George Xing’s Post

Hey LinkedIn, I'm very excited to share that I'm restarting my data blog (How to Data) after a 3.5 year hiatus! For context, I originally started this blog in early COVID days to share some of my learnings from building the data and analytics teams at Lyft with other data practitioners in a more scalable way. As fate would have it, 6 months later I started a startup and I decided to put writing on hold. A couple months ago, our team joined Stripe where I'm currently building data products as a product manager. Hanging up the founder hat means I have some breathing room, and I'm looking to pick up where I left off. My goal with the blog is still the same: practical insights for data practitioners and teams. What's different is that in addition to a pure data leader lens, I now have a data founder, investor, and PM perspective on some of the same problems – which I'll bring into the content. And yes, I'll talk about AI and how that changes things 😊 I would be deeply appreciative if you 1/ subscribe 2/ share with folks in your networks who would be interested in reading 3/ leave comments and feedback on topics you've like me to cover. Thank you!

Who is coming to Stripe Sessions this week? Come by say hi. I’ll be around to chat about all things Stripe and data / reporting. If you are in town but not registered yet, I have some free last-minute tickets. DM me! https://lnkd.in/eGCBMH8Q

I’m excited to share that Supaglue is joining Stripe! Over the last 3 years, we've been on a journey to help customers drive better business outcomes with data. In the process, we've built data platforms, integrations infrastructure, and workflow tools for data teams, marketers, and developers. I’m incredibly proud of the work our team has done to help customers integrate, unify, and take action on their data. After having a front-row seat to all the ways data improves both internal decision-making and user-facing product experiences, we’re more bullish than ever on the impact of data on businesses and the customers they serve. Today, we’re starting a new chapter in our journey by joining Stripe. Our team will help operate and scale the reporting platform and data services for Stripe’s Revenue and Finance Automation (RFA) suite of products. Stripe is a perfect match for Supaglue: both companies share a passion for building tools that reduce complexity for developers and data-centric products that grow revenue for businesses. Lastly, a huge thank you to all of our customers, investors, team members, and family and friends. We wouldn't have made it this far without your support! ❤️ https://lnkd.in/gZHj93cB

It's 2023, and LinkedIn is the best distribution channel for B2B SaaS products for GTM teams. Your buyers (not just the end users) are on LinkedIn, your audience is on LinkedIn, you build more authentic connections on LinkedIn. If you're reading this, you're on LinkedIn right now! But still, LinkedIn automation is a 2nd-class experience in the vast majority of sales engagement tools. Sure, LinkedIn locks down its APIs so it's not easy to automate connecting or DM-ing a prospect on LinkedIn the same way you can send a cold email. Some companies have built browser extensions, a semi-automated approach that still requires reps to take manual action or their laptops to be open. There's a few companies (Dripify, Salesflow.io) that do full LinkedIn automation (send messages, invite connections, read profiles), but they lack a multi-channel approach. We all know that LinkedIn is a great channel. With email decreasing in effectiveness (not to mention the Google restrictions that go into effect in Q1 2024), Linkedin is only going to be more important. The question is, when are sales engagement tools going to invest in first-class LinkedIn automation?

One of the biggest mistakes companies make when buying software is underweighting the importance of long-term roadmap alignment with their vendors in their buying decision. Here's why:👇 People buy software to solve an immediate problem they have. And most of the time, they do choose a solution that solves that problem, assuming they've done their homework and due diligence. When there are multiple options to choose from, a common tiebreaker is price (who is the most cost-effective) or maturity (which company has been around the longest). In some situations, this framework makes sense. The offerings are sufficiently commoditized. I admit, this is how we chose our SOC 2 vendor and our accounting service. I know both will do the job I need them to do 😄 But more often than not, business needs changes. New product requirements come up as companies grow, and companies need their vendors to do more than solve the initial problem. In these situations, there are outsized mutual benefits from having tight roadmap alignment over a multi-year time horizon! The ideal outcome for companies and their software vendors is to grow together. There are so many well-known stories of this happening. Snowflake scaling with Capital One. Stripe growing with Lyft. A symbiotic customer-vendor relationship where the customer helps the vendor prove out new products, and the vendor helps the customer scale faster than it could by itself. A relationship that lasts for more than a decade in some cases. Roadmap alignment is especially important when the vendor is providing core infrastructure. While replacing a productivity tool may be annoying, ripping out deeply embedded core infrastructure is way more painful. The replacement cost can take quarters or even years! Customers can get ahead of this by asking vendors for their product roadmap during the sales discovery process. Because by the time you're 6 months in and realize the mismatched expectations, it's already too late. As a product integrations platform (which falls into the infrastructure bucket), we try our best to make this process easier for potential customers: 1. We make it clear in sales calls that we're focused on serving companies building sales and GTM software. 2. For these customers, we help customers understand how they can solve their immediate integration use cases with our platform, but also their future integration use cases as they grow. 3. We proactively share what we are able to do today, what's on our roadmap, what we cannot do but will eventually support, and what will probably never do. Like Snowflake and Stripe, we hope to work with our customers for many years, and roadmap alignment is a big part of being able to play the long game and succeed together. TLDR: next time you're buying software, ask about the vendor's product roadmap!

Every company building B2B data-centric product for GTM teams either dies, or lives long enough to become a customer data platform (CDP). Here's why: 👇 At their core, B2B products for GTM teams do a few key things: 1. Ingest and centralize customer data from different sources. 2. Clean, enrich, transform the data to make it meaningful 3. Take action against systems of engagement or systems of record. From a product standpoint, this translates into: 1. Product integrations that ingest and sync customer data into a central database. Typically CRM + other data sources. 2. Proprietary models or 3rd party APIs that enrich and generate insights from the data. Often includes identity resolution, ML/AI, and analytics. 3. Product integrations that help end users take action based on insights, rules, and other business logic. Often executed in the context of a workflow. Which also describes the basic architecture of a CDP. To qualify as a CDP, products need to support broader use cases than simply sales engagement, product analytics, marketing automation, or customer success automation – which means more source and destination integrations... Of course, B2B companies that do well naturally expand to cover the entire customer lifecycle - to expand TAM, go after more product use cases, target more customers, and unlock more revenue. None of this is new or surprising. What's interesting today is that every company is also applying generative AI / LLMs in step 2. Companies are delivering new customer value from ingesting unstructured customer data that was previously ignored, because it was too messy to deal with. Things like documents, support tickets, call transcripts from places like Box, Zendesk, and Gong. They're also generating new unstructured content (notes, emails) that previously was not possible to do efficiently (at scale), to the benefit of players like Apollo.io and Outreach. More to come on the rise of the "AI-native CDP"! As an integrations company, it's fun seeing this shift happen in real-time 😄

We're excited to share our customer case study with Warmly,! Learn about how they scaled their CRM and sales engagement integrations from 0-100+ customers in 6 months using Supaglue and saved months of engineering time in the process (link in comments). 👇 Warmly is an autonomous sales orchestration platform that helps sales and marketing teams convert website visitors instantly. They recently announced their Series A financing, and we're glad we to be supporting their growth through powering their CRM and sales engagement integrations. Some highlights: ⏰ Reducing engineering time spent on integrations by 75% 💰 Turning their product integrations into a competitive advantage 🚀 Closing new customers faster and scaling to the enterprise Separately, it's been great working with the Warmly team! maximus greenwald Carina Boo 🎆 Alan Zhao Daniel Wallace Alexander Batis Marcos Cury

The green field opportunity for B2B sales and marketing software is creating their proprietary dataset. Here's why: 👇 Every company building sales and marketing software is using AI — that's a given. And under the hood, most of them leverage the same LLMs (read: OpenAI) to generate personalized emails. But the models are only as good as the input data. For most companies, the input data comes from the same data sources — CRM notes, support tickets, call transcripts. We know this because we're powering many of these integrations! Beyond that, some companies have access to 3rd party data for lead and account enrichment. And historically, companies like Apollo found great success from building a proprietary database of LinkedIn accounts and contacts. But today, there are many more data providers offering search and enrichment APIs. Companies can and do easily act as resellers or even aggregators of Apollo, Zoominfo, 6Sense, People Data Labs, etc. If "traditional" data enrichment is becoming a level playing field, what's the greener pasture companies should run to? The green field opportunity today for B2B companies (and what LLMs make possible) is the long tail of publicly available, unstructured data. Scrape websites, job postings, and other publicly available, potentially vertical-specific datasets that previously were too difficult to parse or turn into useful information. If you're building a proprietary dataset, move quickly! This is a classic first-mover advantage situation. Also, we'd love to chat. 😁

B2B SaaS companies get customer requests for integrations with all kinds of tools they already use, but not every SaaS has a well-documented API you can integrate with... 👇 Some products lack public API endpoints for key features, others don't offer a public API at all. Often the integration is a nice-to-have, but sometimes it's an actual dealbreaker. For example, if your product sends personalized emails through your customers' sales engagement platform, you need to be able to create email templates in those platforms programmatically. Outreach and Salesloft expose this in their public APIs and make this relatively straightforward (you still need to unify the data models yourself), but many sales engagement platforms don't. Apollo.io's public API provides a very limited set of functionality – for example, you can add prospects to existing sequences, but you cannot create new sequences or email templates yourself. HubSpot makes it even harder. They have well-documented public APIs for their CRM product, but they don't expose any functionality for Sequences, their sales engagement product. Customers have been requesting Groove, a Clari Company as an integration recently after its acquisition by Clari. But Groove doesn't have a public API at all! When it comes to offering a public API, it's usually a matter of when, not if. But timing doesn't always line up! And unfortunately, it doesn't help companies close deals or deliver the integrations their customers want. So what to do? There are several approaches ranging from web-scraping to building browser extensions that manipulate the DOM, but the most effective approach we've seen is reverse-engineering private, undocumented APIs. It can work well, but there's a significant cost as well: you have to accept the fact that private APIs can disappear and change unexpectedly, without notice. In practice, few companies put in the effort to reverse-engineer APIs due to the maintenance tax and the upfront time involved in doing research and trial-and-error. Startups are resource-strapped and should not be in the business of building and maintaining product integrations, let alone product integrations with undocumented APIs. At Supaglue, integrations is our core competency and continuous API research is just part of the job, so it makes sense for us to pay the maintenance tax of undocumented APIs. As a result, we're able to offer some features in our API that you might not find building a direct integration yourself...😁 If this is interesting, DM me or book some time on my calendar and I'll be happy to share more!