没有营销手册可抄：OpenAI 和 Stripe 首位营销负责人的 DATE 诊断法

Shankarraman was the first marketing hire at OpenAI, and host Lenny Rachitsky asks why a product growing so fast needed marketing at all. She says awareness was never the issue: everyone knew ChatGPT, but one zoom level deeper people did not know what to use it for, what it replaced, or whether they should use search instead. The work of marketing became creating a "use case epiphany" in which people realize ChatGPT can do something they did not expect.

She stresses that marketers should be diagnostic about what marketing can do rather than reaching for standard top-, middle- and bottom-of-funnel tactics. She also says product management and product marketing should work like a three-legged race from the start of development, not a handoff at the end. Customer insights and the language customers use then become a cheat code for positioning.

On the enterprise side, demand generation was not OpenAI's problem either. When her team turned on the contact-sales form for ChatGPT Enterprise, lead volume rose 40 times overnight, and she used ChatGPT to write a Python script that served as the first lead-scoring model, which stayed in production "way too long". She adds that pricing for AI remains unsolved because the value does not fit neatly into seat-based, SaaS or usage-based models.

Founders often ask Shankarraman how Stripe, Retool or OpenAI did it, and she hesitates to share details because context, competitors and the moment all shaped what worked. Copying the outputs of a strategy ignores the inputs and variables that led to it. Instead she tries to make people diagnosticians who can work out the right strategy for themselves.

Drawing on her engineering background, she lays out four steps. The first is diagnosing the real problem: if a founder wants a demand-gen leader, she asks about the funnel. If people convert once they are in the room, more top-of-funnel spend makes sense, but if they ask why it costs so much and how it compares with competitors, the work is in product-market fit, positioning and sales enablement.

The second step is analyzing competitors for a baseline and for gaps, without becoming so focused on them that you end up at a local maximum. The third is intentionally taking a different path, often by borrowing ideas from domains far from your own. The fourth is to experiment, test and validate, scale what works, and discard what does not, resisting sunk-cost bias and giving teams psychological safety to fail. She names it DATE: Diagnose, Analyze, Take a different path, Experiment.

On differentiation, she warns that being cheaper is a race to the bottom, especially as models get cheaper and more capable. Novelty matters not for its own sake but because people want something aligned with their values and goals. A strong product experience paired with a matching marketing experience can be a superpower.

Shankarraman contrasts Retool with Stripe and OpenAI, which were inbound companies with latent demand. At Retool, awareness was the challenge, so the team had to build outbound demand engines. Rather than simply scaling paid channels, they diagnosed which leads became sales-qualified opportunities and found that paid social did little. She calls impressions, clicks and views vanity metrics and says the numbers that matter are signups, sales leads, pipeline and revenue.

Competitors were investing in content and events, but Retool chose to double down on customer marketing and storytelling because it had real enterprise customers who were paying and expanding, which copycats could not replicate. The team then tested webinars, sales dinners and event formats, scaling those that worked. Lenny's takeaway is to look at what is already working, and Shankarraman adds that you should scrutinize channels that look good but do not drive pipeline.

At Stripe, the competitor for Stripe Connect was companies becoming payment facilitators themselves, and consultancies helping with that leaned on legacy jargon. Stripe wrote a "reverse RFP" piece: here is the secret playbook for becoming a payment facilitator, and if it feels onerous, it is, so use Stripe Connect instead. That let it rank for the search terms people used without positioning itself as a PayFac solution.

Shankarraman spent almost eight and a half years at Stripe, three of them as the only marketer, which she says she does not recommend. She jokes that Stripe's founders John and Patrick were the real first marketers because, as developers, they knew how to reach developers authentically. She had to unlearn lessons from Google and Dropbox, and she learned that developers spot bugs in marketing as well as code, so marketing is an extension of the product and needs high craft and design polish.

She describes chapters of marketing at Stripe. First came a secret backlog of shipped features that had never been communicated, which taught the company that a launch is not complete when the code is complete. Next, launches expanded beyond a blog post to email, other channels, developer relations and developer experience, and then came helping people navigate a multi-product platform. In hyper-growth companies she says to rerun the diagnostic every three to six months and build T-shaped teams that can adapt.

All new Stripe hires did a support rotation, and the company spent about 20% of its time talking to customers and non-users. Questions such as whether Stripe handles subscriptions or can pay people out showed that features existed but were not explained anywhere. They became a stack-ranked backlog of educational landing pages, because for developers a good funnel often never involves talking to sales.

Shankarraman calls good process a hill she would die on, because sufficient process speeds companies up. Marketing is the first touchpoint of the product, so consistency matters. Process also lets new hires be as effective in week two as someone with two years of social capital who knows which conference room to stand outside to get the founder to review something, which she calls unscalable.

She suggests two easy steps. Create a transparent marketing review forum, a weekly meeting, Slack channel or email alias, so others learn from what reviewers look at. And keep checkpoints to two: a 20% review of strategy, audience and approach, and an 80% review when artifacts exist but changes are still possible. A review at 99% is just a rubber stamp and worthless.

On the move-fast objection, she says velocity and brand are interconnected. The brand is not just marketing artifacts but the product, how support resolves tickets and the candidate experience, and understanding it lets the organization move faster. Lenny adds that a strong, trusted brand makes everything easier, such as a new Stripe or OpenAI launch, and Shankarraman says that responsibility sometimes means putting the brakes on to improve quality and safety.

OpenAI had been around for almost a decade as a research organization and launched ChatGPT about a year before Shankarraman joined. Asked when to hire a marketing leader, she says the first criterion is strong product-market fit, because marketing is fuel on a fire. The pillars include product marketing for launch excellence and positioning, demand generation for sales-driven journeys, and brand and community.

She distinguishes capital-M Marketing, the team responsible for channels and funnel engines, from lowercase marketing: what the company stands for, the founder's story to press and the business community, product thinking about what to market, and sales defining the right ICP. It is a whole-company motion rather than one team's job.

Lenny asks why ChatGPT dominates mindshare even if Claude may be better at code and writing. Shankarraman says we are in the early innings of a paradigm shift, and users find products delightful for the same reasons as ever, with loyalty built when the gap between expectation and reality is small. She hopes AI companies think long-term about AI as a positive force for humanity, which needs change management and collaboration, not just a race over the best chatbot.

At OpenAI, she was surprised by the warmth and intellectual curiosity of colleagues and by the sincerity of the mission, while the challenge was being in the eye of the hurricane, with intense scrutiny. Her test for a job, borrowed from Stripe's COO Claire Hughes Johnson, is people, product and potential: people who push your thinking, a product you believe in and potential for your discipline to make an impact.

Shankarraman now is an executive in residence at Thrive Capital, helping portfolio companies with marketing needs. This can mean serving as interim CMO, pressure-testing strategy and growth targets, or reviewing a landing page the day before launch, across companies from unincorporated startups to Databricks, Stripe and OpenAI. The experience reinforced her view that a marketer's best contribution is adaptability: going deep into each company's context rather than applying playbooks.

She describes the "chameleon CMO". Conventional wisdom says to be T-shaped and deep in one pillar such as product, demand or brand marketing, but modern leaders need to be good at many things: analytical, close to data science, creative in service of the buyer journey and revenue goals, and aligned with sales and product. She suggests moving from T-shaped to comb-shaped so you can go deep where the diagnostic says it is needed.

AI can make this easier. A brand marketer who dislikes giant CSV dumps can use ChatGPT as a nonjudgmental partner to understand data, while an analytical growth marketer can use it as a brainstorming partner on creativity.

Asked about taste and creativity, Shankarraman says taste will matter more because AI makes a lot of drivel available to anyone. The companies that stand out will show craft, real understanding of the product and customer, and a meaningful link between them. She prefers AI that augments effort rather than replaces it. Lenny mentions the idea, from Guillermo at Vercel, of of "exposure hours" to great work, and she says Thrive has a channel for sharing things that resonated.

On early-career people relying on AI, she believes the discipline itself is changing, so not using the tools is a disadvantage. But she still strongly believes in STEM education so people understand the fundamental concepts and can choose how to apply them. A learning mindset, rather than going to school just for grades, is essential, and she and Lenny connect this to curiosity, which Shopify's Toby Lutke named as the most important thing to nurture in a child.

On pricing an AI product, Shankarraman says again that there is no playbook and the real answer is experimentation. At Retool, the team made the self-hosted version, previously gated behind sales, available self-serve. That reduced the pipeline the sales team saw but helped them focus on higher-ACV deals. The trade-off was hard, and the team built conviction through experiments and pilots.

The reasoning was twofold. Smaller customers in regulated industries or with private data wanted self-hosting too, so it was not only an enterprise need, and sales should talk to larger customers about larger deals. For AI, she suggests testing whether value is in seats, hours saved or things now possible, and says that with agents, paying for AI workers and the unit of completion are open questions in a "Wild West".

On AI in her own work, she says it helps Thrive talk to dozens of portfolio companies a week through internal AI tools that surface institutional knowledge, and that even at OpenAI she advised enterprises to invest first in their own operational efficiency. As a failure, she cites Stripe Relay, launched in 2014 as a way for e-commerce companies to tap social commerce through buy buttons. It was ahead of its market and failed, because the team had not done enough user research on market dynamics and alternatives.

Her final message is that no single answer exists for marketing problems. The work is hard, you must spend the hours understanding your customer, and AI will not replace that, nor deep understanding of your product and company values. In a lightning round, she recommends April Dunford's *Obviously Awesome*, shares her motto that the gap between expectations and reality drives unhappiness, and says she has reduced but not given up em dashes.