Digital marketing for startups: which channels to start with and how to measure them

The marketing question a startup actually faces is not which channel is best. It is which two or three channels can give you a readable signal before the cash runs out. That single constraint is what separates digital marketing for startups from the marketing a large brand runs, and it changes the channel order, the budget logic and the numbers you report to a board.

What follows is the channel and measurement half of the problem: which channels to shortlist, what each one demands from you, how fast each one answers, and how to calculate the unit economics that decide whether a channel stays or goes. Writing the plan document itself, the objectives, the positioning and the template you fill in, is covered separately in the guide to building a startup marketing plan on a limited budget, so it is not repeated here.

Why startup marketing is a different job from big-company marketing

A large brand in Thailand is usually defending a position it already holds. It knows who buys, it knows what the price should be, and the marketing job is to hold share and stay in the consideration set. Budgets are annual, the calendar is agreed months ahead, and reach is a legitimate objective because the product-market question was settled years ago.

A startup has settled none of that. The first job of your marketing is to find out whether a specific audience will pay a specific price for a specific promise. Every campaign is partly a test, and a test that returns a clean negative is worth as much as one that returns a positive. Four things follow from that.

• Product-market fit comes before brand share. You are not trying to be remembered by everyone. You are trying to find the narrowest group that converts at a rate you can afford, and then widening from there.
• Unit economics come before reach. Impressions, followers and video views do not tell you whether the business works. Cost per acquired customer measured against the margin that customer leaves behind does.
• A growth loop beats a campaign calendar. A calendar spends money and stops. A loop is a mechanism where the output of one cycle feeds the next: a customer produces a review, a referral, a piece of content or a data point that lowers the cost of getting the following customer.
• Speed of learning beats polish. A large team can afford to spend six weeks on a hero video. You are buying information, and information priced in weeks of runway is expensive.

This is also why copying a large competitor's channel mix is a common and expensive mistake. Their mix is optimised for a problem you do not have. Their brand campaigns work because their search demand already exists; yours does not exist yet, so a brand campaign has nothing to catch.

The seven channels worth shortlisting at the start

Almost every early-stage channel decision in Thailand comes down to the same shortlist. What separates them is not quality. It is how long each one takes to give you a usable answer, and what it asks of you besides money.

Search engine optimisation. SEO is good at capturing demand that already exists and at compounding: a page that ranks keeps returning traffic without further spend. It is the slowest of the seven to answer. A new domain has no history, and Thai keyword research is its own discipline because Thai is written without spaces between words, users mix Thai and English in the same query, and brand and product names appear in both scripts. SEO suits a startup whose customers are actively searching for a known category and whose runway is long enough that a compounding asset makes sense. It demands content inventory and patience. If you have neither, it is the wrong first channel, though it is rarely the wrong second one. The mechanics for the Thai market are set out in more detail on the SEO service page.

Google Ads. Paid search buys the same existing demand as SEO, but immediately. It is the fastest honest read on whether people are looking for what you sell and what they are willing to click through to. It suits any startup selling into a category people already name and search for. It demands a landing page that converts, a conversion tracking setup that actually fires, and enough daily budget for the campaign to leave the learning phase. The common failure is not the bidding; it is running so little volume that the automated bidding never gets the conversion data it needs, and then judging the channel on that. The Google Ads service page covers the account structure side.

Meta Ads. Facebook and Instagram ads create demand rather than capture it, which makes them the right tool when nobody is searching for your category yet. They answer quickly, sometimes within days, because the audience is large and the platform will spend your budget fast. They suit startups with a visual product, a clear before-and-after, or an offer that makes sense to someone who was not looking for it. The cost is creative volume. One ad is not a test. You need enough distinct concepts, not just colour variants, for the platform to find the one that works, and you need to keep producing them because creative fatigue is the main reason a working Meta campaign stops working. There is more on account and creative structure on the Facebook Ads page.

TikTok. TikTok reaches attention rather than intent, and its distribution is less dependent on how many followers you have than most platforms, which is unusually friendly to a company nobody has heard of. It answers fast on the attention question and slowly on the revenue question, because the gap between a video performing and a customer paying can be long. It suits founders who are willing to appear on camera and produce at a real cadence. The cost is production rhythm. A startup that can post several times a week gets a signal; one that posts monthly is not really testing the channel. See the TikTok Ads page for the paid side of it.

Content marketing. Content is the raw material the other channels consume. Comparison pages feed paid search quality, explainer posts feed SEO, short-form scripts feed TikTok, and the same material answers the questions your sales conversations keep repeating. It is slow to show a direct return and hard to attribute cleanly, which is why underfunded startups cut it first and then wonder why their ads have nothing to say. It suits businesses with a considered purchase, a long decision cycle, or a category the buyer has to be educated about. It demands writing capacity you may not have.

Email. Email is the only channel on this list you own outright. It has no algorithm between you and the recipient, and its economics improve as the list grows because the marginal cost of another send is close to nothing. It gives you almost no signal in month one for the obvious reason that you have no list, and it becomes one of your best-performing channels later. It suits any startup with repeat purchase, a trial period, an onboarding sequence or a long sales cycle. It demands that you collect addresses from day one, which is a decision you make before you need it.

Micro-influencer. Working with small creators buys borrowed trust and, unlike a large campaign, it can be run one creator at a time. It answers moderately fast and it produces a second asset almost for free: creator content that you can run as paid ads, usually outperforming brand-made creative. It suits consumer products, food and beverage, beauty, apps and anything a person can be filmed using. It demands relationship management and sample or product inventory, and it carries a specific risk in Thailand, where advertising disclosure and health or cosmetic claims are regulated and the brand carries the consequence of what a creator says.

How the seven channels compare

The obvious column to want here is a starting budget in baht per channel. That column is missing on purpose. There is no reliable public benchmark for what a Thai startup should spend on any of these channels; the honest figure depends on your price point, your margin, your category's auction competition and your runway, and any baht number printed as a market rate would be invented. So the fourth column has been replaced with the cost that people usually forget to count, which is what the channel takes from you that is not money. Read the timings as orders of magnitude, not promises.

A workable early shortlist is usually one fast channel that buys existing demand, one fast channel that creates demand, and one slow compounding asset started early because it takes the longest. Which of the seven fills each slot depends on whether your category is already searched for.

A 30, 60 and 90 day playbook

The point of a fixed window is that it forces a decision date. Without one, a channel that is quietly failing keeps drawing budget because nobody agreed in advance what failure looks like.

1. Days 1 to 30, instrument before you spend. Define the one conversion that means money, not a proxy for it. Get that event firing correctly and check it yourself by completing the action end to end. Add a single free-text field to your form or chat flow asking how the person heard about you, because at low volume that field will out-perform your analytics. Set up a way to record cost per channel weekly. Then pick two channels, not five, and put enough behind each one that it can produce conversions rather than a trickle. Write down, before launch, the CAC at which each channel is worth keeping.
2. Days 31 to 60, test the variables that actually move things. The order is offer, then audience, then creative, then landing page, then bid settings. Founders usually start at the bottom of that list because it is the easiest thing to change, and then conclude the channel does not work when the real problem was the offer. Change one meaningful variable at a time, give each test enough conversions that the difference is bigger than the noise, and resist the urge to declare a winner after a good weekend.
3. Days 61 to 90, scale or kill on the rule you wrote down. Scale means increasing budget in steps small enough that performance can recover between them, and watching whether CAC holds as volume rises. It usually does not hold perfectly; the question is whether it stays inside your limit. Kill means stopping, not halving. A channel kept alive on a token budget produces neither results nor information.

State the kill rule as a principle rather than a number, because the number is yours alone: a channel is killed when it has had enough conversion volume for its cost per customer to be distinguishable from noise, and at that cost the payback period is longer than the runway you can fund. Two clauses matter equally. Killing a channel that has not yet produced enough conversions is not a decision, it is a guess. Keeping one whose payback you cannot fund is a slow way to run out of money.

The metrics that decide everything: CAC, LTV and payback period

Three calculations carry the weight, and all three are arithmetic you can do in a spreadsheet. What makes them hard is deciding honestly what goes into them.

Customer acquisition cost. CAC is the acquisition cost of a period divided by the new customers acquired in that period. The argument is always about the numerator. Paid CAC counts media spend and the costs directly tied to it, and it answers a narrow question: is this specific channel buying customers at a price you can live with? Blended CAC counts everything you spend to grow, including salaries, agency fees, tools, creative production and sample costs, divided by every new customer however they arrived. It answers a different question: what does this company actually pay for a customer? Paid CAC is the number you optimise a campaign with. Blended CAC is the number that tells you whether the business works. A startup reporting only paid CAC is usually flattering itself, because organic, referral and founder-sold customers are being carried by costs that never entered the calculation.

Lifetime value. The mistake that ruins most LTV numbers is building them on revenue. Revenue is not what a customer gives you. Use contribution margin instead: revenue per order minus cost of goods, payment processing fees, packaging, delivery, returns and the variable support cost of serving that order. A workable form is average contribution margin per order multiplied by purchase frequency in a period, multiplied by the number of periods the customer is expected to stay. Where you have a stable churn rate, the same idea is expressed as contribution margin per period divided by the churn rate for that period. Either way, an LTV built on revenue can be several times larger than the truth, and the gap is exactly the amount by which you will overspend.

Payback period. Payback is CAC divided by the contribution margin the customer produces per month, and it is the number that matters most when you are cash-constrained. A healthy LTV to CAC ratio describes a good business eventually. Payback describes when the money comes back, and a startup does not die of a bad ratio, it dies of an empty bank account. A ratio can look excellent while the cash returns over three years, which is useless if your runway is nine months. When the two disagree, follow payback.

No reliable public benchmark exists for a good CAC, a normal LTV to CAC ratio or a standard payback period for Thai startups. Anyone quoting one is quoting a foreign market, a different category or a number they made up. Derive your own instead: take the monthly contribution margin per customer, decide how many months of payback your cash position can survive, and multiply the two. That product is your maximum allowable CAC, and it is the only benchmark that has any authority over your account.

Attribution is unreliable at low volume, and you should plan for that

Every platform reports the conversions it believes it caused, and at startup volumes those reports overlap, disagree and quietly include modelled estimates. Consent choices, browser restrictions and cross-device journeys remove part of the raw data, and platforms fill the gap with modelling that is more reliable at thousands of conversions a month than at thirty. Add the Thai pattern where a purchase is discussed and closed in chat, off the website entirely, and the attribution picture gets worse.

The practical responses are unglamorous and they work.

• Treat blended CAC as the number of record and platform-reported CAC as a directional input. If the platforms claim more conversions than your accounting shows customers, believe the accounting.
• Ask people directly. A one-line self-reported source field on the form or in the first chat reply is imprecise, but at low volume it is often closer to the truth than the dashboards.
• Test by turning things off. Pausing a channel for a defined period and watching what happens to total new customers is a blunt instrument and a surprisingly honest one.
• Judge on the total, not the sum of the parts. If every channel reports a good cost per acquisition and blended CAC is bad, the channels are claiming the same customers twice.

What is different about running this in Thailand

A Thai startup's channel mix leans on chat more than a Western playbook expects. A large share of consumer and small-business purchases are negotiated in a chat thread, where questions about stock, price, delivery and payment are answered by a person before money changes hands. That has consequences for every channel above: your ads are not selling the product, they are buying a conversation, and the conversion rate that decides your CAC is largely determined by how fast and how well someone replies. Sales capacity, not media budget, is often the real constraint. A startup with two people answering chat cannot usefully scale a channel that produces more conversations than they can handle, and pouring more budget into it just increases the number of people who are ignored.

Search behaves differently in Thai as well. Thai script has no spaces between words, so keyword tools and search engines both have to guess where words break, and a phrase that looks like one keyword in English can fragment in Thai. Users switch scripts inside a single query, transliterate English terms inconsistently, and often search in English for technical or B2B categories while searching in Thai for consumer ones. Building a keyword set by translating an English list produces a set that no Thai person would type. Content has the same problem in a subtler form, because a translated article reads as translated, and trust is the thing you have least of as a new company.

Marketplace and social commerce behaviour also change the shape of the funnel. Buyers compare on marketplace platforms and in social feeds before they ever reach a website, so a startup with no marketplace presence may be invisible at the exact moment a decision gets made, while a startup that sells only on marketplaces owns no customer data and cannot build the email channel described above. Neither position is automatically right, but choosing one by accident is expensive.

FAQ on startup channels and metrics

How many channels should a startup run at the same time?

Two channels at a level that can actually produce conversions, plus one slow compounding channel started early, is the usual answer for an early-stage team. Splitting a small budget across five channels gives you five results too small to interpret, and a signal you cannot read costs the same as one you can. Add a channel only when an existing one has either proven itself or been killed.

What is a normal marketing budget for a startup in Thailand?

There is no honest public benchmark for this, and any percentage-of-revenue rule you find is imported from another market and another category. Work it out from the other direction: calculate the contribution margin a customer produces per month, decide how many months of payback your cash can survive, and the product is the most you can pay for a customer. Multiply that by the number of customers you need this quarter and you have a budget derived from your own business rather than someone else's.

Should a startup do SEO or paid ads first?

Run paid first if you need to know quickly whether anyone wants what you sell, and start SEO at the same time if your runway allows, because it takes the longest to pay off. Paid search answers the demand question in weeks and gives you keyword and message data that makes the SEO work sharper later. The one situation where SEO comes first is a category where the auction is expensive and the buyer researches for a long time before purchase.

How do you know when to kill a channel?

Kill it when it has produced enough conversions for its cost per customer to be distinguishable from random variation, and at that cost the payback period is longer than the cash you have available to fund it. Both halves are required. Stopping before there is enough volume to read is guessing, and keeping a channel alive on a token budget produces neither customers nor information.

Why does payback period matter more than the LTV to CAC ratio?

Because a ratio says nothing about when the money arrives, and a cash-constrained company is killed by timing rather than by lifetime value. A business can show a strong LTV to CAC ratio while the cash takes years to return, which is unusable if your runway is measured in months. Use the ratio to describe whether the business model works, and payback period to decide how fast you are allowed to spend.

Where to take this next

The order of work is the same for most early-stage teams: instrument the one conversion that means money, pick a short channel list you can actually fund, write the kill rule before you launch, and calculate CAC, LTV on contribution margin and payback period from your own figures rather than from a benchmark somebody published. Everything else is a variation on that.

Relevant Audience works with startups and small teams in Bangkok across the channels described here, from search and paid media to content, email and creator work, and on the measurement setup that makes the numbers above trustworthy. If you would rather talk it through against your own product, margin and runway than work from a general article, start with the startup digital marketing page and get in touch.