Overseas paid ads: how to split budget across Google, Meta and TikTok

Almost every founder who starts running overseas ads asks the same question: "Should I put my money in Google, Meta, or TikTok, and what is the right ratio between them?" The question itself hides a trap. It assumes the answer is a fixed split, like fifty-fifty or forty-thirty-thirty. The truth is that budget allocation is not an arithmetic problem. It is a strategy problem about where your customer is in their journey and what you want each dollar to actually do. The same ten thousand dollars, spent on a mature, demand-clear category versus a brand-new concept that needs market education, can be split in completely opposite ways.

This guide is the long version of a conversation we have with almost every seller who comes to us with a product, a store, and a budget, but who has been spun in circles by three platform dashboards and a hundred "media-buying secrets." We will make a few things concrete: what role each of the three platforms actually plays in the funnel, the logic for splitting budget across funnel layers, how to spend the first dollar when you only have one or two thousand a month, how to read ROAS, CAC, and LTV without fooling yourself, and the pitfalls that quietly burn the budget. Where we give numbers, we label them as illustrative or typical ranges, because category, market, and average order value vary enormously, and pinning a precise-looking number on your account would be irresponsible.

One scoping note first. This is mostly about paid-media budget allocation for DTC, independent sites, and brand ecommerce, and it applies to most B2C going-global cases. If you are a factory selling B2B to importers and distributors, the acquisition logic is very different, so start with the China B2B export playbook instead. If you are at the very earliest stage with no customers yet, cold-starting your first 100 customers is the more relevant piece. Ready? We start with the single most important idea.

What job does each platform actually do?

Short answer: the three platforms are not three options for doing the same thing. They are three different stations on the funnel. Google captures demand that already exists, Meta creates demand and handles retargeting, and TikTok seeds demand by waking up latent wants with content. Confusing their roles is the most common and most expensive mistake in overseas media buying, because it makes you hold a demand-capture expectation against a demand-creation channel, then conclude "this platform doesn't work."

Once you internalize this, you understand why you cannot measure them with the same ruler. A Google Search ad will usually post a higher ROAS than a TikTok in-feed ad by nature, because it serves the bottom of the funnel, people who already raised their hands. TikTok's value often shows up as new demand it created and fed downstream for other channels to harvest. If you compare all three on first-order ROAS alone, you will almost certainly kill the demand-creating channels, then wonder why the bottom-of-funnel channels keep getting more expensive and harder to scale. Let's take each platform in turn.

Google: catch the customer the moment they raise a hand

Google, including Search, Shopping, and Performance Max, is fundamentally about demand capture. When a US consumer searches "best wireless earbuds under 100" or types your brand name, they have already expressed clear intent. Your job is not to convince them they want it; it is to make sure you are in front of them at the moment they raise their hand. That is why search ads usually show the highest conversion rate and ROAS of the three platforms: you are buying ready-made, warmest demand.

But Google has a ceiling: it can only capture demand that already exists. If nobody is searching for your category, or you sell a brand-new concept, the search volume itself is tiny and no amount of optimization conjures demand out of nothing. So Google is your "floor." It gives the steadiest, most predictable return and is the most worth protecting in your budget, but it is usually not the source of your growth ceiling. Shopping matters a lot for ecommerce with clear products and prices, and brand-term coverage is a must: if you do not show up when people search your own name, you hand that traffic to a competitor's conquesting ad.

Meta: both a demand creator and the strongest recycling bin

Meta, meaning Facebook and Instagram, plays two jobs in the funnel, which is exactly why it is so underrated. The first job is demand creation: through precise interest and lookalike targeting, it pushes your product to people who do not yet realize they want it but are likely to be tempted, something Google cannot do. The second job, and often the more profitable one, is retargeting and recycling: pulling back the people who saw an ad, visited your site, or added to cart but never checked out.

That recycling function is a profit source many beginners badly underestimate. A cold prospect seeing your ad for the first time converts at a naturally low rate, but someone who already put an item in their cart is one nudge away. Meta's retargeting ads can often recover those high-intent users at very low cost, frequently posting a ROAS far above cold prospecting. This is why "drive traffic in with other channels, then close with Meta retargeting" becomes such an efficient combination. Meta is also a great place to test creative angles, because its audience scale and feedback speed make it quick to judge which selling point or visual lands.

TikTok: seed the demand with content

TikTok's role is top-of-funnel seeding and demand creation. What it does well is plant a "want" into someone's head with one infectious short video, at a moment when they had no intent to buy, the famous "TikTok made me buy it" effect. For products with strong visual appeal, a demo-worthy moment, or a story (beauty, home gadgets, novelty goods), TikTok's breakout potential is something the other platforms cannot match.

But TikTok is also the hardest to please, and that is a direct consequence of its role. It is brutally creative-dependent: whether a piece breaks out depends far more on content quality than on media-buying technique, so it demands high content output. Its attribution is also dirtier: many people TikTok seeds will not click and buy on the spot, but will Google your brand name days later or get recovered through Meta retargeting, which means the value TikTok actually created often gets booked to downstream channels. If you only read TikTok's isolated in-dashboard ROAS, it is easy to undervalue and cut it too early. This is why TikTok usually should not be your first dollar, but the channel you commit to seriously once you have content capacity and attribution that can see cross-channel contribution. Influencer content is an efficient way to feed TikTok, which we cover in detail in the overseas influencer marketing playbook.

So what logic do you actually split budget by?

Short answer: split by funnel layer, not by platform average. Decide whether each dollar goes to capturing demand (bottom), creating demand (top), or recovering people who didn't convert (retargeting), then plug in the right platforms and amounts. The same platform can appear in multiple layers (Meta both creates demand and retargets), so "split by role" is far more accurate than "split by platform."

A useful mental model is to picture three pools of budget. The bottom pool is the harvest layer: mainly Google Search and Shopping plus Meta retargeting, the spend with the most certain return and highest ROAS, which you should fund first and which rarely struggles to spend, as long as existing demand is there to harvest. The middle pool is the demand-creation layer: mainly Meta cold audiences, keeping fresh prospects flowing into the top of the funnel, because otherwise the harvest layer eventually starves. The top pool is the seeding and expansion layer: TikTok and influencer content, creating brand-new, larger demand, slow to kick in and dirty to attribute, but the thing that decides your ceiling. A healthy account has all three turning, not just money piled in the easiest-to-see harvest layer. Harvesting without seeding is eating your stock and saving no seed, which drains the pool eventually.

This logic also explains why the "right ratio" changes over time. Early in cold start you need certainty and fast validation, so the money should lean clearly toward the harvest layer. Once the market is proven and you start chasing scale, the demand-creation and seeding layers should grow, because the harvest layer hits a ceiling quickly. The table below lays out where to weight at different stages. The figures are illustrative ranges that show direction only.

The honest way to read this table is not to copy the percentages but to see the trend: the further along you go, the more weight seeding and demand creation deserve, because pure harvesting eventually hits the ceiling of existing demand. Note too that categories of different demand maturity start in different places. A demand-clear standard category (say, a class of consumer-electronics accessory) can put more on Google from day one, while a new-concept product that needs market education may have to spend more on Meta and TikTok from the start, because there is simply no search volume to harvest yet. This is the same tradeoff as "build a brand or rent platforms," which we develop in the B2C growth playbook.

With only one or two thousand dollars, how do I spend the first dollar?

Short answer: don't spread out, concentrate. On a small budget, fund a few high purchase-intent Google Search keywords first, push your limited money into producing clean conversion data, and derive a rough CAC. Then put a small amount into Meta to test retargeting and lookalikes. Do not open three platforms and ten ad sets on day one; once budget is spread thin, no set gets enough data, so nothing learns and nothing performs.

The biggest enemy of a small budget is data dilution. Ad systems, whether Google or Meta, need enough conversion samples to learn how to put your ad in front of the right people. If an ad set cannot even gather single-digit conversions a week, the system is essentially flying blind, and most of your money is tuition. So the first principle on a small budget is to concentrate spend on as few, most-likely-to-convert places as possible, get one channel running, and only then talk about expansion. That is why we have you start with high-intent Google Search keywords: it is the warmest, easiest-to-convert demand on the web, with the cleanest learning samples.

A cold-start test rhythm you can run directly

Here is "how to spend a small budget" broken into an order you can check off, suited to sellers starting somewhere between one or two thousand and five thousand dollars a month.

1. Protect your brand terms first. However small the budget, spend a little to defend your own brand searches. The cost is tiny, and it blocks a competitor from conquesting your hottest traffic. This is the highest-return dollar you will spend.
2. Concentrate on 3 to 5 high-intent keywords. Pick terms with clear buying signals (containing "buy," "best," "price," a specific model), put most of your budget on them, run two to three weeks, and aim to accumulate enough conversions to estimate a rough cost per customer.
3. Add Meta retargeting once Search works. By now your site has visitors and add-to-cart data. Open one Meta retargeting set to recover those high-intent people. This is usually your prettiest early ROAS.
4. Test Meta cold audiences in small steps. Use 2 to 3 creative angles to test lookalikes and core interests, capped so it does not disturb the harvest layer. The goal is to find creative that lands, not to demand ROAS immediately.
5. Add TikTok only once you have content capacity. When you already have steady conversions and can produce short videos consistently, put a small budget into TikTok for expansion. Entering before that is mostly paying tuition to the platform.

The inner logic of this order is "certainty first, exploration second": use the easiest-to-convert channel to revive data and cash flow, then use the confidence you earned to test higher, less certain channels. Reversing this order, going heavy on TikTok expansion from the start, is the most common way small-budget sellers crash. For the earliest "no customers yet" problem, cold-starting your first 100 customers has a more detailed playbook.

How should I actually read ROAS, CAC, and LTV?

Short answer: do not stare at first-order ROAS alone. First calculate your break-even ROAS from your gross margin as a floor, then use CAC (cost to acquire a customer) against LTV (lifetime value) to decide whether a channel is worth keeping. Many channels lose on the first order yet are extremely worth funding because repeat purchases lift LTV, and a first-order-only view wrongly executes them.

Start with the biggest trap in ROAS (return on ad spend): the number in your platform dashboard almost never equals your real profit. That figure is usually just "ad-driven revenue divided by ad spend." It does not subtract product cost, shipping, returns, or payment fees. A 3x ROAS in a category with only 30% gross margin may actually be losing money. So the first thing to do is calculate your break-even ROAS, which is the inverse of your gross margin. The illustrative table below shows that relationship.

The use of this table is to convert "what ROAS is healthy," a question with no universal answer, into one you can compute: does your actual ROAS stay above the break-even line for your margin? The thinner the margin, the higher that line and the smaller your room for error. This is also why thin-margin categories are especially hard to run paid acquisition on: you have almost no slack to make mistakes.

CAC and LTV: move your gaze from first order to lifetime

What really decides whether a channel is worth keeping long term is not first-order ROAS but the relationship between CAC and LTV. CAC is the average ad spend to win one customer; LTV is the total gross margin that customer contributes over the whole life of their relationship with you. A frequently cited rule of thumb is that LTV to CAC should be roughly 3 to 1 to be healthy (an illustrative heuristic, not a hard threshold), meaning the value a customer contributes should be clearly larger than the cost of winning them.

This lens corrects a fatal misjudgment. Suppose a channel only breaks even, or loses slightly, on the first order. On first-order ROAS you would shut it off immediately. But if that channel brings customers with high repeat rates and long LTV, it is actually the channel you should fund most, because you are trading a slightly unprofitable first order for a long-term, high-value customer. Categories with weak repeat and low order value almost must profit on the first order, while those with strong repeat, high order value, or subscription can fully absorb a first-order loss to win customers. Figure out which one you are first, then decide how to read these numbers. Getting this math right requires clean tracking and attribution, which we cover in overseas analytics and attribution with GA4.

Why is my ROAS fine but I am still not profitable?

Short answer: most likely inflated attribution plus ignored real margin. Platforms all tend to credit a sale to themselves, so the three dashboards added together over-claim; and dashboard ROAS does not subtract product cost, returns, or fees, so the pretty number never lands in the bank.

This is one of the most maddening forms of "paper prosperity" in overseas media buying, and it is usually caused by two layers stacking. The first is double-counted attribution: every platform's dashboard tends to claim a sale as its own. A user seeded by TikTok, caught by a Google brand term, and finally clicked through Meta retargeting may appear in all three platforms' reports for the same order. So when you sum "revenue" across the three dashboards, you find it exceeds your true total revenue in Shopify. This is why any single platform's isolated ROAS is distorted, and why you need one independent, unified attribution standard (such as GA4 or your ecommerce backend as the source of truth) rather than listening to three platforms each telling their own story.

The second layer is ignored real margin. Dashboard ROAS uses revenue, not profit. Once you subtract product cost, cross-border shipping, returns and refunds, payment and platform fees, even support cost, that 4x ROAS may be barely break-even. Return rates in some categories (apparel, footwear) are vicious enough to eat all the paper profit. So the only reliable way to judge whether you actually made money is to pull the math down to real landed gross margin, not stop at the reassuring number in the ad dashboard. Chargebacks in cross-border payments are another hidden cost that often gets overlooked, which we develop in cross-border payment and chargebacks.

Should I use automated bidding or manual control?

Short answer: it depends on data volume. During cold start, when conversion samples are scarce, handing the account to fully automated conversion-goal bidding too early means the system cannot learn and spends wastefully; first accumulate enough conversion data with cost-controlled bidding, then switch to smart bidding and let the system take over.

Smart bidding (Google's Maximize Conversions and target ROAS, Meta's highest value, and the like) is fundamentally machine learning, and it needs to be fed data to learn. That is the catch: during cold start you have exactly no data. If you open a fully automated, conversion-value-goal strategy now, the system is like a model asked to predict after seeing only a handful of samples; it can only guess, and the cost of guessing is your real money. The steadier path is two phases. Early on, use cost-cappable, relatively controllable bidding (or first optimize toward a denser shallow event like add-to-cart or landing-page view) to accumulate a meaningful volume of conversion samples. Once a single ad set can produce a steady number of conversions per week, switch to conversion or conversion-value smart bidding; now the system has data to learn from and finally shows its power.

There is a common misread here too: many people blame "smart bidding doesn't work" on the algorithm, when the real reason is usually that the ingredients fed to it were too thin, either too few conversions or conversion tracking installed wrong in the first place. Before you doubt your bid strategy, confirm your conversion events and pixel or conversions API are installed correctly and counting accurately. An account that cannot count conversions accurately cannot be saved by any bid strategy.

Underperforming: switch platforms or change creative?

Short answer: fix the creative and landing page first; do not rush to switch platforms. The vast majority of performance problems are about whether the content lands and whether the landing page catches, not the platform itself. Diagnose in the order "click-through rate, then landing-page conversion rate, then audience or platform," so you neither blame the platform unfairly nor flip back and forth wasting data.

This is another hidden channel through which budget slowly burns: every time performance dips, you switch platforms and start over. The result is that each platform runs only three to five days, none accumulates enough data, and you fall into a loop where you have tried everywhere and nothing took off, while the money is fully spent. The more effective approach is to break "underperformance" into a diagnostic chain, starting with the most likely cause.

• Look at click-through rate (CTR) first. Low CTR means the creative or audience did not match; people simply were not hooked by your ad. What to change here is the creative angle, the hook, the visual, or to narrow the audience, not the platform. Creative is the single highest-return thing you can optimize in overseas media buying.
• Then look at landing-page conversion rate. If people click through but do not buy, the problem is the handoff: slow page load, weak trust, price or shipping scaring them off, a checkout that is too long. When this layer leaks, the harder you spend up front, the faster you lose.
• Only then suspect the audience or platform. Only when the creative hooks and the page converts, yet volume still does not come, is it worth reconsidering whether the platform or audience is wrong, for instance your product is simply not suited to TikTok's content ecosystem, or your target audience is not on that platform at all.

Keep that order and you avoid "frequent platform switching," the most expensive bad habit. Creative and landing page are the cause; the platform is just an amplifier. Swapping amplifiers before fixing the cause just replays the same problem in a new place. The landing page and the independent site itself directly decide how efficiently all your paid traffic is caught, which you can dig into in the DTC independent site launch guide.

The pitfalls that quietly burn ad budget

Most overseas media programs do not collapse at once; they leak dry through a pile of small holes. Here are the ones we see most, each a quiet tax on your profit.

Racing the three platforms and keeping only the highest ROAS

The most common error is to line Google, Meta, and TikTok up and compare first-order ROAS, then cut the ugliest number, usually the upper-funnel channels that create and seed demand. That is keeping the harvest and cutting the seeding. Short-term ROAS looks better, but months later the harvest layer gets more expensive and harder to scale because no new demand is fed into the funnel. The fix is to set expectations by role: do not demand a harvest channel's ROAS from a seeding channel.

Pouring traffic onto a landing page that cannot catch it

This is the same pitfall as in the B2B piece, and it is even more lethal in DTC. Every click you paid for, landing on a slow, cluttered, low-trust, checkout-heavy page, is water poured into a leaky bucket: the harder you spend, the faster it drains, and the leak is invisible because spending always looks productive. Always fix the landing page and checkout before scaling spend.

Too little creative, and unwilling to keep producing

On Meta and TikTok especially, creative fatigues. Run the same set long enough and CTR and ROAS necessarily decay because the audience is sick of it. Many accounts stall not because of bidding or budget but because creative output cannot keep up: the same two or three videos for a month, run into the ground. Building "a steady stream of new creative angles" as an ongoing capacity, not a one-time task, is the foundational discipline that keeps media healthy. Influencer and creator content is an efficient source of that creative supply.

Chasing vanity metrics instead of profit

Impressions, video views, likes, follower growth all look like progress, yet rarely correlate with money in the bank. What to watch is CAC, ROAS at the real-margin level, repeat purchase, and LTV. An account can post huge view counts and lose money the whole time; it can also look unremarkable and quietly make money. Do not let the cheerful big numbers in a platform dashboard define "success" for you.

Betting everything on paid traffic with no owned asset

Paid media is fundamentally renting traffic: the day you stop spending, the traffic hits zero. If your growth is 100% dependent on ads, you work for the platform forever, at the mercy of its price swings and policy changes. The smart move is to use paid media for fast validation and ramp while using SEO, GEO, content, email, and owned channels to slowly reduce your dependence on paid traffic. Paid buys the immediate; organic builds the asset. We unpack the difference and the interplay in GEO versus SEO in 2026.

A stage-by-stage roadmap you can run

Here is the thinking assembled into one executable path, using a typical DTC going-global seller as the example. The figures are illustrative; real timelines depend on your category and budget size.

1. Stage 1, get the harvest layer working (about weeks 1 to 6). Protect brand terms, concentrate on Google high-intent keywords and Shopping, open one Meta retargeting set. Goal: install conversion tracking correctly, derive a first real CAC, confirm the harvest layer produces orders steadily.
2. Stage 2, add the demand-creation layer (about weeks 6 to 12). On a stable harvest base, open Meta cold audiences, systematically test creative angles, and find 2 to 3 that land. Goal: keep fresh prospects entering the top of the funnel rather than only eating the stock.
3. Stage 3, add the seeding/expansion layer (about week 12 onward). With content capacity and cross-channel attribution in place, enter TikTok for expansion and layer in influencer content to amplify validated creative. Goal: lift the growth ceiling while a unified attribution view shows each channel's true contribution.
4. Throughout, monthly review plus organic in parallel. Review each channel monthly on real margin, tilt money toward what is working; meanwhile start SEO/GEO and email/owned channels to steadily cut long-term dependence on paid traffic.

This order deliberately puts certainty before exploration: use the steadiest harvest layer to grow cash flow and data first, then expand layer by layer into less certain but higher-ceiling channels. Doing it in reverse, going heavy on seeding from the start, is the most common way new sellers burn money.

Metrics to watch and metrics to ignore

You optimize what you measure, so measuring the wrong thing means optimizing the wrong thing. The goal of paid media is sustainable, profitable growth, so your dashboard should track signals that move toward that outcome, not vanity numbers that look busy but cannot prove profit.

A simple monthly ritual: pull these few numbers every month, including real-margin ROAS, CAC trend, LTV to CAC, channel contribution under unified attribution, and blended profitability. Ask one question of each: month over month, is it moving in the right direction? You are not chasing a magic number but watching a small set of leading indicators move together toward sustainable profit. When CAC falls, LTV to CAC rises, and blended profitability turns positive, your budget allocation is working the way it should.

How this fits your larger growth picture

Paid media rarely lives alone. Its healthiest state is feeding and being fed by organic, content, and owned channels: customers bought through paid get deposited into email and owned audiences for repeat purchase; owned traffic from organic search and GEO reduces your long-term dependence on paid; influencer content both feeds the creative library for paid and drives a seeding effect of its own. If you want to assemble this combination into a repeatable system rather than a set of disconnected accounts, the B2C growth playbook gives the bigger-picture view. The point is that the allocation logic here is not a one-time deal but a system that keeps adjusting with your stage, onto which other growth moves can plug.

How Ignite runs it, and what we don't do

We want to be precise about how we help, because going-global is full of over-promising vendors. Ignite Consulting LLC is a US-registered growth and AI-visibility consultancy serving Chinese brands expanding overseas, and our job is to assemble paid media, search and AI visibility, PR, and influencers into one interlocking system, not to sell you any single channel as a cure-all.

On paid media specifically, we help you plan budget by funnel role rather than platform logo, build clean conversion tracking and unified attribution, find creative that lands through disciplined testing, and add or subtract based on real margin rather than vanity dashboard ROAS. We do not promise any "guaranteed return" or "fixed ROAS"; anyone making that promise is over-selling. On influencer and creator work, we charge only an agency fee for managing the relationship, and the creators' own fees are listed separately and fully transparently. The goal of every layer is the same: that every dollar you spend goes toward sales at a defensible margin.

If you want to see where you actually stand before deciding how to split budget, the fastest path is a free visibility and media audit: we map your account structure, attribution setup, and where the landing-page handoff leaks. For what we deliver, see our Paid Media and SEO and GEO services.

Frequently asked questions