The Problem
I’m sure this sounds familiar, revenue is growing but so is your CAC...and your net profit is still slim despite the revenue growth.
You push more budget into Meta & Google ads and they’ll happily take it...
But your return on that investment seems softer & you're wondering what the ACTUAL bottleneck
So you get into the cycle of spending more with lower and lower returns.
Or you pull back spend and your ROAS get better but your growth is flat or declining.
Your blended & channel CAC/ROAS number are a trap
Even when we split out new & existing customers (which all high LTV brands should do) your blended CAC or ROAS numbers averages out your cheapest customers and your most expensive customers.
It always looks good until it suddenly isn’t you go into panic mode.
Assuming you have a good business doing $150k+ per month the problem may very well be not to do with you ‘spending to much’ but more so not knowing what your diminishing returns curve looks like!
Which is to say - Do you know what each marginal or incremental dollar of ad spend it contributing to new customers and revenue.
The cool thing is, you can look back at your data and do the math to predict pretty accurately what an extra $20,000 in ad spend for the month will bring in and what the CAC & ROAS will be specifically for that extra $20,000 investment only, not blended with your existing spend and revenue.
Here’s how I did it when I worked at PetCircle.com ($600m+ ecommerce pet brand)
I used to work at PetCircle.com, they are an Aussie pet brand mirroring what Chewy.com ($10 billion market cap) did in the US.
When I worked there they had raised a lot of money & the strategy was to gain market share by knowing how much net profit a customer was worth over 12, 18 & 24 months.
Then they could out spend competitors to acquire customers, which meant loosing money on acquiring new customers most of the time. We'd look it our CAC as a payback period, if our our CAC was $120, we'd know that it would take 12 months to get positive return for example.
They of course had enough cash in the bank though to be able to keep the lights on whilst they waited for the lifetime net profit of each customer to come true overtime.
Part of why this strategy works is because of volume of customers and 100%+ growth rates year on year. In order to grow that much we needed to spend a lot on reaching new people.
Because we we’re forced to grow & spending so much, we needed to be very intentional and know the marginal/incremental impact of more spend overall as well as within each channel.
That lead me to doing an incremental CPA/ROAS analysis for Meta & Google ads and seeing what the incremental CPA of more spend was for each channel.
For example if we spend an extra $100k this month on Google ads, what is will that incremental investment actually return looking in isolation.
If all other variables like sales, seasonality etc are similar and we get an extra 2,000 customers the incremental CAC would be $50 ($100,000 / 2,000 = $50)
"This incremental CPA shows the cost of your extra investment"
That distinction is everything.
Because the decision you're making every week isn't "should I keep my average spend level?"
It's "should I spend this extra $50k? And will that be profitable?"
And the only number that answers that question honestly is the incremental CPA forecasting.
If your LTV:CAC target caps you at $100 per customer, and your incremental CPA with an extra $30,000 weekly spend is $120, every dollar you spend above that threshold is actively destroying margin.
Let’s break this down.
Current state:
- Current spend is $50k
- CPA: $100
- Weekly New customers: 500
With $30k incremental spend:
- Weekly spend: $80,000 ($50k + $30k)
- CPA: $120 (rises with extra spend)
- Weekly new customers: 667
From that extra $30k spend we got a extra 167 customer at an incremental CPA of $180 ($30,000 / 167 = $180) so it actually cost you $180 to get those extra 167 new customers…
Not the blended $120 you’d normally see ($80k / 667 = $120)
$120 is uncomfortable but close to target. Manageable, right?
Wrong. Your actual cost to acquire each of those 167 additional customers is $180. Against a $100 breakeven, you're losing $80 per incremental customer…
167 customers × $80 loss = $13,360 in the hole every single week on that extra spend alone, all from a 20% budget increase.
You'd never see the $180 without doing this analysis.
Knowing your incremental CPA curve gives you three things average CAC never can:
A hard ceiling. The exact spend level where your next dollar stops making sense. Not a gut feeling, a solid number.
A forecasting tool. If you want to grow new customer volume by 20%, you can read off the curve exactly what that will cost you in incremental CPA before you spend a dollar.
A budget defence. When someone in the business wants to cut marketing spend, you can show precisely what customer volume you lose per dollar cut and at what spend level cuts actually improve efficiency rather than hurt it.
Here’s how to do the diminishing returns analysis
I’ll show you how to do it with an example using the Meta ads channel, it’s the same process for doing it for all channels.
Remember no model is perfect so the forecasted marginal CAC is directional.
1. First download by week and month all your data with spend & new customers in Meta ads for the past 2-3 years. The more data the more accurate your model will be.
Note down any outlier spend months like BFCM, external factors like COVID and so on. Flag this weeks/months and run the analysis twice, once with and once without the outlier months to see if there is a big difference.
2. Organise your data by month and week with spend and new customers, if you don’t have a new customer optimisation event just use all purchases and filter out any retention campaigns you may have.
3. Once you have all the data in a Google sheet sort by week oldest to newest and tag any outlier months you may have.
4. Now filter by highest spend to lowest spend & calculate the incremental, spend, avg CAC, Incremental CAC & incremental ROAS.
5. Highlight the spend, avg CAC & Incremental CAC columns & create a scatter chart with spend on the X axis and Avg CAC & Incremental CAC on the two Y axis’s.
6. Click a data series → Edit → Trendline → Logarithmic. Do this for both avg CAC and incremental CPA series.
7. Check "Show R² value" — this is your model fit score. You want R² above 0.75.
8. Add a third data series, two rows only, your minimum and maximum spend values, both with a Y value equal to your CAC target (e.g. $100). Format it as a dashed line. This is your ceiling, where the incremental CPA line crosses it is your efficiency cliff.
Where your incremental CPA trend line crosses the $100 reference line, that X value is your weekly spend ceiling. Every dollar above that number is acquiring customers at a loss.
What the analysis found:
The efficiency cliff for this brand sits at approximately $80-120k AUD per month. At that level, incremental CAC hits ~$80. Every dollar above that threshold is acquiring customers above your breakeven target. The CPA table and spend explorer tabs show this in detail.
What this means practically:
The curve is directionally correct, diminishing returns are real and visible, but the exact numbers carry varied confidence intervals.
The R² problem and why it matters:
R² came in at 0.71. That's below but on par with 0.75 target. This tells us that spend is a pretty good predicator vs external other events. This isn't a flaw in the method, it's the data telling you something real:
Sometimes if your R² is lower it can mean customer acquisition efficiency is driven by more than spend level alone.
You should look at the raw data again and you may find, the July–September 2024 period was dramatically more efficient than the same spend levels in 2025, same spend, completely different customer volumes.
That's creative performance, product mix, or audience saturation moving the results, not spend. You can't capture that in a spend-only curve.