patch and I tag them ahead. 00:10:13.100 --> 00:10:15.400 So that's where earnings calls come in. 00:10:15.400 --> 00:10:17.267 Job postings. 00:10:17.267 --> 00:10:19.667 If your competitor is hiring for example, 00:10:19.667 --> 00:10:22.933 like AEO, SEO, GEO, whatever it may be. 00:10:22.933 --> 00:10:24.800 Then, you know for a fact, they're 00:10:24.800 --> 00:10:28.300 investing money to take market share in AI space. 00:10:28.300 --> 00:10:32.733 So that's why you need to keep an eye on the job postings. 00:10:32.733 --> 00:10:36.200 Similarly, research papers and regulations. 00:10:36.200 --> 00:10:38.733 Research paper, why? 00:10:38.733 --> 00:10:41.333 People write research papers. 00:10:41.333 --> 00:10:45.400 They use a lot of data, certified data, actual data, 00:10:45.400 --> 00:10:47.300 in order to write papers. 00:10:47.300 --> 00:10:54.300 example. I built a tool, the open source tool. No email required, no subscription, no follow required. 00:10:54.300 --> 00:11:00.300 I'll make sure that anyone can access it. So, I used these datasets to build a personal tool. 00:11:00.300 --> 00:11:03.267 And there can be one product, different personas, 00:11:03.267 --> 00:11:06.900 different queries for the region. 00:11:06.900 --> 00:11:23.467 I analyzed 5.6 million queries, no one really cares if I say that, but if I classify that map with personas and build a product, then the returns would be, the returns in the graph speaks about 00:11:23.467 --> 00:11:23.900 that. 00:11:23.900 --> 00:11:32.900 And classification is the core thing. But I made myself classify the data because I know my industry better than [inaudible]. 00:11:32.900 --> 00:11:39.900 So, rather than topics, keywords, clusters, I try to [inaudible] the whole industry. 00:11:39.900 --> 00:11:46.900 Starting from the company, there are social media platforms. Wherever a person is talking about that particular thing, I try to collect the data. 00:11:46.900 --> 00:11:51.900 And this has datasets of zero search volumes. I don’t care about volume. 00:11:51.900 --> 00:11:58.900 I care about, is there one user that is going to buy my product? I'm going to answer them. 00:11:58.900 --> 00:12:06.900 So I built the whole cluster. For the next, probably one or two years, I created a self-learning mechanism 00:12:06.900 --> 00:12:10.900 where it will automatically schedule the crawls, collect the data, and update itself. 00:12:10.900 --> 00:12:15.900 Okay, then you go to "person who writes checks". 00:12:15.900 --> 00:12:19.900 [inaudible] “Okay man, you have 5.6 million queries. Why should I write a check for you?” 00:12:19.900 --> 00:12:26.900 Easy hack is: do you want to do market share for your company? 00:12:26.900 --> 00:12:29.900 The one question that you write a check for. 00:12:29.900 --> 00:12:33.900 Hey, do you want to invest double money to recover? 00:12:33.900 --> 00:12:36.900 Or do you want to invest money? Be ahead of them. 00:12:36.900 --> 00:12:39.900 Okay, this is one of my favorite slides in the whole thing. 00:12:40.867 --> 00:12:45.867 I was curious and this research is not funded by any organization except myself. 00:12:45.867 --> 00:12:52.867 I was analyzing 100,000 reviews on all the local telecom stores in Texas. 00:12:52.867 --> 00:12:57.867 What I find, I don't know yet, but what I found was amazing. 00:12:57.867 --> 00:13:02.867 So the second pain point was, I'm sorry, long wait time. 00:13:02.867 --> 00:13:07.867 The pain point of the users when they walk into the stores is that they hate waiting. 00:13:07.867 --> 00:13:15.867 None of the traditional tools tell a problem or pain points, but analyzing and classifying data tells that. 00:13:15.867 --> 00:13:19.867 So, long wait time. What I have to do with the long wait time? 00:13:19.867 --> 00:13:26.867 You can build a product such as digital check-ins, or you can build an in-app virtual appointments. 00:13:26.867 --> 00:13:32.867 That can bring revenue, reduce pain points, and help [inaudible] as well. 00:13:32.867 --> 00:13:41.867 So that's the beauty of you, collecting the data from different sources, you classifying them and building the products based on pain points. 00:13:42.900 --> 00:13:53.900 So, it's good, I did it all myself, and I have a great mentor who I work with. His name is Gene. He challenges me every time. 00:13:53.900 --> 00:13:59.900 Okay, this is great. How can this be used by 10,000 people at the same time? 00:13:59.900 --> 00:14:07.900 So I created a multi-level AI agent orchestration, where agents can talk to each other. 00:14:07.900 --> 00:14:18.900 Is it great? No, I'm still learning and fixing it, but I was able to build a great agents that can 60 to 70% do my job, and then I can verify that. 00:14:21.033 --> 00:14:27.033 So, I'm not going to talk about all these things because I want to use [inaudible] for the most important person in my life. 00:14:27.033 --> 00:14:30.033 It's my mom. 00:14:30.033 --> 00:14:36.033 So, she's the first person to believe in me. 00:14:36.033 --> 00:14:42.033 I came from India, a whole different country, did my masters. 00:14:42.033 --> 00:14:46.033 And I'm here, this is like a dream for me. 00:14:46.033 --> 00:14:50.033 But one person trusted me when I was zero, is my mom. 00:14:50.033 --> 00:14:58.033 is my mom. So, I would like to take this platform and give it to my mom. [inaudible] So, that's it. 00:14:58.067 --> 00:15:00.000 [Applause]