Strategy & Craft
How much advertising budget is the right amount
Percentage-of-revenue rules are widely quoted and mostly wrong. The answer comes from your own numbers in four steps.
2 min readMay 7, 2026
Strategy & Craft
Splitting an ad budget: the 70-20-10 rule
Seventy per cent on what works, twenty on scaling, ten on experiments. The rule is old, simple, and violated in almost every account.
Key points
- Without a fixed experiment share, nothing new ever gets tested.
- The 70 per cent must be earned by evidence, not by habit.
- Review the split quarterly, not annually.
Advertising budgets tend to ossify. What was booked last year gets booked again, because changing it requires an argument and keeping it does not.
The three tiers
70 per cent: proven. Channels and formats where you have evidence of effect, not merely attributed conversions. This is the load-bearing part of the plan.
20 per cent: scaling. Things that worked in a test and are being expanded. New regions, new audiences, higher spend in a channel that showed promise.
10 per cent: experiments. Deliberately uncertain. New channels, new formats, new messages. Most of this fails, and that is the point.
The ten per cent is not a luxury. It is the only mechanism by which next year's seventy per cent gets discovered.
Earning the seventy
The dangerous part is tier one, because habit disguises itself as evidence. A channel that has run for four years and is never questioned is not proven, it is entrenched.
Apply a simple test once a year: for each channel in tier one, what would happen if we switched it off for six weeks? If nobody can answer, it belongs in a switch-off test, not in the protected majority.
What counts as a real experiment
| Not an experiment | An experiment |
|---|---|
| A new creative in a running campaign | A channel you have never used |
| A slightly different headline | A different audience definition |
| Raising the budget | A different offer structure |
| A new agency doing the same thing | A different medium entirely |
Experiments need a written hypothesis, a duration, a budget cap and a success threshold agreed before the start. Without those four they become opinions.
A worked split
On a £8,000 monthly budget for a regional service business:
- £5,600 proven: search advertising on service terms, local directory listings, the newsletter.
- £1,600 scaling: extending search to three neighbouring districts that tested well.
- £800 experiments: this quarter, local podcast sponsorship; next quarter, addressed mail to a bought business list.
Why quarterly review beats annual
An annual review locks a mistake in for twelve months. Quarterly review lets a failed experiment die in three months and a successful one move up a tier while it is still growing.
Set four fixed dates in the calendar. The review takes 90 minutes and consists of three questions per channel: what did it cost, what did it produce, and what would happen without it.
Frequently asked questions
What if nothing is proven yet?
Then start at 40-30-30 and shift towards 70-20-10 as evidence accumulates over two or three quarters.
Is ten per cent enough for experiments?
On a £10,000 monthly budget, £1,000 tests one new channel properly. On £2,000 it is too little; test one thing per quarter instead.
More from Strategy & Craft
Keep reading
Strategy & Craft
The brief agencies actually need
A bad brief produces bad work at full price, and the client pays twice. Eight questions, two pages, three hours of your time.
2 min readMay 5, 2026
Strategy & Craft
Positioning before the first ad runs
Advertising amplifies whatever you already are. If that is indistinguishable from four competitors, the budget amplifies the confusion.
2 min readMay 3, 2026