How to Allocate Your Ad Budget Using Data (Not Gut Feel)

Budget allocation comes down to one concept (diminishing returns), one metric (marginal ROAS), and a step-by-step reallocation process. A worked example with simulated data comes first, then the method.

Worked Example (Simulated): Where Does Added Spend Reach Break-Even?

Question: At what weekly spend does added ad spend fall to the break-even line in this made-up example?

Inputs: Simulated data, seed 37. The script builds weekly spend and revenue for one channel, in thousands of dollars. The revenue curve (90 x (1 - exp(-spend / 6))) and the 40% gross margin are set in the script, not measured. The seed sets the random noise (standard deviation 0.7) so the run can be repeated.

Method: Average ROAS is revenue divided by spend. Marginal ROAS is the change in revenue divided by the change in spend. Break-even ROAS is one divided by gross margin.

import numpy as np

# Simulated weekly data, in thousands of dollars. The seed makes the noise
# repeatable. The revenue curve and the 40% margin are set here, not measured.
seed = 37
rng = np.random.default_rng(seed)
spend = np.array([4, 6, 8, 10, 12, 14], dtype=float)
base_revenue = 90 * (1 - np.exp(-spend / 6))
noise = rng.normal(0, 0.7, size=spend.size)
revenue = base_revenue + noise
roas = revenue / spend
marginal_roas = np.diff(revenue) / np.diff(spend)
break_even_roas = 1 / 0.40
# Where the built-in curve's slope, 15 * exp(-spend / 6), equals break-even.
cross = 6 * np.log(15 / break_even_roas)

print(f"Seed: {seed}")
print("Spend ($000): " + ", ".join(f"{x:.0f}" for x in spend))
print("Revenue ($000): " + ", ".join(f"{x:.2f}" for x in revenue))
print("Average ROAS: " + ", ".join(f"{x:.2f}" for x in roas))
print("Marginal ROAS between spend points: "
      + ", ".join(f"{x:.2f}" for x in marginal_roas))
print("Assumed gross margin: 40%")
print(f"Break-even ROAS: {break_even_roas:.2f}")
for i, value in enumerate(marginal_roas):
    low, high = spend[i], spend[i + 1]
    status = "above" if value >= break_even_roas else "below"
    print(f"${low:.0f}k to ${high:.0f}k spend: marginal ROAS {value:.2f}, "
          f"{status} break-even")
print(f"Built-in curve crosses break-even at about ${cross:.2f}k spend")

Printed output:

Seed: 37
Spend ($000): 4, 6, 8, 10, 12, 14
Revenue ($000): 44.42, 56.57, 65.36, 73.46, 78.49, 81.64
Average ROAS: 11.11, 9.43, 8.17, 7.35, 6.54, 5.83
Marginal ROAS between spend points: 6.07, 4.39, 4.05, 2.51, 1.58
Assumed gross margin: 40%
Break-even ROAS: 2.50
$4k to $6k spend: marginal ROAS 6.07, above break-even
$6k to $8k spend: marginal ROAS 4.39, above break-even
$8k to $10k spend: marginal ROAS 4.05, above break-even
$10k to $12k spend: marginal ROAS 2.51, above break-even
$12k to $14k spend: marginal ROAS 1.58, below break-even
Built-in curve crosses break-even at about $10.75k spend

Result: Break-even ROAS is 2.50. The $10k to $12k step (2.51) sits on the break-even line. Noise in the weekly revenue can move each step by about half a point, so the crossing point is not firm. The script's built-in curve crosses at about $10.75k. The $12k to $14k step (1.58) is below the line.

Limit: The data are simulated and this is one run. A real spend and revenue pattern can reflect season, offers, or other changes, so a curve like this alone does not show that spend caused the revenue change.

The Problem With Even-Split and Last-Year-Plus Budgets

Most ad budget allocation follows one of two patterns:

To see which of your own marketing channels drive sales, try the free media mix model tool.

• Even split: Divide total budget equally across channels. "We have $100K and five channels, so $20K each."
• Last year plus: Take last year's allocation, add 10-15%, maybe shift a few thousand dollars around based on gut feel.

To see which of your own marketing channels drive sales, try the free media mix model tool.

Both approaches ignore the fundamental reality of advertising: every channel has a point of diminishing returns. The first $5,000 you spend on Google Ads can target your highest-intent keywords and generate strong returns. The next $5,000 may target slightly less valuable keywords. At higher spend you may be bidding on broad-match terms with weaker purchase intent, and your cost per acquisition can climb.

Consider a made-up example. A DTC brand allocates $60,000/month evenly across three channels. The figures in the next two tables are invented to show the arithmetic:

Google Ads looks like the best channel on average ROAS (4.2x). But its marginal ROAS is 1.1x: the last $5,000 brings back about $5,500 in revenue, which may not cover costs after margin. TikTok's marginal ROAS is 0.6x: the last $5,000 spent there brings back about $3,000 in revenue, which is $2,000 less revenue than was spent, before any margin is counted. Meta, despite a lower average ROAS, returns 2.9x on marginal spend in this made-up example. In this example, that is where the next dollar would go.

Diminishing Returns: The Concept That Changes Everything

Returns from a channel often flatten as spend grows, and the curve may be S-shaped or simply concave. At low spend levels, returns increase rapidly as you reach the most receptive audience. As spend grows, you exhaust the best opportunities and start reaching less qualified prospects. Eventually, more spend produces almost no incremental return.

If you have spent different amounts on a channel over recent months (due to seasonal budgets, testing, or gradual scaling), you may already have the data points to plot a response curve.

A Worked Example: Google Ads Response Curve

A made-up e-commerce retailer tracked weekly Google Ads spend vs. revenue over 16 weeks. Spend varied from $8,000 to $22,000/week due to seasonal adjustments. These figures are invented to show the arithmetic. Here is the pattern:

Average ROAS at $22,000/week is still 3.2x, which looks healthy. But marginal ROAS on the last $2,000 is 0.2x. That $2,000 generates only $400 in revenue. Assuming the 50% gross margin used in the break-even example in Step 3 below, break-even ROAS is 2.0x. The table shows marginal ROAS of 2.0x from $12,000 to $14,000 (at break-even), then 1.25x from $14,000 to $16,000. The table supports spending up to about $14,000/week before other costs.

How to Identify Channel Saturation

You need three things to identify saturation: spend variation, outcome data, and a simple analysis framework.

Step 1: Gather Historical Spend and Outcome Data

For each channel, pull as many weeks of data as you can, with these columns:

• Week: The time period (weekly granularity works best)
• Channel spend: Total ad spend for that channel that week
• Revenue (or conversions): Attributed revenue or conversion count
• External factors: Note any promotions, holidays, or anomalies

The critical requirement is spend variation. If you spent exactly $15,000 on Google Ads every week, you have one data point, not one per week. You need weeks where spend was meaningfully higher and lower.

Step 2: Calculate Marginal Returns at Each Spend Level

Sort your data by spend level (low to high). For each incremental spend band, calculate the additional revenue generated:

Marginal ROAS = (Revenue at Spend Level N) - (Revenue at Spend Level N-1)
                 ------------------------------------------------
                 (Spend at Level N) - (Spend at Level N-1)

When marginal ROAS drops below your break-even threshold, you have found the saturation point.

Step 3: Determine Your Break-Even ROAS

Your break-even ROAS depends on your gross margin. The worked example at the top used a 40% margin (break-even 2.50). Here, as an example, if your gross margin is 50%, you need at least 2x ROAS to cover cost of goods. If your margin is 33%, you need 3x. Add a buffer for overhead and target margin:

Break-even ROAS = 1 / Gross Margin %
Target ROAS     = Break-even ROAS + Margin Buffer

Example (50% gross margin):
Break-even = 1 / 0.50 = 2.0x
Target     = 2.0x + 0.5x buffer = 2.5x

Any channel spend where marginal ROAS falls below your target ROAS threshold should be reallocated.

Step-by-Step Data-Driven Budget Reallocation

Once you have response curves (or at least directional saturation estimates) for each channel, reallocation follows a clear process.

Step 1: Rank Channels by Marginal ROAS

At current spend levels, which channels still have the highest marginal returns? Order them from highest to lowest marginal ROAS.

Step 2: Identify Surplus and Deficit Channels

Surplus channels have marginal ROAS below your target threshold. They are over-funded. Deficit channels have marginal ROAS well above your threshold. They are underfunded and can absorb more spend productively.

Step 3: Shift Budget from Surplus to Deficit

Move budget in increments (10-20% shifts, not wholesale changes) from surplus channels to deficit channels. The goal: equalize marginal ROAS across all channels at a level above your break-even threshold.

Example: Reallocation in Practice

Using our made-up DTC brand example with $60,000/month total budget. The new budgets and marginal ROAS values are invented for illustration:

Same $60,000 total budget. In this example, $6,000 moves from Google Ads and $4,000 from TikTok to Meta, and the three channels end with marginal ROAS values closer together. These are not results from any account.

Step 4: Monitor and Iterate

Reallocation is not a one-time event. Response curves shift as audiences change, creative fatigues, and competitors adjust. Recalculate marginal ROAS monthly and make incremental adjustments quarterly.

Common Budget Allocation Mistakes

Beyond even-split budgeting, three patterns can hurt ad budget efficiency:

Mistake 1: Chasing Average ROAS Instead of Marginal ROAS

In a made-up case, a channel with 5x average ROAS gets more budget. The marginal ROAS on the increase is 0.8x. Total average ROAS drops to 4.2x but still looks "good." The team celebrates while losing money on the incremental spend. Average ROAS is a lagging indicator that hides marginal waste.

Mistake 2: Cutting Channels Based on Last-Click Attribution

Display advertising can show weak last-click ROAS. Teams cut it entirely. Later, search ROAS may fall if display was feeding the top of the funnel. Before cutting a channel, run an incrementality test to separate direct and assist value.

Mistake 3: Ignoring Saturation Thresholds When Scaling

For example, a channel works at $10K/month. The team scales to $40K/month expecting 4x the results and gets less because of diminishing returns. The response curve is concave, not linear. Always model expected marginal returns before committing to a scale-up.

Working on this for your business? See the marketing analytics tools.

Building Your Own Response Curves

If you want to build response curves in-house, here is the practical approach that works without a data science team.

The Spreadsheet Method (Good Enough for Most Businesses)

1. Export weekly data: Pull as many weeks as you can of channel_name, spend, and revenue from your ad platforms
2. Create scatter plots: One per channel, spend on X-axis, revenue on Y-axis
3. Add a trendline: Use a logarithmic or polynomial trendline (not linear) in Excel/Sheets
4. Read the curve: Where the trendline flattens is your saturation zone
5. Calculate marginal ROAS: Use the slope of the trendline at your current spend level

This method is not perfect: it does not control for seasonality or cross-channel effects. It is still a step up from gut feel. For more sophisticated modeling that accounts for cross-channel interactions, media mix modeling is the next step up.

What Data You Need

At minimum, a useful budget allocation analysis requires:

• Enough weeks of data (more weeks give a steadier picture)
• Meaningful spend variation (a clear range between the lowest and highest weeks)
• More than two channels (with only two, reallocation options are limited)
• Consistent attribution (same attribution window throughout the period)

If you lack spend variation, create it intentionally. Run a test where you increase one channel and decrease another for a few weeks. The change in performance gives you additional spend levels to compare.

The Bottom Line: Marginal Returns Drive Optimal Budgets

Ad budget allocation is not about finding the "best" channel and pouring money into it. It is about finding the optimal spend level for each channel: the point where the next dollar still generates a return above your break-even threshold. Beyond that point, every additional dollar is waste, regardless of how good the channel's average ROAS looks.

Three habits help:

1. Track marginal ROAS, not just average ROAS. Average ROAS tells you how good the first dollar was. Marginal ROAS tells you how good the next dollar will be.
2. Build response curves from real data. Even rough curves from your own spend variation can be a better guide than equal splits or gut feel.
3. Reallocate incrementally. Shift budget in steps from saturated channels to underfunded ones. Monitor the impact. Repeat.

The math is simple. If Google Ads returns $1.10 on your last dollar and Meta returns $3.50, move budget from Google to Meta until those marginal returns converge. That convergence point is your optimal allocation. The single insight is to spend to equalize marginal returns, not to reward past average performance.