How to allocate marketing budget by what each channel's next dollar will buy.

A dental group with four locations had a strong year on Google Ads and planned to double the budget for the next one, and that plan is a good test of how to allocate marketing budget.

Before the budget meeting, I ran the numbers in a forecasting spreadsheet. The ads were already showing on about 85% of the searches that mattered, which is what impression share measures, and more money would mostly buy pricier clicks on broader searches. The forecast showed three options:

1. Current spend brought in about 60 leads a month, at about $95 each.
2. Double the spend reached about 64 leads, at about $178 each.
3. A 25% increase plus a test on Microsoft Advertising landed near 70 leads, at about $102 each.

The group went with the third option and saved the difference.

A bigger budget helps when there are good searches left to buy.

Why the next dollar matters more than the total.

If your company follows Gartner's numbers, next year's total will look a lot like this year's. Gartner, which sells advisory services to CMOs, surveyed 401 marketing leaders in North America and Europe for its 2026 CMO Spend Survey (January to March 2026, mostly at companies above $1 billion in revenue). Budgets held at 7.8% of revenue, against 7.7% in 2025, and 56% of CMOs said their budget isn't enough for their 2026 strategy.

The money for a channel that's working usually comes out of one that isn't.

Channels also don't scale in a straight line. Google researchers (Jin and colleagues, 2017) wrote that advertising "usually has lag effects and diminishing returns," which is why channels get compared on their next dollar, over a window long enough for the lag.

A line commonly attributed to John Wanamaker, and to Lord Leverhulme, says half the money spent on advertising is wasted and no one knows which half. Quote Investigator traced it to a 1919 speech and found no record of Wanamaker saying it. Measurement answers it, not more money, and the wasted half is often a channel full of cheap leads that never become good clients.

How to allocate marketing budget in seven steps.

Build one sheet ahead of the budget meeting, named with the year, for example "2026-channel-review".

1. List every channel that brought in clients this year.

Include referrals, outbound, events, partners and email to existing clients.

At a larger company, map the CRM's source values and finance's cost centers into eight to ten agreed channels.

2. Add up the full cost of each channel.

Add the tools, agency or contractor fees, event fees and travel, and people's time at their loaded hourly cost to the ad spend.

3. Count the clients each channel brought in.

Count clients, not leads, from two sources side by side: the CRM's first touch (Original Traffic Source in HubSpot, the source field on the lead or contact in Salesforce, a column in a spreadsheet with no CRM) and the answer to "How did you hear about us?" When they disagree, keep both, and the gap shows which channels your tracking can't see.

At a larger company, count closed deals by the source recorded on the opportunity or deal, and count each account once.

4. Mark the clients you'd want more of.

Write the criteria before you look at which channel each client came from: fit (size, sector, the problem they bought for) and how much they're likely to grow, not only what they've spent to date. Mark each one yes or no in a "best-fit" column, or at a larger company, use the account tiers sales already works from.

5. Divide to get the cost per best-fit client.

Full channel cost divided by its best-fit clients. A channel with none this year gets "none yet" and a note on how long its clients take to arrive. With a long sales cycle, match each year's spend to the clients it produced.

6. Rank the channels, and check headroom before moving money.

The top channel earns the first look if it can take more at the same cost. For paid search, three Google Ads columns answer it:

• Search impr. share, the share of eligible impressions your ads received.
• Search lost IS (budget), the share of the time your ads didn't show because the budget ran out, reported at the campaign level only.
• Search lost IS (rank), the share lost to Ad Rank.

If little is lost to budget, new money sits unspent until bids, targets or keywords loosen, and then it buys broader, pricier searches. When the loss to rank tops 50%, "increasing budget is unlikely to drive incremental value," Navah Hopkins wrote in MarTech (June 2026), and the work goes into ads, landing pages and bids. With target CPA or target ROAS bidding, a target set too tight can cut traffic, Google notes, and more budget won't fix that.

Performance Planner forecasts from simulated auctions over the last 7 to 10 days, adjusted for seasonality, and Google's simulators use the last 7 days and may show nothing for a campaign that hit its budget that week. Both windows are short for a year's plan, and your 90-day numbers check them.

7. Hold 10 to 15% of the budget to test one new channel.

Test one channel at a time, with the result defined before it starts, because with two at once you usually can't tell which did the work. Testing one channel at a time covers the plan.

SEO, events and brand work can look expensive in a single year. Les Binet and Peter Field's 2019 B2B report for the LinkedIn B2B Institute (page 36) cites their earlier, mostly consumer analysis showing long-term effects on sales "only start to dominate short-term effects after six months," and a one-year review will usually rank the slow channels last. Give SEO and events a two-year view before cutting them.

If referrals rank first, the budget goes to what produces them, like the client experience or a partner program (past clients and your network).

A worked example with numbers.

In the dental forecast, the second half of a doubled budget would have bought about 4 more leads, from about 60 a month to about 64. Example numbers: at $6,000 a month for 60 leads ($100 each), reaching 64 at $12,000 means paying $6,000 for 4 added leads, or $1,500 each, against a $100 average. Bring that marginal cost to the budget meeting, not the averages.

Example numbers: a B2B services firm with a $189,000 budget and 22 best-fit clients this year.

Referrals rank first, and the extra $6,000 funds a partner program. Google Ads loses only 4% to budget and stays flat, and outbound is capped by one salesperson's hours. The weakest event, in its second year with no best-fit client, is dropped, LinkedIn Ads keeps $10,000 for a stated reason (reach with the outbound accounts), and $20,000, about 11%, goes to one new channel. The total stays at $189,000.

Keep the budget moving after it's set.

A campaign can underperform for months while its share stays where the annual plan set it. A monthly ranking fixes that:

1. Send qualified leads back to the ad platforms. HubSpot's ad conversion events sync lifecycle stages to Google Ads, and Salesforce or a plain Google Sheet connect through Google's Data Manager with enhanced conversions for leads, the upgrade to offline conversion import.
2. Rank every campaign on the first business day of the month by cost per qualified lead over the last 90 days, in Data Studio (formerly Looker Studio) or a spreadsheet.
3. Move up to 10% of the budget from the bottom campaigns to top campaigns that still lose impressions to budget, with the total unchanged. Campaigns under 60 days old, or with too few qualified leads to compare, sit out.
4. Review any campaign at the bottom three months in a row. It's paused, rebuilt or kept for a stated reason.
5. Log every change in a shared sheet, for example "2027-budget-moves", with the reason next to it.

Once it's set up, the review takes about 20 minutes a month. If Microsoft Advertising runs from a scheduled Google Ads import, set the import options to leave budgets alone, or the next import can overwrite your move. At a larger company, agree on the 10% rule with finance up front, and the monthly moves won't need new approval.

Common mistakes, and the pushback a specialist will raise.

Ranking on the last 30 days. Google Ads reports conversions by the date of the click, and qualified-lead imports arrive weeks later, which drops slow-qualifying campaigns to the bottom before their results show up. Google's guidance is to judge bid strategy results over two conversion cycles or two weeks, whichever is longer, and for B2B lead quality 90 days is safer.

"Attributed isn't incremental." That's right, and eBay's field experiment (Blake, Nosko and Tadelis, Econometrica, 2015) shows the gap. It switched off non-brand search ads in about 30% of US metro areas and measured a return of −63%, where standard regression had suggested more than 1,400%. eBay is a consumer brand with heavy brand demand, and B2B search may do better. The step test below gives a first read on the next dollar, and at a larger company, switching off a region gives the causal one.

"First touch ignores everything in between." It does, which is why the self-reported column sits next to it. This is a ranking for a budget decision, not a credit split, and a one-page report keeps both in view.

A test to run before the budget meeting.

1. Pick the campaign with the lowest cost per qualified lead that still shows some impressions lost to budget.
2. Record its 90-day baseline (spend, qualified leads, cost per qualified lead) and, where you can, the same weeks last year to separate the budget from the season.
3. Raise its budget by about 15% and change nothing else.
4. Run it for four weeks or two conversion cycles, whichever is longer, then wait out your usual time from lead to qualified lead before counting.
5. Count qualified leads in the CRM from every source, not only the ones Google Ads credits, by click or form-fill date, and divide the added spend by the added qualified leads, against the baseline weekly rate.
6. As a rule of thumb, an added count under about three times the square root of the baseline count for those weeks is noise. Example numbers: on a baseline of 100, that bar is 30, and a 15% step usually adds 15 or fewer, which makes any result below a baseline of about 400 a direction to retest, not a verdict.
7. It passes if it clears that bar and the marginal cost per qualified lead is at or below what a qualified lead is worth to you (what a lead is worth has the formula). If it clears the bar at a higher cost, the next dollar goes elsewhere.