How to split budget across Meta, Google and LinkedIn

Ask ten media buyers how to split a budget across Meta, Google and LinkedIn and nine will tell you "it depends." Then they'll allocate by habit anyway. The split ends up wherever last year's split was, plus or minus whatever channel the loudest stakeholder likes. That's a lot of money to allocate by habit: paid media is the single biggest line in the marketing budget, and Gartner's 2025 data puts it at 30.6% of total marketing spend.⁴ Nothing else you decide this quarter moves as much money as this one ratio.

To be fair, "it depends" is the right answer. It just depends on three knowable variables, not on vibes. Once you've answered them, the starting split is close to mechanical, and the real work becomes rebalancing it every month without breaking what already works.

The three variables that decide it

Every defensible budget split is an answer to three questions. If you can't answer them, no benchmark will save you. If you can, most of the allocation falls out on its own.

1. How much demand already exists? Google search captures intent that's already formed: someone typed your category into a search box. Meta and LinkedIn mostly create demand in people who weren't looking. Capture is cheaper, faster to convert and finite, because there are only so many searches per month for what you sell. The size of that pool relative to your growth target is the single biggest driver of the search-vs-social ratio.

2. What is a customer worth? Deal economics decide which channels you're even allowed to use. B2B cost per lead across paid channels averaged $84 in 2025, roughly $70 on Google and $110 on LinkedIn.¹ A $110 CPL is a rounding error on a $40k ACV and a death sentence on a $500 one. Work backwards from your allowable CAC and some channels eliminate themselves before you've spent a dollar.

3. How big is the audience? A precise audience of 30,000 decision-makers favors LinkedIn's targeting over Meta's scale. A product everyone with a credit card could buy favors the platform with the cheapest attention. Audience size also sets your ceiling: a channel can be your best performer and still only absorb so much budget before saturating.

Answer those three and the rest of this article is just arithmetic plus the discipline to act on it.

Where the first dollar goes

Order of operations matters more than percentages. The first dollars go to capturing demand that already exists, because it's the highest-probability money you will ever spend: brand search, then high-intent non-brand search, then shopping if you have a catalog. Only once capture is funded to the point of diminishing returns do you start paying to create demand on social.

The platform pricing makes the logic concrete:

Google's average search CPC sits around $4.22 across industries,² Meta's global median CPC ran about $1.11 from January 2025 through January 2026,³ and LinkedIn clicks cost $5–12, climbing past $15 for C-suite targeting.⁵ Don't read those as quality rankings. They're prices for three different products: Google sells intent, Meta sells reach, LinkedIn sells precision. The mistake is paying LinkedIn prices for reach, or expecting Meta prices to deliver intent.

Starting splits by business type

With the three variables answered, most businesses land near one of four archetypes. These are starting points; your own data should pull you away from them over time:

The high-ACV B2B split is the one I've seen teams resist most, because LinkedIn's sticker prices feel wrong next to Meta's. Drawn out, it looks like this:

It holds up because lead quality diverges harder than lead cost: LinkedIn leads convert from MQL to SQL at roughly 20–30%, versus 8–15% for Meta in B2B accounts.⁵ Price the lead at the opportunity stage instead of the form fill and the "expensive" channel often turns out to be the cheap one. Whatever archetype you start from, sanity-check your costs against vertical benchmarks before concluding a channel "doesn't work". Half the time the channel is fine and the account is mis-built.

The 70-20-10 layer

The channel split answers where the money goes; 70-20-10 answers how confidently. The framework, which Coca-Cola ran as "NOW / NEXT / NEW", puts 70% of budget behind proven channel-and-tactic combinations, 20% behind bets showing early promise, and 10% behind genuine experiments.⁶ McKinsey research cited alongside it found companies with this kind of balanced allocation delivered 2.7× higher shareholder returns over ten years than those with static budgets.

Two caveats from running this on real accounts. First, the ratios should flex with maturity: a startup with no proven channel yet is better off at 50-30-20, while a mature account defending efficiency can run 80-15-5. Second, and this is where most teams fail, the framework only works if money actually moves between buckets. A "10% experimental budget" that funds the same TikTok test three quarters running isn't an experiment anymore, it's a subscription.

Decide upfront what an experiment has to show, and by when, to earn real budget. If nobody wrote that down, the 10% bucket becomes a place where pet projects go to live quietly.

Set a fixed monthly review where experiments either graduate to the 20% bucket with real budget, or get cut. The discipline matters more than the exact percentages.

When to add a channel (and when you're just bored)

New channels get added for one of two reasons: the data says the incumbent is saturating, or someone saw a conference talk. Only the first is a reason to spend money. The trap is that average ROAS hides saturation: a channel can report a blended 4× while the last dollar in returns well under 1×, because early spend on cheap, high-intent audiences subsidizes expensive marginal reach.⁷ I've watched teams scale their "best channel" on blended numbers and buy the flat part of the curve for months.

Concrete triggers that justify opening a new channel:

• Marginal cost per result has crossed your allowable CAC. Compare your last budget increase's results to the account average. If the increment converted at 2× the blended CPA, the next increment will be worse.
• Audience exhaustion on prospecting. Frequency creeping up while reach plateaus means you're paying to show ads to the same people again instead of finding new ones.
• CPM inflation without an auction-wide cause. If your CPMs rise quarter over quarter while category benchmarks hold flat, the platform is straining to find you net-new attention.
• A structurally unreachable segment. Some buyers simply aren't addressable on the incumbent — no Meta budget reaches a CFO the way a LinkedIn job-title filter does.

When a trigger fires, fund the new channel from the 20% bucket at a size that can actually produce a verdict. That means enough spend for roughly 50 conversions per month; below that, the platform's delivery system never exits learning and you'll mistake an underfed account for a failed channel.

Rebalancing without thrash

The difference between rebalancing and thrashing is step size and cadence. Reallocate weekly on last week's CPA and you're chasing noise, resetting learning phases, and paying the re-learning tax on every move. Never reallocate and you're donating margin to a saturated channel. The rules for the middle path fit in one table:

Two refinements make this much easier to run. First, move budget within a channel before moving it between channels; shifting spend from a fatigued campaign to a fresh one is cheaper than crossing platforms. Second, write the trigger down before the month starts ("if LinkedIn's marginal CPL exceeds $140 for the full month, shift 10% to Google") so the decision is already made and the monthly meeting is just execution. This is the kind of policy worth encoding once in your campaign strategy rather than re-litigating every month, and it only works if you can see all three platforms' pacing and marginal costs in one view instead of three tabs and a spreadsheet.

Reading cross-channel signals

Budget allocation treats channels as competitors for the same dollar, but they're also each other's best research departments, and almost nobody takes advantage of that.

• Winning hooks travel well. A Meta ad whose hook doubles the account's CTR is a tested message with thousands of impressions behind it. Rewriting it for LinkedIn's context (or as a search headline) is a much better-odds test than a fresh brief.
• Search terms tell you what the market wants to hear. Rising query themes in your search terms report are next month's social angle, validated by people who typed it themselves.
• Let the cheap channel do the follow-up. LinkedIn is costly for repeated touches. Audiences that engaged there can be re-reached on Meta at a fraction of the CPM, so let the premium channel find them and the cheap one nurture them.
• One channel's fatigue predicts the next. A concept dying on Meta after three weeks will usually die on other feeds too, so schedule the refresh before the second platform proves it.

Most teams lose these signals to platform silos: the Meta buyer and the LinkedIn buyer are different people with different dashboards. You can close the loop manually with a standing monthly ritual where each channel owner brings one learning the others can use. Closing it automatically is what cross-channel learning in Adside does; it treats every platform's results as test hypotheses for the others.

That's the full system: three variables to set the split, a grid to start from, 70-20-10 to keep it honest, triggers to expand it, step-size rules to rebalance it, and cross-channel signals to learn from it. None of it requires software, just a calendar and the discipline to judge channels on their marginal dollar instead of their reputation.

Frequently asked questions

Sources

1. B2B cost-per-lead averages by channel, 2025 — Sopro, B2B Cost Per Lead Benchmarks
2. Cross-industry average Google Search CPC — PPC Chief, PPC Benchmarks by Industry
3. Global median Meta CPC, Jan 2025–Jan 2026 — AdAmigo, Meta Ads CPM and CPC Benchmarks by Country
4. Paid media share of marketing budgets (Gartner 2025) — Single Grain, 2025 Marketing Budget Insights
5. LinkedIn CPC, CPL and lead-quality comparisons — Stackmatix, LinkedIn Ads Cost: Complete CPC Pricing Guide
6. The 70-20-10 framework and McKinsey findings — Growth Method, The 70-20-10 Rule
7. Average vs. marginal ROAS and saturation — mbuzz, How a 4x ROAS Channel Hides 0.6x Marginal ROAS