Amber Chu’s Digital Corner

Growth Hacking: What the Book Gets Right — And What Small Teams Actually Need to Know

Fri, 25 Jan 2019 00:00:00 GMT

📖 Reading time: ~11 minutes

💡 Quick Answer Growth hacking is a data-driven approach to finding your fastest, cheapest path to user growth — originally developed in Silicon Valley startups but applicable to teams of any size. The core tools that actually work without an engineering team: finding your product's "aha moment," defining a single North Star Metric, and running small structured experiments before committing to any channel. Here's what the book teaches well and where it needs translating for real-world use.

The Client Who Couldn't Figure Out Why Growth Had Stalled

A few years back I was working with a small e-commerce brand — three people, doing decent revenue, stuck. Traffic was coming in. Sales were happening. But growth had plateaued in a way nobody could explain, and the founder's instinct was to spend more on ads.

Before recommending anything, I asked her one question: what does a user do in their first week that predicts whether they'll come back?

She didn't know. She had revenue data and traffic data but no understanding of what moment — what specific action — turned a new visitor into a customer who returned. She was optimising the top and bottom of her funnel without knowing what was happening in the middle.

That question — and the process of answering it — is essentially what Growth Hacking by Sean Ellis and Morgan Brown is about. Ellis coined the term "growth hacker" and built the methodology behind it at companies like Dropbox and Eventbrite before writing this down. The book formalises an approach that had previously been practiced informally at a handful of Silicon Valley companies and makes it accessible to everyone else.

Whether "everyone else" can actually use it depends a lot on their situation. More on that shortly.

What Growth Hacking Actually Is

The term gets thrown around loosely enough that it's worth being precise about what the book actually means.

Growth hacking is not a collection of clever marketing tricks. It's a systematic approach to finding growth through data and experimentation rather than intuition and assumption. The core loop: form a hypothesis about what might drive growth, design the smallest possible experiment to test it, measure the result, keep what works, discard what doesn't, repeat.

What distinguishes this from normal marketing is the emphasis on speed and data at every stage. Traditional marketing campaigns take months to plan and execute and often produce results that are hard to attribute clearly. Growth hacking experiments are designed to be fast, cheap, and measurable — the whole point is to generate signal quickly so you're not committing large resources to approaches that haven't been validated.

The Airbnb case that gets cited constantly — where the founders discovered that poor-quality host photos were suppressing bookings and personally visited hosts to take better ones, resulting in a significant jump in reservations — is a good illustration of the method in miniature. They didn't assume the problem was marketing spend or platform features. They looked at the data, found the friction point, tested a specific fix, and measured the result. The solution was obvious once the problem was correctly identified. Finding the right problem is the hard part.

The Three Ideas From This Book That Actually Hold Up

After fifteen years of working with brands across North America and Greater China — from international consumer names entering the Chinese market to small founder-led businesses trying to grow their presence on Xiaohongshu or TikTok — three concepts from this book have proven genuinely durable.

1. The "Aha Moment"

Every product that retains users has a specific moment where the user gets it — where the value proposition stops being abstract and becomes real. Ellis calls this the "aha moment." For a ride-hailing app, it's the first time a car arrives faster than you expected. For a recipe platform, it's the first time you cook something from it and it actually works. For a B2B tool, it's the first time it saves you an hour you needed.

The aha moment matters because it's the gateway to retention. Users who reach it come back. Users who don't, leave. And most products have no idea where their aha moment is because they've never tried to find it systematically.

The practical exercise: look at your retained users — the ones who have been using your product or service for more than 90 days — and find what they did in their first week that your churned users didn't. That pattern is usually the aha moment, or close to it. Once you know it, you redesign your onboarding to get every new user there as fast as possible.

I ran a version of this exercise with that e-commerce founder. Her retained customers had almost universally browsed the "how to use" content before buying. New customers who went straight to product pages and bought without that context had much higher return rates and much lower repeat purchase rates. The aha moment was understanding the product before using it, not the purchase itself. We added guided content into the post-purchase flow. Repeat purchase rate went up measurably within six weeks.

2. The North Star Metric

One of the most useful and most misapplied ideas in the book. The North Star Metric is a single number that best represents the core value your product delivers to users — not revenue, not traffic, but the metric that most directly indicates whether users are getting what they came for.

Facebook's is daily active users. Airbnb's is nights booked. A messaging app's might be messages sent per user per week. The point is to choose one metric that the whole team optimises toward, because most growth teams fail not from lack of effort but from lack of focus — they're pulling in slightly different directions because they're watching different numbers.

For small businesses, this concept is liberating rather than constraining. Instead of tracking fifteen metrics and feeling overwhelmed, you pick the one that matters most and ask, every week: did that number go up or down, and why?

The caveat Ellis himself acknowledges: choosing the wrong North Star Metric can be damaging. Optimising for a proxy metric that doesn't actually reflect user value can produce numbers that look good while the underlying business weakens. The work of choosing the right metric is not trivial, and it requires genuine understanding of what your users actually value — which loops back to the aha moment work.

3. Structured Experimentation Over Gut Instinct

The part of growth hacking methodology that most small teams underuse is the experimental mindset — the habit of treating every marketing decision as a hypothesis to be tested rather than an assumption to be acted on.

This doesn't require an engineering team or expensive tools. At the most basic level it means: before you commit significant time or money to a channel or approach, run the smallest possible version of it and measure the result. Post three times a week on a new platform for a month before building a full content strategy for it. Test two different email subject lines before scaling an email campaign. Run a small paid budget on two different audiences before deciding which one to invest in.

The discipline of treating these as experiments — with a specific hypothesis, a specific metric, a specific timeline — changes how you interpret the results. You're not "trying things and seeing what happens." You're testing a specific prediction. When the result comes in, you learn something specific, whether it confirms or contradicts what you expected.

What the Book Doesn't Prepare You For

I want to be direct about this because the book's case studies — Airbnb, Dropbox, Facebook — can create an impression that growth hacking is primarily a technology company methodology.

A lot of the tactics in Growth Hacking require engineering resources that most small businesses simply don't have. Referral programmes with automated tracking. Personalised onboarding flows based on user behaviour. A/B testing at scale. These are powerful tools and they're worth knowing about. They're also largely inaccessible to a one-person operation or a small team without a developer.

What IS accessible to any team, regardless of size or budget:

The aha moment discovery process — this is a question of looking at your data, not building new infrastructure

The North Star Metric — choosing one number to care about most is a decision, not a technical project

The experimental mindset — running small, structured tests before scaling is a habit, not a tool

The customer lifecycle thinking (the AARRR framework: Acquisition, Activation, Retention, Revenue, Referral) — mapping where users drop off is something any business can do with basic analytics

The honest reality is that the social media platforms have done a lot of the infrastructure work for smaller operators. TikTok's analytics, Xiaohongshu's creator insights, META's ad testing tools — these give small teams more experimentation capability than Dropbox had when Sean Ellis was growing it. You don't need to build the testing infrastructure. It's already there.

Growth Hacking in the Chinese Market — One Difference Worth Noting

One thing the book doesn't address, which has been relevant in my work with brands entering or growing in Greater China: the platforms themselves are more active growth partners in the Chinese market than their Western equivalents.

On Xiaohongshu and Douyin, the platform's growth incentives and the brand's growth incentives are more explicitly aligned. Platforms actively surface and amplify content that performs well, provide direct data on what's working, and create commercial tools that small brands can access relatively easily. The growth hacking principle of "find what's working and do more of it" maps very directly onto how these platforms want their brand partners to operate.

The aha moment concept is particularly useful in this context. On Xiaohongshu especially, the content that converts — the post that takes a browser to a buyer — often hinges on a very specific moment of recognition: a user who is exactly like me, solving a problem I have, using this product. Finding that specific signal and replicating it deliberately is growth hacking applied to content strategy. The methodology is the same. The tools are different.

Who Should Read This

Growth Hacking is worth reading for anyone who manages growth for a product, service, or brand — which, if you're a founder or a solo marketer, means you. It's clearly written, well-structured, and the case studies are genuinely instructive even when they're not directly applicable.

Read it for the frameworks, not the tactics. The aha moment, the North Star Metric, and the experimental mindset are ideas that will change how you think about marketing problems, regardless of what tools you have available. The specific tactics — the referral mechanisms, the automated onboarding sequences — treat as things to work toward rather than things to implement immediately.

The book that pairs well with it: Hacking Growth (same author, different title in some markets) covers similar ground with more recent examples. If you want to go deeper on the metrics and analytical side, Lean Analytics by Croll and Yoskovitz is the best companion read.

The Pyramid Principle: The Thinking Framework I Wish I'd Learned Before My First Big Pitch

Wed, 17 Jun 2020 00:00:00 GMT

📖 Reading time: ~12 minutes

The Memo That Came Back With Red All Over It

Early in my career, I spent three days writing a campaign proposal for a senior client. I was proud of it. It was thorough. It started with the market context, moved through the research findings, walked through the analysis in careful detail, and arrived — on page eleven — at the recommendation.

It came back the same afternoon. One comment at the top, in red: "What are you actually proposing? I had to read to the end to find out."

That comment stayed with me for a long time. I understood what the client meant. I just didn't know how to fix it structurally — I kept writing the way I'd been taught to think, which was: gather evidence, build the argument, arrive at the conclusion. Start at the bottom and work up.

Barbara Minto's The Pyramid Principle is the book that explained, clearly and specifically, why that approach fails in professional communication — and what to do instead.

I read it about two years into my consulting work. It's one of those books that feels obvious in retrospect and transformative at the time.

What the Book Actually Is

The Pyramid Principle was developed by Barbara Minto during her time at McKinsey in the 1970s, and it has remained the foundational text on structured business communication ever since. Most major consulting firms still teach some version of it. Business school writing courses reference it. If you've ever read a McKinsey report or a well-constructed strategy deck and wondered why it felt so clear and easy to follow, this is usually the reason.

The core idea is simple enough to state in one sentence: lead with your conclusion, then support it.

But the book goes considerably deeper than that, into the specific mechanics of how to organize a complex argument so that every piece of it earns its place and the reader always knows where they are. That depth is what makes it useful for working professionals rather than just an interesting concept.

One important caveat Minto states upfront, which I appreciated: the book assumes you already know how to write clear sentences and paragraphs. It's not a writing guide. It's a thinking and structuring guide. If your sentences are unclear, no amount of pyramid structure will save them. What the framework does is organise clear thinking into a form that other people can actually receive and use.

The Idea That Changed How I Write Everything

The thing that landed hardest for me when I first read this book was the explanation of why "save the conclusion for the end" fails in professional contexts.

Minto's argument: when you present information without a clear conclusion upfront, you're forcing the reader to hold all of your evidence in working memory while trying to figure out where it's going. That's cognitively expensive. By the time they reach your recommendation, they've spent so much mental energy tracking the argument that they have less capacity to actually evaluate the conclusion.

When you lead with the conclusion — "we should do X, for three reasons" — you give the reader a frame before they encounter the evidence. They know what they're reading toward. Each piece of evidence slots into a structure they already have. The whole thing is easier to follow, easier to remember, and easier to push back on if they disagree.

In marketing and agency work specifically, this matters constantly. Every brief, every proposal, every client presentation, every internal recommendation — all of these live or die by whether the person reading them can quickly grasp what you're asking them to think or do. Burying the recommendation on page eleven doesn't demonstrate rigour. It demonstrates that you haven't done the work of deciding what actually matters most.

The professional habit shift the book requires is real and takes practice: finish your thinking before you start writing, not during. The structure is only as good as the clarity of thought behind it. When I catch myself writing in circles, it's almost always because I haven't actually decided what I think yet — and I'm using the document to figure it out rather than to communicate a conclusion I've already reached.

The SCQA Framework — The Most Useful Tool in the Book

The part of The Pyramid Principle that I've probably used most frequently in actual work is the SCQA structure for introductions. It stands for: Situation, Complication, Question, Answer.

The idea is that every effective professional communication starts by establishing a shared context (Situation), introducing the tension or problem that makes action necessary (Complication), raising the question that naturally follows from that tension (Question), and then answering it directly (Answer). The rest of the document supports the Answer.

In practice, a campaign brief introduction might run something like:

"Our Q4 campaign on TikTok is currently underperforming against benchmarks — reach is on target but conversion to purchase intent is 40% below last quarter. [Situation] The primary driver appears to be a mismatch between content format and the platform's current algorithmic preferences, which have shifted significantly toward video-first formats since our last campaign cycle. [Complication] How do we close the conversion gap before the end of Q4? [Question] We recommend reformatting 60% of planned content into short video within the next two weeks, prioritising the three product lines with the strongest existing community signal. [Answer]"

This sounds obvious when you read it. The reason it's worth formalising is that most professional writing skips one or more of these elements — usually the Complication (which explains why action is needed now, not just what the situation is) or the explicit Question (which makes clear what kind of answer the document is actually providing).

Minto also offers four variations on this structure depending on context — Standard (Situation → Complication → Answer), Direct (Answer → Situation → Complication), and two others suited to specific situations like emphasizing urgency or pre-empting resistance. The flexibility is genuinely useful once the base structure is internalized.

Deductive vs Inductive: The Distinction That Cleans Up Messy Arguments

The other part of the book that I found genuinely clarifying — and that I still use as a diagnostic when a document isn't working — is the distinction between deductive and inductive reasoning structures.

Deductive reasoning builds an argument step by step, where each point follows from the previous one, leading to a conclusion. Major premise → minor premise → conclusion. The classic syllogism. In a document, this looks like: "X is true. Y is true given X. Therefore Z." The problem with pure deductive structure in long documents is that it's slow and the reader has to wait through the whole chain before they get the conclusion — which is exactly the problem SCQA and conclusion-first structure solve.

Inductive reasoning groups several independent points that all support the same conclusion. "We should do X for three reasons: A, B, and C." The points don't depend on each other — they each stand alone as supporting evidence. This is the structure that works better for most business documents because the reader can see the conclusion immediately and each supporting point adds independent weight.

The practical diagnostic: if you're writing a document and it feels like it's going in circles, check whether your supporting points actually stand independently or whether they depend on each other to make sense. If they depend on each other, you have a deductive chain and you need to be more explicit about the logical sequence. If they're independent, you have an inductive structure and you should lead with the conclusion they support.

This sounds technical. In practice, once you've applied it a few times, it becomes second nature — the kind of thing you notice immediately when a document isn't working and can usually fix in under ten minutes.

How I Actually Use It Now

After fifteen years of writing proposals, briefs, strategy documents and client presentations across North American and Greater China markets, here's what has stuck from this book:

Always write the conclusion first — in your head if not on the page. If you don't know what you're recommending before you start writing, stop and figure that out first. The document exists to communicate a decision, not to reach one.

Use SCQA for anything that requires context-setting. Briefing a new client. Explaining a platform strategy change. Presenting unexpected results. Recommending a budget reallocation. The Situation-Complication-Question-Answer structure makes the stakes immediately clear and the recommendation immediately legible.

Check your supporting structure when something isn't landing. When a client pushes back in a way that feels like they've missed the point, the problem is often structural rather than substantive. They received the information but not in an order that allowed them to process it correctly. This is fixable.

The framework matters more in high-stakes communication. For a Slack message or a quick email, overthinking structure creates friction. For a board presentation, a major proposal, or a brief that will be briefed to a team — the investment in getting the structure right pays back many times over.

The one limitation I'd note: this book was written in a consulting context, and some of its examples and assumptions reflect that world. For social media marketers, content creators, or anyone whose communication is more conversational than formal, the framework requires some translation. The underlying logic — lead with what you want the person to think or do, then give them the reason to believe it — is universal. The specific mechanics need calibrating to context.

Who Should Read This

The Pyramid Principle is not a quick read and it's not a light one. It rewards careful, slow reading and benefits enormously from being applied in real work as you go through it rather than read end-to-end and shelved.

The audience it's most useful for: anyone who regularly writes documents that need to persuade or inform — proposals, strategy briefs, campaign post-mortems, internal recommendations, client presentations. If you've ever sent something carefully constructed and received a response that suggested the other person had read something entirely different, this book will tell you why and give you specific tools to fix it.

If you're early in your marketing career and building a writing practice: read this book before you develop too many bad structural habits. The habits the framework builds are much easier to form from scratch than to replace later.

If you're running a small business and writing your own proposals and pitches: the SCQA structure alone is worth the price of the book. Apply it to your next proposal. See what happens.

Embedded Power: What This Book Finally Explained About China's Economy — And My Own Missed Opportunities

Thu, 01 Dec 2022 00:00:00 GMT

📖 Reading time: ~10 minutes

The Conversation I Wish I'd Had Earlier

Two moments from my own life keep coming back to me while reading this book.

The first: 2008 or 2009, just starting university. Property prices in Wuhan were rising in a way that felt unusual even to an 18-year-old with no savings and no real understanding of how any of it worked. I noticed the numbers going up. I didn't understand why. Tianjin followed a few years later. These weren't random fluctuations — in hindsight, they were part of a very legible cycle, one that anyone who understood how local government financing actually worked in China could have seen coming. I didn't understand it then. Nobody in my immediate circle explained it to me. I did nothing.

The second: about 2015, a few years into working life, living in Beijing. By this point I had some savings, some sense of how the world worked, and a growing awareness that the property developers were doing something aggressive — the land auctions were becoming spectacles, prices were being pushed hard, leverage was being stacked in ways that felt unsustainable even from the outside. I had the instinct. I did not have the framework to act on it. And crucially, I didn't have Beijing household registration — hukou — which meant that even if I'd had the down payment, I didn't have the legal standing to buy.

I already knew by then that I'd likely spend time living abroad. So I told myself it was fine. I moved on.

Society, as we say in Mandarin, beat me soundly regardless.

Lan Xiaohuan's Embedded Power — originally published in Chinese as 置身事内, a title that roughly translates as "being inside the matter" or "embedded in the game" — is the book that would have given me the framework I was missing in both of those moments. It's a book about how the Chinese government actually works in the economy: not as regulator standing outside it, but as a participant embedded within it. It's written by a Fudan University economics professor in language that is remarkably clear given its subject matter, and it's become one of the most widely discussed economics books in China in recent years.

As of writing, there is no official English translation. I hope there will be.

What the Book Is Actually About

The central argument of Embedded Power is stated clearly and early: in China, the government does not merely influence how the economic "cake" gets distributed. It participates directly in baking it. This is not a criticism or a defence — it's a description of a structural reality that shapes almost every significant economic phenomenon in the country.

Lan organises the book in two parts.

The first half works at the micro level — the mechanics. What does a local government in China actually do day to day? Where does its money come from? How does it spend? The land financing system gets significant attention here, and rightly so: the way local governments in China have historically used land sales as a primary revenue mechanism is one of the most important and least-understood features of the country's economic model from the outside. It's not corruption (mostly). It's a structural feature of how the fiscal relationship between central and local government was designed, and it has downstream consequences for almost everything — infrastructure investment, debt accumulation, the behaviour of property developers, the trajectory of housing prices.

The second half zooms out to the macro level — urbanisation and industrialisation, regional inequality, debt risk, the structural imbalances in domestic demand, and the origins of trade tensions with the West. These are phenomena that show up in headlines constantly. What Lan does is connect them back to the micro-level mechanisms described in the first half. Once you understand how local governments are incentivised and constrained, the macro picture becomes considerably more coherent.

The book doesn't require an economics background to follow — this is one of its genuine achievements. Lan writes with an academic's rigour but a teacher's instinct for clarity. Complex fiscal arrangements get explained with examples grounded enough that a non-specialist can follow the logic.

Why This Matters for Anyone Working in or With China

I've spent over ten years building marketing strategies for brands operating across both North American and Chinese markets — working with everything from major international consumer brands entering China to Chinese founders trying to grow a presence elsewhere. And one of the most persistent gaps I've seen on both sides is a structural misunderstanding of how China's economic incentives actually work.

Western brands entering China frequently approach the market with frameworks built for contexts where government is primarily a regulator — something you need to comply with, navigate around, occasionally lobby. The reality in China is more complex and in some ways more interesting: local governments have their own economic interests, their own revenue needs, their own development timelines. Understanding those interests doesn't just help you navigate compliance — it can tell you which cities are actively trying to attract foreign investment, where infrastructure investment is likely to flow, which sectors have political tailwinds, and which relationships are worth building.

For small business owners and marketing professionals trying to understand why certain Chinese platforms, content categories, or consumer trends emerge when and how they do — the underlying logic often runs back to exactly the kind of government-economy relationship Lan describes. RedNote's growth, the rise and regulation of the live-commerce sector, the government's ambivalent relationship with the domestic tech giants — none of these is fully legible without understanding the structural context.

This is not a book that will tell you how to run a campaign or close a deal. It will tell you something more foundational: how the system your campaign or deal is operating inside actually works. That knowledge compounds.

The Real Estate Chapter and the Personal Reckoning

The book spends considerable time on housing and land — perhaps more than on any other single topic. For readers who lived through China's property boom of the 2000s and 2010s, these chapters land differently than they would for someone reading about the period purely historically.

Lan's explanation of why housing prices rose when and where they did is not about market dynamics in the conventional sense. It's about the fiscal pressures facing local governments, the structure of land auction mechanisms, the way infrastructure investment creates value that gets captured through subsequent land sales, and the feedback loop this created over two decades. When you understand the mechanism, the regional patterns of price increases — why Wuhan moved when it did, why Tianjin followed — stop looking like chaos and start looking like a logical output of a particular set of incentives.

I'll be honest: reading this section while remembering both of my missed moments was not entirely comfortable. The opportunities weren't invisible — they were legible to anyone with the right framework. I didn't have the framework.

What I take from this now, working in a market advisory capacity with clients navigating the China business environment, is that the structural knowledge in this book is genuinely perishable. The specific mechanisms Lan describes — land financing, local government debt, the particular shape of the urbanisation model — are themselves changing. The government has been actively trying to shift the economy away from infrastructure-and-property-investment-led growth toward domestic consumption and technology. That transition is ongoing and uneven and genuinely uncertain. Reading Embedded Power tells you how the system worked and, importantly, why — which gives you a basis for thinking about what changes as the model evolves and what doesn't.

Who Should Read This

Embedded Power is not a light read, but it's not an inaccessible one either. The audience it seems written for is exactly the audience it deserves: thoughtful people who want to understand China's economy at a structural level rather than through the simplified narratives that dominate most English-language coverage.

For international readers: if you work with China in any professional capacity — sourcing, investment, brand expansion, platform strategy — this book will give you a foundation that most of your competitors don't have. That's worth something concrete.

For marketers and business owners working in the Greater China market: the political economy backdrop Lan describes is the operating environment you're inside. The more fluent you are in its logic, the better your decisions will tend to be.

For anyone who, like me, has a few "I should have understood this earlier" moments from their own relationship with China's economic development — this book is a rewarding and occasionally humbling experience.

I'll be watching for the English translation. In the meantime, the Chinese original is worth the effort if you can manage it.

DeepSeek R1 vs ChatGPT: What the Hype Gets Wrong — A Marketer's Real Assessment

Wed, 05 Feb 2025 00:00:00 GMT

📖 Reading time: ~11 minutes

💡 Quick Answer DeepSeek R1 is genuinely impressive — fast, cheap, open-source, and capable of strong reasoning. But its hallucination rate is significantly higher than competing models, which makes it a risky daily driver for marketing, content, or any work where accuracy matters. For research and brainstorming it has real value. For anything client-facing or decision-critical, proceed carefully.

The Week Everyone Asked Me If They Should Switch

The third week of January 2025, my phone wouldn't stop.

Clients, collaborators, a few founders I'd been advising — all some version of the same message: "Have you seen DeepSeek? Should we be switching? Is ChatGPT dead?"

It had been maybe five days since DeepSeek R1 hit the App Store. Within a week it had knocked ChatGPT off the top of the download charts — which, if you follow the AI space at all, felt roughly equivalent to a regional brand showing up at the Super Bowl and outselling Nike. US technology stocks dropped sharply on the news, with estimates of around $1 trillion wiped from valuations in a matter of days. The reaction was somewhere between genuine excitement and low-grade panic, depending on who you asked.

I've been using AI tools daily for client work since the early wave — content workflows, campaign analysis, research, platform strategy across TikTok, META, Google and RedNote. So the question of whether DeepSeek R1 was worth switching to wasn't abstract for me. I needed an actual answer, not a hot take.

Here's what I found after spending real time with it.

What DeepSeek R1 Actually Is — And Why It Caused a Stir

DeepSeek R1 was officially released on January 10, 2025, with its model weights open-sourced under the MIT license ten days later. It was built by a Chinese AI lab and developed at a reported cost of approximately $5.6 million — a figure that got passed around in every headline because, for context, GPT-4's estimated development cost was around $100 million.

That cost gap is the story everyone latched onto, and it's a real story. The efficiency of what the DeepSeek team achieved with comparatively limited compute is genuinely impressive from an engineering standpoint. In benchmark tests across coding, mathematics and scientific reasoning, R1 performed competitively against models that cost far more to build and run.

The open-source angle matters too. Released under MIT licence, R1 can be freely used, modified and deployed — which is a meaningful difference from the closed systems that dominate the market. For developers, for researchers, and for companies that want to run models on their own infrastructure, that flexibility has real value.

So the excitement wasn't manufactured. Something significant happened.

But benchmarks and headlines are not the same as usefulness in practice, and this is where the story gets more complicated.

The Number Nobody Was Talking About

In the middle of all the coverage — the App Store charts, the stock market reaction, the think-pieces about whether this was America's "Sputnik moment" in AI — there was a data point that kept getting buried.

DeepSeek R1's hallucination rate is significantly higher than the models it was being compared to.

Hallucination, in AI terms, means the model generates information that sounds plausible but is factually incorrect or entirely fabricated. It's the AI equivalent of a very confident colleague who occasionally just invents things without realising it.

The research on R1's hallucination rate surfaced a few specific patterns worth understanding:

The "over-helpful" problem. R1 tends to add information that wasn't in the source material, even when that addition might seem reasonable. About 71.7% of R1's hallucinations fall into this category — the model essentially filling gaps with plausible-sounding content it generated rather than knew. For comparison, DeepSeek's own V3 model shows this pattern in only 36.8% of its hallucinations. That gap is significant.

The reasoning chain creates more opportunities to go wrong. R1 shows its thinking process — which is genuinely useful for transparency and for understanding how it reached a conclusion. But longer chains of reasoning also mean more steps where a small error can compound. R1's average token output per query is around 4,717 — compared to 191 to 462 for most competing models. It's doing a lot more work, which creates a lot more surface area for mistakes.

The real cost calculation. R1's per-token pricing is dramatically lower than GPT-o1: approximately $0.55 per million input tokens versus $15. That sounds like a no-brainer until you factor in that R1 is generating 10 to 25 times as many tokens per query. The actual cost differential is much smaller than the headline number, and in some use cases disappears entirely.

What This Means If You're Running a Small Team or a One-Person Operation

Let me be direct about why this matters practically, because I think the hallucination conversation often gets too abstract.

If you're using an AI tool to help you draft social media captions or brainstorm campaign angles, a higher hallucination rate is a manageable inconvenience. You're going to read and edit the output anyway. A plausible-sounding fake statistic in a first draft is annoying, not catastrophic — you catch it before it goes anywhere.

But if you're using AI to research a market, fact-check competitor claims, summarise a report for a client presentation, or support any kind of decision that has real consequences — the stakes are different. A 14% error rate on factual content is not a minor quirk. Imagine building a client proposal around a figure that turns out to be fabricated by your AI tool. Or sending a report to a brand director with incorrect data you didn't catch because you trusted the output. That's not a hypothetical risk; it's a pattern that will eventually catch up with anyone who doesn't account for it.

I've worked with clients ranging from early-stage startups to regional brand teams where one person is managing the entire marketing function. In both contexts, trust is the most fragile and most valuable asset you have. One credibility slip — one piece of content or one client-facing document with an AI-generated error you didn't catch — can cost you more than weeks of recovered productivity from using a cheaper tool.

This isn't an argument against using DeepSeek R1. It's an argument for using it with a clear-eyed sense of where it's appropriate and where it isn't.

When DeepSeek R1 Is Actually Worth Using

Having spent time with it, here's where I think R1 genuinely earns its place in a working toolkit:

Exploratory research and brainstorming. When you're in early-stage thinking — generating ideas, exploring angles, mapping out a topic — the hallucination rate matters less because you're not treating the output as ground truth anyway. R1's reasoning chain is actually valuable here. Watching it think through a problem produces ideas you might not get from a model that just delivers a clean answer.

Coding and technical problem-solving. Its benchmark performance in coding is legitimate, and for developers or technically-oriented marketers building tools or automations, it's worth testing directly against your specific use cases.

Price-sensitive high-volume workflows. If you're running large volumes of lower-stakes text generation — internal documents, first-draft outlines, content that will be heavily edited — the cost advantage can add up, as long as you're building in verification steps.

When open-source deployment matters. If you need to run a model on your own infrastructure for data privacy reasons — which is increasingly a concern for brands operating in both North American and Chinese markets — the MIT licence is a meaningful advantage.

Where I'd be more cautious: anything client-facing, anything involving factual claims, anything that goes out under your name or your brand's name without a thorough human review pass. The efficiency gains don't justify the risk if you're not building in that review step.

The Bigger Question the DeepSeek Moment Is Asking

The reaction to DeepSeek R1 revealed something about how we talk about AI that's worth pausing on.

The narrative moved incredibly fast — from release to "ChatGPT killer" to stock market panic to backlash to nuanced reassessment — and at almost every stage, the conversation was more about the meaning of DeepSeek than about what the model actually does well and where it falls short. The geopolitical angle, the cost angle, the open-source angle — all of these are real and interesting. But they're not the same as "is this useful for the work I'm actually doing?"

This is a pattern worth noticing because it's going to repeat. AI development is moving fast enough that there will be another DeepSeek moment — another release that causes a wave of "should I switch?" messages. The useful question to ask each time isn't "is this impressive?" but "what specifically is this better at, and what does it cost in accuracy or reliability to get that advantage?"

For marketers and small business owners especially, the temptation to chase the newest and cheapest option is real when budgets are tight. But AI tools are infrastructure now — they're embedded in research, content, client communication, analysis. Optimising for cost without accounting for error rate is like buying the cheapest possible accounting software because the subscription is lower, without checking whether it gets the numbers right.

The right framework isn't "which AI tool is best?" It's "which AI tool is best for this specific task, at this specific risk tolerance, with these specific verification steps in place?" DeepSeek R1 answers that question well for some tasks. For others, the answer is still ChatGPT, Claude, or something else entirely. And that's fine — a toolkit is supposed to have more than one tool in it.

❓ FAQ

Q: Should marketers use DeepSeek R1 in 2025? A: For brainstorming, research exploration, and lower-stakes content drafting — yes, it's worth testing. For anything client-facing, factual, or decision-critical, build in strong verification steps or stick with models that have lower hallucination rates. DeepSeek R1 is a useful tool in the right context, not a universal upgrade.

Q: Is DeepSeek R1 actually cheaper than ChatGPT? A: The per-token price is dramatically lower, but R1 generates 10 to 25 times more tokens per query than most competing models. The real cost difference is much smaller than headlines suggest, and for some use cases the cost is similar or higher. Factor in actual token usage, not just the headline price, before making a decision.

Q: What is AI hallucination and why does it matter for marketing? A: AI hallucination is when a model generates confident-sounding content that is factually incorrect or fabricated. For marketing, this matters because incorrect data in a client report, a campaign brief, or a published piece of content can damage your credibility seriously. DeepSeek R1's hallucination rate is notably higher than competing models — a real consideration for professional use.

Q: How does DeepSeek R1 compare to ChatGPT for social media content? A: For generating creative options, exploring angles, and drafting content that you'll edit heavily, R1 performs well and the lower cost can be attractive at volume. For anything requiring accurate facts, statistics, or specific claims, the higher hallucination rate means more careful review is needed versus GPT-4 or Claude.

Q: Is DeepSeek R1 safe to use for client work? A: With appropriate verification, yes — but that verification step is non-negotiable. The same way you'd fact-check research from any AI tool before it goes to a client, R1 requires the same diligence, arguably more. The risk isn't that it's unusable; it's that its confident output style can make errors easy to miss if you're moving fast.

Finite and Infinite Games — The Philosophy Book That Changed How I Think About Marketing

Thu, 14 Nov 2019 00:00:00 GMT

📖 Reading time: ~10 minutes

The Client Meeting That Made Me Think Differently

A few years into running the agency, I sat across from a brand director who was frustrated. Their Douyin numbers were strong — follower growth on target, engagement rates healthy, content performing well against benchmarks. By every metric we'd agreed on at the start of the year, the account was succeeding.

But she felt like something was missing. The brand had an audience. It didn't have a community. People were watching but not talking. They were clicking but not caring. The numbers were good and the relationship with the platform felt hollow.

I didn't have clean language for what she was describing at the time. I do now, mostly because of a short, strange, deeply useful book by an American philosopher named James Carse.

Finite and Infinite Games was published in 1986 and runs to fewer than 200 pages. It has been recommended to me by multiple people over the years in completely different contexts — a founder, a strategist, someone who works in education. The book has that quality of being useful to almost everyone who reads it, for reasons specific to their own situation. What I took from it was a framework for understanding a tension I'd been observing in marketing for years without having the right words for it.

The Core Idea — And Why the Chinese Translation Gets It Wrong

The book's central distinction is simple: there are finite games, played to win and end, and infinite games, played to continue.

A quick note on terminology: the Chinese translation renders "game" as 游戏, which implies something light and recreational. A more accurate translation would be 博弈 — the term used in game theory, which carries the sense of strategic interaction with genuine stakes. This matters because Carse is not writing about entertainment. He is writing about every form of human activity in which people engage with purpose and consequence.

Finite games have defined rules, defined players, a beginning and an end, and a winner. The purpose of a finite game is to win it. Chess. A job interview. A quarterly sales target. A political election. Finite games are characterised by their boundaries — once you know the rules and the endpoint, you know how to optimise your play.

Infinite games have no fixed endpoint and no final winner. The purpose is not to win but to keep the game going — to ensure that play continues and that more players can join. Marriage. Culture. Language. A creative field. A meaningful career. Infinite games change their rules over time when necessary, specifically to avoid the game ending.

Carse's insight, which sounds obvious until you think about how rarely it gets applied: most of the things that matter most to us are infinite games, but we keep trying to play them with finite game logic.

What This Has to Do With Marketing

When I think about the tension that brand director was describing — good numbers, hollow relationship — the finite/infinite framework maps directly onto it.

Most marketing is organised around finite games. Campaigns have start dates and end dates. KPIs have targets. Quarterly reviews assess whether the number went up or down. Budgets get allocated based on what "won" last period. This is all rational and largely necessary — businesses need measurable outcomes, and campaigns need parameters.

The problem is when finite game logic gets applied to things that are fundamentally infinite games. Brand trust is an infinite game. Community is an infinite game. The relationship between a brand and its audience — the kind of relationship where people actually care what happens to a company, where they defend it when it makes mistakes, where they tell their friends without being incentivised to — that is an infinite game. And it cannot be optimised the way a campaign can be optimised.

I've watched this play out in both directions across the markets I work in.

In North America, I've seen brands with genuinely impressive engagement metrics that have no idea who their audience actually is. They've run enough campaigns to generate numbers. They haven't had enough real conversations to generate understanding. The metrics win the finite game. The relationship, which is the infinite game, stays underdeveloped.

In the Greater China market — on Xiaohongshu and Douyin especially — the brands that build lasting presence aren't necessarily the ones with the best content schedules. They're the ones that engage like a person rather than a brand manager. They respond to comments with genuine opinions. They notice what their community is actually talking about and join the conversation rather than redirect it. They play the infinite game, which means they're willing to invest in the relationship before the return is visible in any metric.

The follower count is a finite game score. The follower who mentions your brand unprompted to a friend because they actually like it — that's an infinite game outcome. Both matter. They require different kinds of attention.

Finite and Infinite in Business and Agency Work

The framework extended further for me when I started applying it to the business of running a consultancy.

Project-based work is, structurally, a finite game. There's a scope, a timeline, a deliverable, a sign-off. When the project ends, the game ends. This is how most agency relationships are structured, and it creates a particular dynamic: both sides optimise for the deliverable rather than the relationship, because the deliverable is what the finite game is scored on.

The client relationships that have meant most to me professionally — the ones that have lasted years and grown in scope and generated real mutual investment — are the ones where both sides consciously shifted to infinite game logic at some point. Not "what do we need to deliver to close this project?" but "what does this brand actually need to grow, and how do we keep solving that problem together over time?" The conversation changes. The stakes change. The quality of the work changes too, usually.

This shift is harder to make than it sounds, particularly when business development incentives are structured around new clients rather than deepening existing relationships. Finite game metrics (new accounts, project revenue, team utilisation) can actively work against infinite game relationship building if you're not paying attention to both simultaneously.

The Part About "Playing Toward Death"

There's a section of Finite and Infinite Games that I've thought about more than any other — Carse's concept of "living toward death."

His point is not morbid. It's clarifying: finite games derive meaning from their endings. A match matters because it ends and someone wins. But if life itself is an infinite game, then death is not the thing that gives life meaning — it's the thing that marks its limit. The question, for infinite games, is not "how do I win before this ends?" but "how do I play in a way that keeps something valuable continuing beyond me?"

For entrepreneurs and small business owners, I find this reframe genuinely useful. There's a version of building a business that is purely a finite game — build to exit, optimise for a transaction, define success as the moment you cash out. That's a legitimate approach and some people play it well. But many people I've worked with, including myself, are actually playing a different game — building something they care about continuing, growing a team they want to keep working with, developing expertise they want to keep deepening. When you're playing an infinite game and using finite game metrics to evaluate it, you'll feel constantly behind even when things are going well.

Naming the game you're actually in is more than half the work of playing it well.

What I'd Tell Someone Reading This Today

Finite and Infinite Games is not a business book, and it shouldn't be read as one. It's a philosophy book — genuinely strange in places, deliberately aphoristic, written by someone who was more interested in asking good questions than providing actionable frameworks. Simon Sinek wrote a more accessible business-oriented version of a similar idea in The Infinite Game, if that's what you're after.

But Carse's original is worth the effort precisely because it isn't practical in the conventional sense. It changes the way you see the categories, not just the tactics. And that shift — from optimising a finite game to understanding which game you're actually playing — tends to be worth more, over time, than any specific technique.

The question I came away with, and still return to: which of the things I'm working on are finite games I'm treating as infinite, and which are infinite games I'm mismanaging by applying finite game logic? The answer is almost always clarifying, and usually uncomfortable in the most useful way.

Being Mortal — What I Understood After My Mother's Diagnosis

Tue, 25 Oct 2022 00:00:00 GMT

📖 Reading time: ~11 minutes

The Book I Wasn't Looking For

I didn't pick up Being Mortal because I was interested in medical ethics or end-of-life policy. I picked it up because my mother had just been diagnosed with cancer, and I was looking for something — anything — that would help me understand what was happening and what came next.

Atul Gawande's book had been on my list for a while, one of those titles that feels important in the abstract. Reading it in the weeks after my mother's diagnosis made it something else entirely. It became a mirror for an experience I was already inside, one I hadn't asked to be part of and had no idea how to navigate.

I want to write about this honestly, which means writing about both the book and what the book didn't prepare me for. Gawande is a gifted writer and a careful thinker. Being Mortal is genuinely important. It also, like most books about serious illness, is written primarily from the outside — from the perspective of the doctor, the observer, the thoughtful witness. The person on the inside of it, in my case, was the one sleeping in hospital chairs.

What the Book Is About — And Why It Matters

Being Mortal is structured around a central argument: modern medicine has become so focused on extending life that it has lost the ability to ask whether the life being extended is one the patient actually wants.

Gawande makes this case through a series of stories — his own father's illness, patients he has treated, friends navigating the healthcare system with aging parents. He is especially good on the gap between what doctors are trained to do (identify the problem, treat the problem) and what patients facing serious illness often actually need (to be asked what matters to them, and then to have that answer actually heard).

The parts of this book that are hardest to read are the ones where a patient has a clear sense of what they want — to be home, to avoid aggressive treatment, to be able to eat dinner with their family without being in a hospital — and the medical system simply doesn't have the infrastructure to honour that. Not out of cruelty. Out of the particular kind of institutional blindness that comes from optimising for the wrong thing.

The question Gawande keeps returning to is a simple one: what does this person actually want their life to look like? Not "what is medically possible?" but "what makes this life feel worth living to the person inside it?" These questions are not the same. In most healthcare settings, the first gets asked constantly. The second rarely gets asked at all.

What My Mother's Year Actually Looked Like

My mother's diagnosis came without warning — the way these things tend to come. One day things were normal. Then they weren't.

The surgery was extensive. She was in the ICU for several weeks afterward. I learned what it means to live in a hospital corridor — the specific geography of waiting rooms, the rhythms of shift changes, the language you develop for communicating with nurses when you've been there long enough that they know your name. I learned what it sounds like when someone you love is struggling to breathe after a ventilator comes out, and what it sounds like — the specific quality of the exhale — when they manage it on their own.

After that came a year of chemotherapy.

I want to describe what that year was like without making it sound more coherent than it was. It wasn't a journey, exactly. It was more like a constant low-grade emergency with brief remissions. I would wake up early to read research papers on her specific cancer subtype before her oncology appointments, trying to understand enough to ask useful questions. I would sit in treatment rooms for hours, go home and work until midnight, sleep a few hours, and start again. I missed things — social things, professional things, things that mattered to me — without keeping track of what I was missing, because keeping track would have required attention I didn't have.

Gawande writes about how patients with serious illness need advocates — people who understand both the medical situation and the person inside it, who can translate between the clinical and the human. He's right. What he doesn't write about at the same length is what that advocacy costs the person doing it.

There's a term in the research literature for what happens to family caregivers over time: caregiver burden. Physical depletion, emotional exhaustion, social withdrawal, professional disruption. The medical system sees the patient. The caregiver is largely invisible to it. You are there every day, often making decisions under pressure with incomplete information, and the system treats you as a resource rather than as a person who is also going through something.

I don't say this as a complaint about the doctors or nurses — the people we worked with were skilled and caring. I say it because Gawande's book is very good at making the patient visible in a system that tends to reduce them to a diagnosis. The caregiver needs that same visibility, and it's mostly absent.

The Decisions Nobody Tells You About

The part of Being Mortal that I read most carefully — and returned to most often — is the section on how to have honest conversations about what a patient actually wants when a prognosis is serious.

Gawande argues that these conversations rarely happen well in practice, despite everyone agreeing they should. Doctors default to presenting options rather than asking values questions. Patients default to saying they want whatever the doctor recommends, because that feels like the responsible answer. And somewhere in the middle, the actual human being — with their actual preferences about what their life should feel like — gets bypassed.

We had versions of this problem throughout my mother's treatment. After each scan, there were decision points about whether to continue, modify, or stop. The medical information was presented clearly. The deeper question — what does she want her life to look like while she's going through this? — often had to be asked by us, not by the system.

I became obsessed, during this year, with preserving the small things that made her life hers. When she was too weak to maintain her usual independence, we reorganised the house so she could move through it on her own terms. When the treatment side effects made her familiar foods intolerable, we tested alternatives until we found things that still gave her pleasure. When pain limited what she could do physically, we found other ways to mark the day, other small rituals that gave the hours some shape.

These things mattered enormously to her. They were also invisible to everyone except us.

This is what Gawande means when he writes about what patients actually want near the end of life — not grand things, but ordinary ones. Continuity. Comfort. The ability to be a person rather than a case. The medical system is not well designed to deliver this, and recognising that gap clearly is one of the most useful things his book does.

A Year Later

My mother completed chemotherapy. She came through it.

I find myself returning to Being Mortal now with different eyes than the ones I read it with the first time. The first time, I was reading it in real time, using it as a field guide for an emergency I was inside. Now I can read it more slowly, and what I notice differently is how much the book is about the gap between the stories we tell ourselves about medicine and what medicine actually is.

We tell ourselves that modern medicine can fix things. Sometimes it can. We tell ourselves that more treatment is almost always better than less treatment. Gawande's evidence suggests this is often not true. We tell ourselves that a longer life is by definition a better outcome. His patients consistently suggest otherwise.

What I took from the year of my mother's treatment, and what the book helped me articulate after the fact, is something like this: the most important decisions in a serious illness are not the medical ones. They are the human ones. What does this person want their days to look like? What are they willing to endure, and what aren't they? What would make the time they have feel meaningful rather than just extended?

These questions don't have medical answers. They have personal ones. And the people who love the patient are often the only ones asking them.

Who Should Read This

Being Mortal is not easy to read, but it is important and it is clear. Gawande writes without sentimentality and without evasion, which is harder than it sounds when the subject is death.

If you are currently a caregiver for someone with a serious illness: this book will give you language for things you are already experiencing, and it will help you ask the right questions in medical settings where those questions otherwise don't get asked. Read the sections on goals-of-care conversations carefully.

If you are not currently in that situation: read it anyway. Most of us will be caregivers at some point, and most of us will also eventually be patients. The frameworks Gawande offers are genuinely useful in both roles.

What I would add, as someone who has been inside the experience the book describes: the book is more useful if you read it before you need it. The year I was living in hospital corridors was not the ideal time to be developing new frameworks for thinking about medical decision-making. The book gave me what it could. Some of what I actually needed, nobody had written yet.