Response-curve methodology for paid media

Find the exact point where
your ad spend stops making profit.

ProfitCurve fits an econometric model to every campaign in your portfolio. It computes the marginal return on every dollar and tells you precisely where to reallocate.

BirdRock Home case: $7.2M/yr waste identified on 3.7M-row dataset
0
Data points
0
Product families
0
Model types
0
Days to handoff
Real data. Real curves. Real verdicts.

Actual BirdRock Home data — 5 product families across ~600 days. Switch families, drag the sliders, watch everything update.

Simulate spend $727/day
Contribution margin 35%

Spend vs. Sales — Response Curve

Actual data
Fitted curve
Current spend
Optimal spend

Marginal ROAS by Spend Level

Marginal ROAS
Breakeven
OrganicMax observed
Same total ROAS. Completely different reality.

Dog Gates — 2.74x total ROAS

Total ROAS: 2.74x
Marginal ROAS: 2.74x
Spend headroom: +$20,621/day
↑ SCALE — every dollar still profitable

Scenario: Same ROAS, overspending

Total ROAS: 2.74x
Marginal ROAS: 0.4x
Waste: $8,200/day past breakeven
↓ CUT — every new dollar loses money

Total ROAS looks identical. Marginal ROAS reveals where the money actually goes.

All five families. One reallocation picture.

Every family shows growth opportunity. Four of five are capped — the true optimal is beyond the observed data.

FamilyModel Current $/dayOptimal $/day mROASHeadroomVerdict
Current total spend
$3,432/day
Across 5 families
Optimal total spend
$89,209/day
4 of 5 capped at max observed
Total headroom
+$85,776/day
Portfolio heavily underspending
We tell you what the model doesn't know.

Every recommendation comes with warnings. If the model is extrapolating, uncertain, or hitting data limits — you'll know.

Optimal Capped

True optimal is beyond observed data. Capped at max historical spend as safety floor.

Lap DesksDog GatesOttomans

Inflection Beyond Data

Model inflection point past observed range. Extrapolation risk if scaling aggressively.

Lap DesksDog GatesOttomans
📊

Ceiling Beyond Data

Model ceiling exceeds observed max. Saturation point can't be verified.

Dog GatesOttomans

mROAS Still Positive at Max

Marginal ROAS above breakeven at highest spend. More data needed for true saturation.

Lap DesksDog GatesOttomans
Discovery to handoff in four weeks.

Week 1

Discovery + Data Audit
  • Channel data audit (90+ day history)
  • Attribution cross-reference
  • Margin mapping per SKU
  • Workflow shadow

Week 2

Infrastructure + Fitting
  • DuckDB + ProfitCurve deployed
  • Data ingested + normalized
  • Per-campaign curve fitting
  • Bootstrap CI computation

Week 3

Reallocation + Review
  • Portfolio optimizer runs
  • Reallocation plan generated
  • Stakeholder narrative
  • Review + sign-off

Week 4

Execute + Handoff
  • Reallocation executed
  • Performance monitoring
  • Team training workshop
  • Full handoff Day 30
Eight deliverables. You own everything.

Per-Campaign Response Curves

Every campaign gets its own fitted model with bootstrap confidence intervals. Any channel.

Portfolio Reallocation Plan

Specific dollar moves. "Move $X/day from overspenders to underspenders."

Marginal ROAS Dashboard

Where every campaign sits on its curve. Verdict bar with narrative. Updated weekly.

Safety & Warning System

Capped-optimal detection, inflection flags, ceiling warnings.

ProfitCurve Engine

Proprietary Python library deployed to your infrastructure. Licensed for your team's use.

Stakeholder Narrative

AI-generated executive summary for CMO/CFO: findings, actions, impact.

Team Training + Runbook

Your team trained to interpret, run, and execute reallocations.

20-Article Methodology KB

Hill curves, saturation, water-filling allocation, bootstrap CIs, safety rails.

Pays for itself in 1–2 months.
$8,000 – $25,000
Depends on channels, portfolio size, and data complexity
Typical ROI: 15–35% of spend recovered → $150K–$350K/yr on $1M/mo portfolio
  • 30-day embedded engagement
  • ProfitCurve engine deployment
  • Per-campaign curve fits with CIs
  • Portfolio reallocation plan
  • Stakeholder narrative
  • Team training + runbook
  • 20-article methodology KB
  • Full source code ownership
Book a Call
Optional retainer: $3–5K/mo for quarterly re-fits
Common questions.
What's the difference between ProfitCurve and Amazon Profit Studio?
Amazon Profit Studio is a turnkey SaaS app for Amazon-only sellers. ProfitCurve is a multi-channel consulting engagement — Amazon, Meta, Google, TikTok, and any channel with spend/conversion data.
Can this work with my existing MMM model?
Yes. MMM tells you channel-level allocation. ProfitCurve tells you within-channel campaign-level allocation. We ingest MMM outputs as priors.
What if I don't have 90 days of data?
Thin-data campaigns get flagged via the warning system. Safety system prevents recommendations on insufficient evidence.
How do you handle "capped optimal"?
When mROAS is still above breakeven at max observed spend, we flag it as "limited visibility" and recommend testing higher spend. We never extrapolate without warning.
Do I need a data science team?
No. The engagement includes training your PPC team to interpret and act on results.
Do I get access to the ProfitCurve engine?
Yes. The engine is deployed to your infrastructure and licensed for your team's ongoing use. You don't need to depend on us after handoff — your team runs it independently.

Bring your spend data.
We'll show you the curve.

30-minute call. We'll look at your portfolio and tell you where the money is.