Contents＋
01 / THE PERCENTAGE TRAP
7.8% can be true and still be useless
Gartner’s 2026 CMO survey places average marketing budgets at roughly 7.8% of company revenue. That is a global management benchmark, not a menu for every Hong Kong SME.
Two companies can each make HK$10 million. One sells high-margin advisory work; the other moves low-margin products. One has spare delivery capacity; the other is booked for three months. The same percentage can starve the first of growth and buy the second customers it cannot serve.
The danger of a percentage is that it looks objective while hiding lifecycle and margin. A young professional-services firm may have unstable revenue but high value per client. A mature ecommerce company may have large revenue while fulfilment, returns and platform fees consume most gross profit. The same ruler misbudgets both.
Ask how soon the spend must create recoverable cash. When a company is already facing higher costs and lower profit, a budget cannot only describe desired reach. It must survive the worst month without squeezing payroll, inventory or delivery.
02 / WORK BACKWARDS
Calculate what a customer is worth before deciding what to spend
Start with an imperfect but honest equation:
Suppose a project earns HK$60,000 and leaves HK$30,000 after direct delivery cost. If the firm allocates 25% of that gross profit to acquisition, it can tolerate about HK$7,500 per new customer. If one in four qualified enquiries closes, the ceiling is about HK$1,875 per qualified enquiry.
The point is not to spend up to the limit. It is to replace “Are ads expensive?” with “Is this kind of customer economically worth acquiring?”
Split close rate as well. If gross profit per client is HK$30,000 and the company allocates 25% to acquisition, the ceiling is about HK$7,500 per client. At one close for every four qualified enquiries, the ceiling is HK 02 / WORK BACKWARDS Start with an imperfect but honest equation: Suppose a project earns HK$60,000 and leaves HK$30,000 after direct delivery cost. If the firm allocates 25% of that gross profit to acquisition, it can tolerate about HK$7,500 per new customer. If one in four qualified enquiries closes, the ceiling is about HK$1,875 per qualified enquiry. The point is not to spend up to the limit. It is to replace “Are ads expensive?” with “Is this kind of customer economically worth acquiring?”Calculate what a customer is worth before deciding what to spend
Then ask about payback. Does the client pay in month one, or renew after six months? A cash-constrained company may theoretically tolerate a cost it cannot practically pre-fund. Separate gross profit from collection timing so future revenue is not mistaken for today’s cash.
03 / THE HIDDEN CEILING
The real budget ceiling is often operational capacity
Many SMEs do not completely lack demand. They have an owner writing every proposal, three approval layers, no developer available and WhatsApp replies arriving two days late.
In QBE’s 2025 survey, almost 60% of Hong Kong SME respondents faced rising costs and reduced profitability; roughly half reported cash-flow and talent pressures. That makes an overlooked fact more urgent: enquiries nobody can handle become operating pressure, not growth.
If delivery can absorb three additional clients a month, optimise for three suitable clients—not maximum reach.
Use a simple capacity check: monthly qualified enquiries the team can handle equals available sales hours divided by hours required per enquiry. If the owner has ten hours a week and a complex B2B enquiry takes two hours, the monthly ceiling is roughly twenty. More reach only makes response time longer.
Talent shortages lower that ceiling further. QBE’s survey found that about half of Hong Kong SMEs were troubled by cash flow and talent issues. When resources are tight, part of the budget may need to buy triage, process and automation—not simply more people at the door.
04 / FOUR WALLETS
Media is only the easiest cost to see
| Budget layer | What it buys | Common omission |
|---|---|---|
| Reach | Media, distribution, events, partnerships | Creative fatigue and testing |
| Reception | Site, landing pages, cases, speed | Development and approval time |
| Judgement | Analytics, CRM, tracking, experiments | Data cleaning and offline updates |
| Execution | People, expertise, follow-up and sales | Real owner and team hours |
HKPC’s 2025 AI-readiness research similarly identifies skills, privacy and integration with existing systems as implementation barriers. Buying a tool is not transformation. Buying media is not completed marketing.
The four wallets also have a sequence: the receiving surface usually comes before reach. If a service page cannot explain who it is for, how delivery works and when someone replies, extra advertising repeats the same unresolved question. One site fix, by contrast, can improve organic, paid and referral traffic at once.
AI tools carry a fifth, hidden wallet: data preparation, permissions, privacy review, human checking and staff learning time. HKPC’s research identifies talent, data privacy/security and integration with existing systems as implementation barriers. A monthly subscription does not make those costs disappear.
05 / THREE STARTING MODES
Fund one complete learning loop, not an arbitrary percentage
Validation
One audience, one expensive problem and one primary channel. The objective is message and lead-quality evidence, not coverage.
Growth
Can support media, site, content and measurement together, while keeping most resources on one battlefield.
System
For firms with a stable sales process, sufficient margin and delivery capacity; the constraint becomes experiment speed and coordination.
These are decision modes, not market quotations or performance promises. A smaller sum can work if scope shrinks with it. The smaller the budget, the less credible an “all channels” plan becomes.
The validation band is defined by a completed learning loop, not a number: a clear hypothesis, trackable contact, a qualified definition and a stop line. A growth band must prove sales and delivery can absorb new demand, or the larger budget only creates a longer queue. A system band also needs a data owner; otherwise cross-channel reporting becomes contradictory arithmetic.
These ranges are not a market price list. If a company can invest HK$5,000, narrow the scope to one service, one area and one learning question. If it can invest HK 05 / THREE STARTING MODES One audience, one expensive problem and one primary channel. The objective is message and lead-quality evidence, not coverage. Can support media, site, content and measurement together, while keeping most resources on one battlefield. For firms with a stable sales process, sufficient margin and delivery capacity; the constraint becomes experiment speed and coordination. These are decision modes, not market quotations or performance promises. A smaller sum can work if scope shrinks with it. The smaller the budget, the less credible an “all channels” plan becomes.Fund one complete learning loop, not an arbitrary percentage
Validation
Growth
System
06 / THE STOP RULE
A mature budget states when to stop before it starts
Write three thresholds in advance: the spend after which no signal means stop; the evidence required to increase investment; and the maximum qualified enquiries operations can handle each month.
A failed test is not the greatest danger. The greater danger is a test with no decision threshold, renewed every month with “give it a little longer.”
A budget should not prove that marketing activity occurred. It should purchase a testable business assumption and make the next dollar smarter than the previous one.
At the monthly review, split spend into three columns: attention purchased, market evidence purchased and revenue produced. The first grows easily. The second appears only when quality is labelled. The third is usually delayed. Keeping them separate stops reach replacing learning and learning replacing cash.
A mature stop rule also includes “do not scale.” If qualified enquiries already exceed delivery capacity, improve screening and fulfilment before buying more demand. A budget is an option on a hypothesis, not a badge proving that the company is active.
Sources
Sources
- Gartner 2026 CMO Spend Survey
- QBE Hong Kong SME Survey 2025
- HKPC AI Readiness in Workplace Survey 2025
- Google Analytics: key events and conversions
Figures describe cited samples or case studies and do not guarantee the same result for every Hong Kong business.