It flags what it cannot measure
A channel without enough spend variation to identify a curve is marked unmeasurable and kept out of the recommendation. It does not get a plausible number and a quiet caveat.
Marginal return, not average
Your best channel is usually the worst place to put the next dollar.
Every dashboard you own reports average efficiency: Search CAC is $170, LinkedIn CAC is $294. That tells you which channel was efficient last quarter. It cannot tell you where the next dollar goes, because that is a marginal question, and the answer is often the reverse.
GainGauge fits a response curve to each channel, takes its derivative, and moves budget until the marginal return is equal everywhere. Then it explains the move in language you can take into a board meeting.
Marginal return, live
Illustrative model, not customer data
Paid Search
$940
next customer · avg $170
LinkedIn
$307
next customer · avg $294
Customers per month
685
on the same $120k budget
Versus today's mix
0
this is today's mix
The argument
Average CAC answers a question nobody is asking.
A channel with a $170 average CAC that is fully saturated may cost $940 for the next customer. A channel with a $294 average CAC that is still climbing may cost $307 for the next one.
Average CAC says double down on the first channel. It is wrong, and it is wrong in the direction that costs the most money, because the saturated channel is usually the one with the longest track record and the loudest internal advocate.
The distinction is not a reporting nicety. It is the difference between a budget that compounds and a budget that plateaus while every dashboard stays green.
| Channel | Spend | Avg CAC | Next customer |
|---|---|---|---|
| Paid Search | $112,000 | $170 | $940 |
| $8,000 | $294 | $307 |
Read the last column, not the third. On this mix, the next $10,000 buys about 9 customers in Search and about 31 in LinkedIn.
How it works
Four stages, and the arrow only points forward.
Nothing downstream is allowed to recompute a number. Every screen, every alert and every sentence of the summary reads from one analysis, which is why two people looking at two screens never see two different figures.
-
Stage 1
Connect
Ad platforms and your CRM, read only. Spend is converted to your currency once, on the way in, and never guessed.
-
Stage 2
Resolve
Click IDs, then email, then company domain. Every deal is traced to the touchpoints that produced it, and the match rate is reported rather than assumed.
-
Stage 3
Fit
A response curve per channel, with adstock and lag, plus a confidence interval from several hundred bootstrap replicates.
-
Stage 4
Decide
A budget where marginal return is equal, the expected gain with its range, and a written record of what you chose and whether it was implemented.
What it refuses to do
The restraint is the product.
Any model can produce a confident-looking reallocation. The work is knowing when not to.
It never invents a figure
Every sentence in the summary cites the computed evidence behind it, and a claim that fails its citation check is repaired or replaced with the deterministic version. No number originates in the narration layer.
It will not touch your ad accounts
Recommendations are read, argued with and applied by your team. Automated push-back to the platforms is deliberately out of scope, because a wrong number that spends money by itself is a different category of mistake.
Sources
Built for B2B, where the sale takes months.
Considered-purchase businesses spending roughly $50k to $500k a month, with a CRM as the system of record and sales cycles measured in months rather than minutes. Lag between the click and the closed deal is modelled, not ignored.
Google Ads
Meta
LinkedIn Ads
GA4
HubSpot
Next step
Bring one account. We will show you the gap.
A walkthrough runs against your own spend and your own closed deals, and ends with a number: what the current mix costs you per month against a mix where marginal return is equal. If the gap is small, we will tell you that too.