A brand sends overseas creators what could be called a textbook-perfect brief: selling points complete, script fixed, which lines must appear and which words cannot be changed — all written out clearly.
Three months later, the video goes live. Views are okay. Conversions are zero. The team reviews and concludes: the creator didn't film it well. Let's try a different batch.
Susan Peng has heard too many stories like this. Her judgment is the opposite.
"The video comes out as a stiff ad — the problem isn't the creator, it was set in motion the moment the project started." The more "correct" the brief, the less the creator can put it in their own words. The audience sees it's an ad at a glance, swipes past, done.
Peng is now Partner and CGO at AhaCreator, doing global social media content marketing. Looking back at her career: five years at Pinduoduo, going through commercialization, Duoduo Grocery, then participating in TEMU's semi-managed business from 0 to scale. Before that: Wall Street CN, Sina Weibo, Douban. Brands, merchants, supply chains, global creators — she's operated them all firsthand.
I asked her: when Chinese brands go overseas and do creator marketing, where is the most common misunderstanding?
Her answer: "The phrase 'find a batch of creators to post videos' isn't wrong, but it's too shallow."
Peng breaks this down into three "not equals."
Finding people ≠ partnership established. Whether the person is a match, whether they have genuine willingness, whether the quote and schedule are feasible — every item needs verification.
Video posted ≠ marketing complete. Whether the content explains the product clearly, whether usage rights are明确, whether the brand's own conversion funnel can catch it — every item changes the outcome.
One campaign done ≠ capability formed. If the first round only leaves "how many videos posted," the next round still has to find people from scratch, request quotes again, and trial-and-error all over.
What's truly valuable are the answers each round leaves behind: which people work, which expressions work, where the problems get stuck, which content can be reused, how the next budget should be adjusted.
"What's truly valuable are the answers each round leaves behind: which people work, which expressions work, where the problems get stuck, which content can be reused, how the next budget should be adjusted."
These answers aren't pretty words in a report. They directly determine who the next batch of budget goes to, what to replicate, what to abandon.
Many teams only count "how many videos posted, how much money spent" in the first round, then start from zero again in the second round — finding people, requesting quotes, trial-and-error — which means paying new-company costs every time, repeatedly paying tuition.
So I asked her a more specific question: a Chinese brand entering the US market for the first time, with a $100,000 budget — how exactly would you spend it?
Her answer was specific enough to surprise me.
First, divide the time: three months. This window needs to cover positioning and brief, phased delivery, review, plus one retargeting decision.
The money needs clear boundaries upfront: $100,000 by default covers creator collaboration and content-related budgets. Whether sample shipping, content usage rights, taxes, and subsequent amplification are included must be written clearly before launch.
Then go through three phases.
First, clarify the positioning. What is the product, who is the core audience, focus on one scenario, validate one content hypothesis this round — and only then form an executable brief.
Second, deliver in batches, never spend it all at once. The first batch only validates a few clear audience segments and content angles. Get real feedback, then adjust the second batch.
Third, review and retarget. Break out effective creators, effective content structures, and funnel problems. Decide whether to add budget, purchase usage rights, or simply stop a certain approach.
How many people to find? She says the headcount is never decided first. Calculate the full cost of each qualified collaboration based on real quotes first. Reserve money for retargeting and supplementary testing. Ensure each major audience segment or content hypothesis has more than one independent sample. Only the remaining budget determines how many creators to find in the first batch.
$100,000 spent on SaaS, consumer goods, or hardware — the headcount is completely different.
When to add budget? She looks at three things:
1. Whether the same audience segment or content angle shows consistent signals across more than one creator
2. Whether the brand's own click, trial, and order data shows matching signals
3. Whether quotes, fulfillment, and usage rights support continued scaling
By the way, for first-time entry into a new market, her default order is mid-tier and long-tail first, then top-tier. Top creators are a bit like minor celebrities — they can create events and open awareness, but mid-tier and long-tail creators provide comparable market samples. Only by testing in layers and batches can you determine which content model actually works.
Form samples first, then let top-tier amplify. Reverse the order, and the money becomes an expensive fireworks show — beautiful view counts, but if the product and backend fulfillment aren't established, not a penny more will be added.
When it comes to the hardest question in creator marketing, pricing is unavoidable. Beyond follower count, what really matters?
Peng's order is interesting: first judge whether this creator can complete this task, then talk about how much they're worth. Price isn't an abstract "market value" either — it's a contract price under specific collaboration terms.
The same creator, the same video — including one-year usage rights, ad authorization, multi-platform distribution, and raw materials — is fundamentally not the same product as posting once on a personal account only.
Her team makes pricing judgments based on five categories of signals:
1. Historical quotes and real transaction prices of similar creators
2. Follower scale and growth trend
3. Median views and stability of recent content
4. Audience country, age, interests, and overlap with the brand's target audience
5. Content tone and brand match
Which metrics are the most deceptive? She says: total follower count, and average views without breaking down country and audience structure. A single historical viral hit, or high engagement driven by giveaways and general entertainment. Transaction screenshots with unverifiable statistical caliber. Quotes that only mention a one-time posting price without explaining usage rights and authorization scope.
A single-point best result only proves they once succeeded — it doesn't prove this time matches your brand.
The creator marketing industry has a randomness that keeps people up at night: invest in 100 creators, and in the end maybe only a few pieces of content actually take off.
I threw this anxiety at her. Her answer was very "Pinduoduo": "Anything can be deconstructed, right?"
But deconstructable ≠ guaranteeable.
First, define what "taking off" means. Is it views and engagement breakthrough, or generating clicks, trials, orders? The former is about传播, the latter is about commercial conversion. The two outcomes can't be mixed into the same "viral probability."
Then deconstruct five things:
1. Is the person right? Are the creator's real audience and the brand's target users the same group?
2. Why did it spread? Was it the opening, scenario, emotion, product proof, or timing?
3. Why did the target behavior happen? Does the content structure make users willing to click through?
4. What did the consumer encounter? From which scenario, which selling point, all the way to the landing page and product experience?
5. Can the result be replicated? Switch to another creator, similar audience, similar content structure — does it still hold?
So the first round of investment should be like an investment portfolio. Test in groups by audience, scenario, content format, and price range. Standardize briefs, delivery, rights, and data collection. Finally, concentrate resources on repeatedly appearing effective signals — rather than switching to a new batch of people every round to try luck again.
"We can't turn viral hits into industrial products, but we can make three things increasingly certain: discovering effective samples, controlling losses, and adding budget."
"Deconstructable" means reducing uncertainty: you know which variables to look at, what data to collect. "Not guaranteeable" means acknowledging that single results will still fluctuate: following the process doesn't guarantee a viral hit.
But looking at three rounds of investment together, discovering effective samples, stopping losses in time, and decisively retargeting — these three things can become increasingly certain.
At the end, I asked a more fundamental question: how do you judge whether a company is selling products overseas, or truly building a global brand?
She says look at four signs.
1. Can users state a reason for choosing you beyond price?
2. Have product, content, customer service, and fulfillment truly been adjusted for the target market — not just language translation?
3. Is there repeat purchase, active search, and user recommendation in growth — rather than complete dependence on the next ad spend?
4. Can effective content, creator relationships, and user feedback enter the next round for cross-cycle reuse?
She specifically mentions a common problem among Chinese overseas companies. Organizations often prioritize immediately quantifiable conversions, and underinvest in aesthetics, creative judgment, and long-term cultural expression. Many businesses look like they're growing every month, but every round requires buying traffic again and proving themselves again — leaving no user memory or brand assets.
"Truly moving toward a brand means making users willing to remember, repurchase, and recommend — so the company doesn't have to spend money every month to prove it exists."
Selling products and building a brand aren't black and white; many business actions are separated by a thin line. But when the operating mindset truly shifts from selling to branding, these four signals appear one by one.
After leaving a promotion or a platform dividend, do users still know who you are? Does the organization also know how to do better next round? That's the test.
AhaCreator itself is inserting AI into every环节 of creator screening, outreach, communication, and fulfillment. I asked her: in this process, how much work can really be handed to AI?
She gave a directional judgment: calculated by the volume of repeatable tasks that can be standardized in the future, 80% of the work can be handed to AI.
But she immediately added the boundary: this doesn't mean 80% of time has already been saved today.
Search and initial screening, information organization, first-round outreach, translation, quote summarization, anomaly alerts — these information-dense, clear-rule tasks are already suitable for deep AI participation today.
But what problem this round of budget actually solves, whether the creative has感染力, which cultural expressions will harm the brand, how to negotiate complex usage rights, and who is ultimately responsible for disputes — these must still be carried by people today.
"You can hand tasks to AI, but you can't hand operating responsibility to AI."
Her principle: AI can be responsible for reading, finding, organizing, and repeat execution. People must be responsible for goals, trade-offs, relationships, and final commitments.
So will traditional agencies be reshuffled? Her judgment: they won't disappear entirely, but the standard execution layer will be significantly compressed. Sourcing, outreach, status syncing — these tasks will increasingly be eaten by software and AI. What remains is strategy, creativity, local culture, long-term relationships, complex negotiation, and result responsibility.
In three years, AI images, AI videos, and digital humans will make content supply approach infinity. Will real creators become increasingly worthless?
She didn't answer yes or no directly. People who only provide generic materials and basic production capabilities will see their value rapidly压低 by AI. But creators with real identity, stable audience, professional judgment, and long-term credibility will become scarcer.
The closer content gets to infinity, the more users need to answer one question: why should I trust this person?
She says she started in brand advertising, went through platform commercialization, product supply, global supply chain, and now does creator supply. The more she works on the front line, the more certain she is of one thing: lists will get cheaper, but trust verified through real collaborations won't.
Susan Peng has over 10 years of frontline experience in brand marketing, platform commercialization, product & supply chain, and global growth. She previously worked at Wall Street CN, Sina Weibo, and Douban, then spent five years at Pinduoduo Group going through commercialization, Duoduo Grocery, Chaoxingxing, and TEMU. She currently leads AhaCreator's commercialization, customer growth, sales organization, marketing & industry expression, and ecosystem partnerships.
Q1: You started in brand advertising and social media, then went through multiple Pinduoduo businesses and TEMU, and now AhaCreator. You've done brand marketing, merchant acquisition, supply, platform operations, creator marketing, and supply chain. If you had to summarize it, what problem have you actually been solving these ten years?
Peng: I think although the industry and job type kept changing, at the core I've been solving the same type of problem: finding order in disorder, breaking down chaotic and scattered things into patterns, then summarizing and abstracting personal judgment into a system that can run continuously and be verified repeatedly.
My path wasn't a direct jump from product supply to creators. During college I did campus content entrepreneurship, and interned at different companies including global top PR agencies Weber Shandwick and Ruder Finn, overseas performance advertising Avazu, and the marketing department of Fortune 500 consumer brand PepsiCo. After entering Pinduoduo, my work extended from brand and commercialization to regional operations, commercial products, product supply, and TEMU's global supply. Now I face global creators.
The objects keep changing — from brands and customers, to merchants and products, to creators — but the problem hasn't: how to make scattered supply and demand meet efficiently, how to make transactions and fulfillment actually成立, and how to make one success not just belong to one capable person, but沉淀 into a system that can be reused next time.
Q2: You participated in TEMU's semi-managed business from 0 to scale. Looking back, what were the factors that truly drove growth but are easily overlooked externally? What experiences do you still use today?
Peng: From my personal experience, TEMU's two most visible pillars are supply chain & pricing capability, and growth capability. Low prices, ad spending, and creator growth were done very well — these are actually visible to the outside world.
What I felt most deeply during the TEMU phase was that the team would try to return to first principles when doing things, break out the most core variables, and filter out noise that doesn't affect core results — rather than rejecting market and user feedback. This is still the method I use today: only grasp the essence of things.
The reason this approach drives growth is that it turns abstract judgments into a fast closed loop: first distinguish process from results, use small-scale validation to separate real signals from noise, then concentrate resources on directions where positive signals have already appeared, while letting product, supply, content, fulfillment, and user experience jointly承接 growth. If any环节 can't catch it, front-end traffic is wasted.
So today when doing Creator Marketing, I also don't believe in a single viral hit. I first form real feedback, then decide where the next money and next batch of resources go.
Q3: After leaving TEMU, why did you choose the Creator Marketing direction? When did you realize that global social media content is becoming new infrastructure for Chinese companies going overseas?
Peng: My most direct judgment for choosing Creator Marketing is: it's changing from an "optional item" to a "must-have item" for more and more overseas companies.
Different reports have inconsistent statistical calibers, so I don't treat some growth number without clear scope as an external fact. But the industry is still growing rapidly. What Chinese companies going overseas increasingly lack isn't just products, channels, and technical ad spending — it's the ability to make local users understand products, build trust, and be willing to try.
In the AI era, technical ad spending won't disappear, but trust between people will become increasingly important. The core is "trust."
This field still hasn't seen a platform that can simply dominate everything. The difficulty isn't just technology — it's how to simultaneously build trust on both the brand and creator sides.
I have three more judgments:
Brand demand for mid-tier, long-tail, and vertical creators is rising. They're closer to specific audiences and real scenarios, and more suitable for forming comparable market samples.
It's essentially still matching between supply and demand sides. This is where it resembles my past platform business; the difference is that creator supply depends more on personality, culture, trust, and long-term relationships, and can't be managed with standardized inventory logic alone.
This isn't something I started关注 from Weibo, nor a sudden trend I'm chasing. I've long stood on the brand and content side, watching why people notice, understand, and trust a product. So I see global social media content as new market infrastructure. It's not a substitute for advertising — it's the前置工程 for a brand entering an unfamiliar market to understand users, build trust, and validate expression.
Q4: In the past, Chinese companies going overseas were better at buying traffic, distributing products, and doing conversions. Do you feel this model is hitting a ceiling now? What's the biggest change from "selling products overseas" to "brand globalization"?
Peng: China's supply chain is strong, and it's indeed good at buying traffic, distributing products, and doing conversions — these capabilities won't失效. But if a company wants long-term pricing power and operating compound interest, relying only on transaction efficiency will show an increasingly clear ceiling.
I think the new upper limit is mainly in three capabilities:
1. Aesthetics and creative judgment. How the product is expressed, what temperament the brand presents — can't be solved by data tables alone.
2. Value shaping and premium capability. Premium itself is the result; the premise is that the company can give reasons for choosing beyond price, build brand trust and long-term pricing power.
3. Ability to挖掘 deep user needs. Not directly translating domestic selling points overseas, but understanding how local users live and make decisions.
I've seen in many Chinese overseas companies that organizations often prioritize immediately quantifiable conversions, and underinvest in artistic sense, aesthetics training, creative judgment, and long-term cultural expression. So I always emphasize: don't just do one-time business. Truly moving toward a brand means making users willing to remember, repurchase, and recommend — so the company doesn't have to spend money every month to prove it exists.
Q5: How do you judge whether a company is "selling products overseas" or has truly started building a global brand? Are there clear signs?
Peng: Selling products and building a brand aren't black and white; many business actions look like they're separated by a thin line. But when the operating mindset truly shifts from selling to branding, relatively clear signals appear. I look at four signs:
1. User choice reasons. Can users state reasons for choosing it beyond price?
2. Real localization. Have product, content, customer service, and fulfillment been adjusted for the target market — not just language translation?
3. Natural demand starting to form. Is there repeat purchase, active search, natural discussion, and user recommendation in growth — rather than complete dependence on the next ad spend?
4. Organization starting to accumulate assets. Can effective content, creator relationships, content usage rights, and user feedback enter the next round for cross-cycle reuse?
These aren't an absolute industry standard, but they help me judge: after leaving a promotion or a platform dividend, do users still know who you are? Does the organization also know how to do better next round?
Q6: In the past you faced brands, merchants, and suppliers; now you face global creators. Are there similar platform operating patterns between merchant and creator supply?
Peng: Very similar. The most important one: supply quantity ≠ available supply. More merchants doesn't mean they all have products, prices, and fulfillment capabilities suitable for this campaign; more creators doesn't mean they're willing to take orders, have matching audiences, reasonable quotes, or can complete content as required.
The second pattern: supply must be operated in layers. Different creators suit different markets, categories, and content tasks — can't just use follower count for a total ranking.
The third: incentives and credit are equally important. After a collaboration is completed, real quotes, response speed, content quality, fulfillment performance, and mutual evaluations should all be沉淀下来, making the next match more accurate.
The real barrier of a platform isn't how many names are in the database — it's how much supply has been verified through real transactions, and whether it can simultaneously protect the long-term experience of both brands and creators.
Q7: AhaCreator now focuses on global social media content marketing. Without using terms like "creator marketing platform" or "AI marketing," how would you explain to a Chinese company going overseas for the first time what problem you actually solve for clients?
Peng: I'd say: we help a Chinese product be understood, trusted, and willingly tried by real people in an unfamiliar market; while leaving behind the effective candidates, real quotes, content feedback, and collaboration experience generated in this process, so the next round doesn't start from zero.
Specifically, the brand still decides which market to enter, what problem this budget solves, and what content meets brand requirements. We connect the大量 scattered things in between: finding the right people, confirming collaboration willingness, getting quotes, advancing production, checking requirements, and summarizing publish and content data available from the platform, then reviewing with brand feedback.
We don't make all judgments for the brand — we let the most valuable people leave their time for truly important judgments.
Q8: Many Chinese brands still understand overseas creator marketing as "find a batch of creators to post videos." From your real projects, what's the biggest problem with this understanding?
Peng: The phrase "find a batch of creators to post videos" isn't wrong, but it's too shallow — it only mentions the surface layer. The complexity of this matter is far higher than the literal meaning.
Finding people ≠ partnership established. Whether the person is a match, whether they have genuine willingness, whether the quote and schedule are feasible — all need verification.
Video posted ≠ marketing complete. Whether the content explains the product clearly, whether usage rights are明确, whether publish timing can配合, and whether the brand's own conversion funnel can catch it — all change the outcome.
One campaign done ≠ capability formed. If the first round only leaves "how many videos posted," the next round still has to request quotes again and trial-and-error all over.
The truly valuable approach is to leave answers every round: which people work, which expressions work, where the problems get stuck, which content can be reused, how the next budget should be adjusted.
Q9: When Chinese brands do overseas Creator Marketing for the first time, what are the three most common mistakes? Any failure cases that left a deep impression?
Peng: For first-time entry into overseas markets, the three most common mistakes are:
1. Directly copying successful experience from the original market. Not understanding the local market, but having path dependence because of past success.
2. Not thinking clearly who the TA and creators分别 are. The brand doesn't know its target audience, and doesn't know what creators are good at expressing.
3. Rushing to scale before the funnel is ready. Product experience, landing pages, attribution, inventory fulfillment, and content usage rights aren't ready, but the first round has already spread budget and audience widely.
One type of failure that left a deep impression is a common pattern反复 appearing across multiple projects: the brand gives the creator a very complete and very "correct" brief, stuffed with many selling points, fixed scripts, and unchangeable requirements — but product permissions, materials, or actual experience conditions aren't ready. In the end the team repeatedly revises the script, and even if the video is posted, it's just a stiff ad.
For first-time entry into a market, validating one audience segment, one scenario, and one core selling point first is more important than proving everything at once.
Q10: Should a brand find top creators, or lots of mid-tier and long-tail creators? For brands at different stages, how should this budget be divided?
Peng: I look at creators with two dimensions: one is reach scale — that is, top, mid-tier, and long-tail; the other is vertical match degree. The two can't be mixed into one hierarchy.
Top creators: A bit like minor celebrities, with stronger public attention and awareness anchor effects. Suitable for creating events, opening awareness, or承担 clear brand anchor tasks.
Mid-tier and long-tail creators: More suitable for providing comparable market samples. Can do large-scale testing in layers and batches, increasing the probability of discovering viral content patterns.
Vertical match degree: It's not a certain tier — it runs through all tiers. A top creator with the wrong audience may not be more valuable than a truly matching mid-tier creator.
For brands entering a new market for the first time, with the goal of validating audience and expression, my default order is clear: mid-tier and long-tail first, then top. Form samples first, validate audience and expression, then let top-tier amplify.
Budget can't follow a fixed ratio. Besides creators at different tiers, also reserve a sum of money not spent in advance,用于 retargeting effective candidates, purchasing content usage rights, or amplifying content that has already taken off.
Q11: Suppose a Chinese brand enters the US market for the first time, with a $100,000 creator marketing budget. How would you design the first round of testing? How many people to find, how long to test, and under what conditions to start adding budget?
Peng: $100K — I'd say do three months. More precisely, three months is the default observation window for the first round in this hypothetical scenario, not a fixed cycle for all categories; actual timing should also be adjusted based on sample shipping, product experience, content format, and purchase decision cycle.
This window should at least cover positioning & brief, phased delivery, review, and one retargeting decision. Here, $100K by default mainly covers creator collaboration and content-related budgets, not including product R&D and large-scale performance advertising; whether sample shipping, content usage rights, taxes, and subsequent amplification are included should be written clearly before launch.
I'd divide these three months into three phases:
1. Clarify positioning first. Define product, core audience, one main scenario, and the content hypothesis to validate this round, then form an executable brief.
2. Then deliver and compare in batches. Don't spend the budget all at once in the first batch; only validate a few clear audience segments and content angles. After getting real feedback, adjust the second batch of candidates and expressions.
3. Finally review and retarget. Break out effective creators, effective content structures, and funnel problems, then decide whether to add budget, purchase usage rights, do subsequent amplification, or stop a certain approach.
Headcount isn't decided first. I'd first calculate the full cost of each qualified collaboration based on real quotes, reserve budget for retargeting, content usage rights, and supplementary testing, then ensure each major audience segment or content hypothesis has more than one independent sample; only the remaining available budget determines the first batch creator count.
Whether to add budget, I look at three things:
1. Whether the same audience segment or content angle shows consistent signals across more than one creator
2. Whether the brand's own click, trial, order, or retention data shows signals consistent with goals
3. Whether quotes, fulfillment, and content usage rights support continued scaling
Only beautiful view counts, but product and backend fulfillment not established — I won't add more money.
Q12: One of the hardest questions in overseas creator collaboration is "how much is this creator really worth?" Beyond follower count, what metrics do you actually look at? Which metrics look beautiful but are actually very deceptive?
Peng: How much is a creator really worth? Beyond followers, there's a lot to look at. My order is: first judge whether this creator can complete this task, then talk about how much they're worth. Price also isn't the creator's abstract "market value" — it's a contract price under specific collaboration terms.
We no longer rely solely on follower count or some operator's experience to拍 a number. We mainly综合 five categories of signals in pricing decisions:
1. Historical prices of similar creators. When there are verifiable records, reference historical quotes and real transaction prices of creators in the same track, same scale, same platform that we can cover, to establish market anchors.
2. Follower and account scale. Look at follower规模 and growth trend, judge the account's stage, rather than just a static total.
3. Content view performance. Look at median views, stability, and commercial content performance of recent content, avoid being misled by a single viral hit.
4. Audience quality. Within available data, look at audience country, age, interests, consumption characteristics, and overlap with the brand's target audience.
5. Creator-brand match degree. Look at content tone, vertical field, brand safety, and whether the product scenario matches.
The system provides decision information around recommended price ranges, suggested starting quotes and negotiation space, comparable market references, brand match and expected value prompts; actual available output is limited by product version, specific Campaign conditions, and data coverage, and humans then adjust based on collaboration conditions and make final decisions.
Most deceptive metrics: total follower count, and average views without breaking down country and audience structure; a single historical viral hit, or high engagement driven by giveaways and general entertainment; transaction screenshots with unverifiable statistical caliber; quotes that only mention a one-time posting price without explaining usage rights, authorization scope, and specific delivery conditions.
A single-point best result only proves they once succeeded — it doesn't prove this time matches your brand.
Q13: Creator marketing inherently has strong randomness. Invest in 100 creators, and maybe only a few pieces of content actually take off. How do you understand this "viral probability"? Is there a way to gradually turn it into a more certain business?
Peng: I think viral probability can be deconstructed — anything can be deconstructed, right? But deconstructable ≠ guaranteeable.
First define what "taking off" means: is it views and engagement breakthrough, or generating target behaviors like clicks, trials, orders? The former mainly looks at传播, the latter also looks at commercial conversion. The two outcomes can't be mixed into the same "viral probability."
Specifically I deconstruct five things:
1. Is the person right? Does the creator profile match the product, are the brand's TA and the creator's real audience matching?
2. Why did it spread? Was it the opening, scenario, emotion, product proof, platform distribution, or timing that drove views and engagement?
3. Why did the target behavior happen? Do CTA and content structure make users willing to click, try, or buy?
4. What did the consumer encounter? Which scenario, information, selling point, or proof did they encounter in the content, and then through which entry, landing page, and product experience did they continue?
5. Can the result be replicated? Is it an accidental peak, or does it still hold after switching to another creator, similar audience, or similar content structure?
So the first round should be like an investment portfolio: test in groups by audience, scenario, content format, and price range; standardize briefs, delivery, rights, and data collection; finally concentrate resources on repeatedly appearing effective signals — rather than switching to a new batch of people every round to try luck again.
We can't turn viral hits into industrial products, but we can make three things increasingly certain: discovering effective samples, controlling losses, and adding budget.
Q14: You mentioned creator marketing should shift from "one-time investment" to a "trackable, reviewable, retargetable" growth system. After the first round ends, how exactly do you decide who to continue investing in, what to replicate, and what to abandon in the second round?
Peng: Before the first round starts, I'd first write a main metric and continue/stop conditions for each of four layers: audience match, content understanding, target behavior, and quotes/fulfillment/usage rights. After it ends, only compare creators under similar audience, content format, and collaboration conditions — can't put views, clicks, or orders from different tasks on the same leaderboard.
What to continue investing in: creators with correct audience, effective content, established backend signals, and willingness to collaborate again. What to replicate isn't word-for-word scripts, but verified content structures — for example, which question opening, which real scenario, which demonstration method is easiest for users to understand. What needs modification is specific bottlenecks in the funnel — if there are clicks but no trials, first check the entry and product experience, rather than immediately switching creators.
What should be abandoned: creators whose audience isn't in the target market at all, who can't complete core demonstrations, whose fulfillment持续失信, whose rights costs can't support scaling, or who only have a single unreplicable peak performance.
Each round should leave at least one judgment of what to continue doing, and also one clear practice to stop.
Q15: What kind of content truly drives conversions? Why do some content have millions of views but don't sell products, while some videos don't have high views but can consistently generate orders?
Peng: Some content doesn't have high views but can consistently generate orders because it found the right customer group and the right selling point. And some content although has millions of views, may be more entertainment-oriented, and its customer group is wrong.
I break this into two layers. The first layer is whether the content has generated qualified intent: is the audience accurate; are the selling points truly explained clearly; is the expression authentic and credible. The second layer is whether there's ultimately a transaction: after the content generates interest, can links, landing pages, pricing, inventory, trials, and product experience smoothly take users to the next step — determines whether intent can be caught.
High views but no sales usually means entertainment value and purchase intent aren't in the same group of people, or there are problems with audience region, product explanation, and backend fulfillment. Content with not-high views but consistent orders often hits a very narrow but high-intent audience, and has long-term search value.
Final conversion and retention must be confirmed by the brand's backend data — can't use platform view counts as a substitute.
Q16: AhaCreator now uses AI in creator screening, outreach, communication, content review, fulfillment, and other环节. In the entire Creator Marketing process, how much work do you think can truly be handed to AI today, and which环节 still must rely on people?
Peng: If calculated by the volume of repeatable tasks that can be standardized in the future, my directional judgment is: 80% of the work can be handed to AI. But if the question is "what proportion can already be stably handed to AI today," I wouldn't treat 80% as the current automation rate.
What's already suitable for AI to承担 or deeply participate in today: search and initial screening, historical information organization, matching suggestions, first-round outreach, translation, status reminders, quote and schedule summarization, brief completeness checks, and anomaly alerts.
What still must be responsible for by people today: what problem this round of budget actually solves, what the product's real value to users is, whether the creative has感染力, which cultural expressions will harm the brand, how to negotiate complex quotes and usage rights, how to manage important creator relationships, and who ultimately bears responsibility for disputes and anomalies.
My principle: AI can be responsible for reading, finding, organizing, and repeat execution. People must be responsible for goals, trade-offs, relationships, and final commitments. You can hand tasks to AI, but you can't hand operating responsibility to AI.
Q17: After AI removes a lot of repetitive labor, how much will human efficiency in overseas creator marketing teams change? Is measuring by "how many creators one operator can manage" accurate enough?
Peng: My directional judgment is: after the process is fully standardized, at least three-quarters of repetitive execution work hours can be compressed in the future. This is a future direction, not that the current full process has already achieved 75% work hour savings.
Sample-shipping short videos, long videos requiring complete product experience, hardware projects, and complex usage-rights collaborations aren't the same type of work to begin with, so I wouldn't give a "how many managed before, how many in the future" applicable to all projects just to sound good.
More important than headcount is: how many qualified collaborations can each operator simultaneously推动; how high are the complexity and anomaly rates of different projects; have on-time delivery rates, rework rates, and brand risks worsened; after the first round ends, how many creators and content can enter retargeting.
If I must measure how many creators one person can manage in the future, I'd look at the qualified concurrent collaboration count weighted by project complexity, then bind it to delivery rate, rework rate, and retargeting rate.
The efficiency change AI brings isn't just letting one person chase more emails — it's letting operators exit from finding email addresses, copying messages, chasing status, and organizing spreadsheets, and put their time into understanding products, calibrating content, and managing key relationships.
Q18: If AI can in the future automatically find creators, communicate, review content, track data, and even complete retargeting decisions, what value is left for traditional overseas creator agencies? Will there be an obvious reshuffling in this industry?
Peng: My judgment isn't "will agencies be replaced across the board," but which values will remain. At least in the visible stage, traditional agencies won't disappear entirely; but in the AI era, no fixed process or job division is naturally immune from technological restructuring. The core is still people-oriented, oriented to people's needs.
The real question isn't defending old jobs, but after technological restructuring, can people put their time into higher-value judgment, creation, and relationships.
Specifically, three things will happen in the future:
1. The standard execution layer will be compressed. Sourcing, outreach, status syncing, and basic process management will increasingly be automated by AI and software.
2. Complex professional capabilities remain important. Full-case marketing, strategy & creativity, local culture & public opinion, long-term creator relationships, complex negotiation & production, content rights, compliance, and crisis handling can't be simply replaced in a short time.
3. Agencies will become more like brand operating partners. They'll become an extension of in-house teams' capabilities, leveraging platform and AI infrastructure to concentrate human resources on high-difficulty judgment, relationships, and result responsibility.
Q19: You cover many different countries and regions. From actual projects, what cultural and content differences are Chinese brands most likely to underestimate? Are there cases where the Chinese team thinks a brief is very good, but it completely fails overseas?
Peng: What Chinese brands most easily underestimate is: translating the language doesn't mean translating the scenario, trust, and expression habits. Selling points that headquarters considers very complete may look like just corporate self-introduction to local users; scripts that headquarters considers very精致 may not sound like the creator's own words when read out, and the audience can easily treat it as a stiff ad.
The same brief problem has appeared across multiple projects: functions written very completely, every line required to appear, but without answering "why would this person use it in their own life."
A more effective approach is to clearly state non-negotiable information — facts, compliance, core functions, and forbidden areas — while giving creators enough space to complete expression with scenarios familiar to local users and their own language.
The localization I recognize isn't letting overseas creators imitate Chinese teams, but letting the brand's core value be naturally spoken by local people.
Q20: Looking three years ahead, when AI images, AI videos, and digital humans make content supply approach infinity, do you think real creators will become less valuable, or instead become scarcer because of personality, trust, and real relationships? What will global Creator Marketing ultimately become as a business?
Peng: My judgment is that both changes will happen simultaneously. People who only provide generic materials and basic production capabilities will see their value rapidly压低 by AI; but creators with real identity, stable audience, professional judgment, and long-term credibility will become scarcer.
The future isn't that all real creators become more expensive — rather, value will further differentiate. What's truly valuable isn't just being able to produce one piece of content, but being able to represent a细分 group's real experience, influence user choices, and be willing to bear reputation costs for their own recommendations.
My career started in brand advertising and social media, then platform commercialization, merchant & product supply, commercial products, and global supply chain, now creator supply. The more I work on the front line, the more certain I am: lists will get cheaper, but trust verified through real collaborations won't.
I think global Creator Marketing will ultimately form an interconnected network of transactions and decisions, rather than being simply dominated by one platform. What brands ultimately need isn't just one video — it's the ability to better enter a market, understand a group of users, and continuously improve decisions.
*Full-text fidelity gate passed: 0 substantive omissions*
*Delta public-web standing authorization granted 2026-09-05*