The Backtest
Make the Model Prove Itself
Every measurement vendor asks you to trust them about the future. We'd rather be tested on the past.
The Backtest replays 12 to 36 months of your historical data through the Mediaura Causal Engine and shows you, specifically, what the evidence would have changed: the budget cut that cost more than it saved, the channel quietly cannibalizing another, the spend that was buying demand you already owned. You already know what actually happened. That's exactly why this works: you can check our answers.
Why a Backtest, Not a Demo
Anyone Can Demo a Dashboard
A demo shows you software. A backtest shows you your own business with the fog removed. If the causal engine can't find anything in your history worth acting on, you'll know before you've committed to a platform. If it can, you'll see precisely what running on evidence would have been worth. Either way, the analysis is yours to keep.
How It Works
Three Steps, No Surprises
The Fit Check
Before you pay a dollar, we assess whether your data can support causal conclusions: transaction volume, history depth, channel mix, and tracking coverage. Causal methods need enough events to separate signal from noise, and the honest answer varies by business. If your volume can't support confident conclusions, we'll tell you exactly that, explain what would need to change, and part friends. Typically 2 business days from data access.
The Backtest
A fixed fee and a fixed timeline, both scoped to your footprint after the Fit Check and locked in writing before we start. No hourly billing, no scope creep, no surprise invoice. We replay your history through the same pipeline that runs Signal in production: signal repair, identity stitching, revenue mapping, then causal modeling validated against the natural experiments already sitting in your past (the location that opened, the campaign that paused, the competitor that closed). Every finding is tied to a decision you actually made or could have made.
The Verdict
A working session and a written readout in operator language: contribution margin, cost per admission, cost per cover. What was causal, what was coincidence, what was cannibalization, and what it means for next quarter's budget. No statistical vocabulary at the podium. The full diagnostics ride along in an appendix for anyone on your team who wants to check the math.
From there, if the findings earn it, Signal runs continuously so every future budget meeting starts where the Backtest ended: from evidence.
What You Receive
Everything, yours to keep.
- The Verdict readout, written for the room where budgets get decided.
- A quantified list of the decisions the evidence would have changed, each with its estimated dollar impact and the confidence behind it.
- A tracking and data-quality audit: what's broken in your measurement today and what it's costing you.
- The methodology appendix, auditable by your analyst, your auditor, or your board.
- Yours to keep, whatever you decide next.
Common Questions
The Questions Buyers Actually Ask
What data do you need?
Ad platform exports, your analytics history, and the system where revenue actually lands: POS, CRM, EMR, or call tracking. The Fit Check produces the exact list for your stack. Read-only access is all we ever take.
We're in healthcare. Can you even touch our data?
Yes. Signal's infrastructure is HIPAA-compliant by design and a BAA is included before any PHI-adjacent data moves. The Backtest runs inside the same compliance boundary as the production platform. Details on the security page.
What if the model finds nothing?
Then you've learned your current allocation is closer to right than most, and the audit portion still pays for the exercise by finding what's broken in your tracking. In practice, most first passes uncover unmeasured revenue or a misallocated channel inside the first review.
What if we're too small?
The Fit Check exists to answer that honestly before money changes hands. Causal confidence depends on event volume, and we would rather decline the engagement than deliver confident fiction. That's the whole point of the company.
Who does the work?
Mediaura engineers and the production Signal pipeline. Not an offshore analyst with a spreadsheet template.
Start With the Fit Check
Free, honest, and binding on us, not you. If your data can support causal conclusions, you'll get the fixed fee and timeline in writing. If it can't, you'll know that too.
Make the Model Prove Itself
Every measurement vendor asks you to trust them about the future. We'd rather be tested on the past.
What happens next:
- Free, honest, and binding on us, not you.
- Read-only access is all we ever take.
- The Backtest runs inside the same compliance boundary as the production platform.
- Mediaura engineers and the production Signal pipeline. Not an offshore analyst with a spreadsheet template.
Security team reviewing us? Read our security posture →
Bold patterns. Boring rigor.