Deciding how much to spend on marketing is step one. Deciding where it goes is where most service businesses get it wrong. A simple allocation framework: foundation, proven, promising, and a small slice for tests.
You've settled on a number. Six percent of revenue, or four thousand a month, or whatever the budget guide helped you land on. Now comes the question that actually decides whether it works: where does it go?
Most service businesses answer by accident. The ad rep who called gets some. The website person gets some. Whatever the owner read about last month gets some. A year later there's a bill for everything and a result from nothing in particular.
Here's a way to allocate on purpose.
First, Count Hours as Money
Your marketing budget isn't just the ad spend. It's the ad spend plus the hours you and your team put in, valued at what those hours are worth. An owner who bills $150 an hour and spends five hours a week on marketing is spending roughly $3,000 a month before the first ad runs.
Add that in. It changes the picture, and it usually reveals that the "free" channels are the most expensive ones you have.
Free toolMarketing budget calculatorEnter revenue and a goal and see the four buckets in your own numbers.
Run your numbersThe Four Buckets
Split the total into four buckets. The percentages below are a starting point for a service business with something already working; adjust once your own numbers come in.
Foundation (about 20%)
The things every other channel depends on. A website that turns visitors into conversations, a finished Google Business Profile, a review system, basic tracking so you know what's working, an email list with a way to capture addresses. Most of this is one-time setup plus light maintenance. If it isn't done, it comes first and it takes a bigger slice until it is, because paying for traffic to a broken foundation is the most reliable way to waste money in marketing.
Proven (about 50%)
Whatever you can already trace customers to. If your attribution says referrals and Google bring the customers, half the budget goes to making those two better: a referral system, a properly run search campaign, service pages that rank. The proven bucket is where you spend to grow, not to experiment. It should be boring, because boring is what predictable looks like.
Promising (about 20%)
Channels that fit your customer and your economics but you haven't fully worked yet. For most service businesses this is where email, content and a retargeting campaign live in year one. They get real money and a real time frame (two quarters), with a defined result that would move them into the proven bucket or out of the budget.
Tests (about 10%)
Small, time-boxed bets on things that might work. A geofencing test around one event. A local sponsorship. A new platform your customers seem to be on. Ten percent, ninety days, a written definition of success before you start, and no exceptions to the time box. Most tests fail. That's the point of keeping them small.
What Goes Where, for a Typical Service Business
A rough map, assuming the foundation is in place:
- Search ads (Google, including Local Services Ads): Proven for most businesses with urgent or high-intent services. WordStream's 2026 benchmarks put the average search cost per lead at $66.69 across industries and $90.92 for home services, which is a bargain against a $1,500 job and a disaster against a $60 one. Let the math decide the size.
- SEO and content: Promising in year one, proven by year two if you keep at it. Mostly hours, not dollars.
- Email: Promising until the list is a few hundred people, then proven. Cheap in dollars, needs a monthly hour.
- Meta ads: Promising as retargeting and seasonal awareness. Rarely proven as cold lead generation for a service business.
- Referrals: Proven almost everywhere and funded almost nowhere. Give the referral system a real line: the thank-you gifts, the printed cards, the follow-up sequence.
- Everything else (sponsorships, print, radio, geofencing, a new platform): Tests, until they prove otherwise.
The Rules That Keep It Honest
- Nothing moves to "proven" without a traced customer. Impressions, likes and "brand awareness" don't count. A customer whose source you recorded does.
- Nothing stays in "promising" for more than two quarters. It graduates or it goes.
- The test bucket doesn't grow. The urge to chase the new thing is the urge to spend the test budget three times over. Ten percent, and the rest waits its turn.
- Rebalance quarterly, not weekly. Channels need time to show what they can do. Quarterly is often enough to act and rare enough to be fair.
"A budget with no allocation is a list of vendors. A budget with an allocation is a strategy."
A Worked Example
A residential contractor doing $1.2M a year sets aside $6,000 a month, hours included. Foundation is done, so it drops to maintenance: $800 for the profile, reviews, tracking and site upkeep. Proven is search ads and referrals: $2,400 on Google (Local Services first, then a tight search campaign) and $600 on the referral program. Promising is email and content: $1,200 mostly in a part-timer's hours building a monthly answer and a seasonal reminder sequence. Tests: $600 on a spring geofence around a home show and a single neighborhood mailer.
At the end of the quarter, the attribution table says the mailer produced two jobs and the geofence produced nothing traceable. The mailer moves to promising with a bigger slice next quarter. The geofence goes. Search ads produced eleven jobs at a cost per customer the owner would pay all day; the proven bucket grows. Boring, predictable, working.
When You Can't Afford Four Buckets
If the total budget is small, collapse it to two: foundation and proven. Finish the profile, the review system and the tracking, then put every remaining dollar and hour into the one channel you can already trace customers to. Add buckets as revenue allows. A small budget spread across six channels does nothing. The same budget on one channel, done properly, works.
Allocation isn't complicated. It's the discipline of spending where the evidence is and testing where it isn't, and of writing it down so next quarter's decision is better than this one's.