'%20opacity='0.55'/%3e%3crect%20x='66'%20y='66'%20width='892'%20height='892'%20rx='168'%20ry='168'%20fill='none'%20stroke='%231C3A40'%20stroke-width='4'%20opacity='0.55'/%3e%3crect%20x='243'%20y='560'%20width='74'%20height='140'%20rx='24'%20fill='url(%23barTeal)'/%3e%3crect%20x='359'%20y='470'%20width='74'%20height='230'%20rx='24'%20fill='url(%23barCyan)'/%3e%3crect%20x='475'%20y='370'%20width='74'%20height='330'%20rx='24'%20fill='url(%23barTeal)'/%3e%3crect%20x='591'%20y='260'%20width='74'%20height='440'%20rx='24'%20fill='url(%23barCyan)'/%3e%3crect%20x='707'%20y='140'%20width='74'%20height='560'%20rx='24'%20fill='url(%23barTeal)'/%3e%3cpath%20d='M%20280%20540%20C%20330%20505,%20372%20468,%20396%20450%20C%20432%20424,%20470%20380,%20512%20350%20C%20556%20318,%20598%20262,%20628%20230%20C%20666%20190,%20708%20152,%20744%20120'%20fill='none'%20stroke='%23A5F3FC'%20stroke-width='16'%20stroke-linecap='round'%20stroke-linejoin='round'%20opacity='0.92'/%3e%3c/svg%3e){kind=small}
SignalRig.
A media mix model that shows its work · $249 once
SignalRig turns spend and outcome CSVs into calibration curves, holdout accuracy, and plain-language budget verdicts. Fully offline on your Mac. No account. No cloud.
Coming soon to the Mac App Store.
The receipts
Results from the fully worked sample model that ships in the app. Your data earns its own receipts.
02 / Proof
Fig. 01 · MMM Lab
2480 × 1830 px
MMM Lab
Pull the slider. Fitted curves answer.
Move budget between channels and watch two fitted response curves answer back, in leads and in dollars, with a credible interval attached to every number. No adjective grades the trade; a probability does. Every scenario carries a plain-language verdict, not just a chart to decode.
Fig. 02 · Data Drop-In
Shown here: a failing package, on purpose.
1784 × 2340 px
Data Drop-In
Drop a CSV. Watch it get graded.
The screen above isn't the happy path, it's a package that fails on purpose, because that's the honest demo: specific, actionable errors instead of a vague "invalid file." Readiness is scored 0 to 100 against the same contract the model trains on, so the fix list is exact and the score never lies to make you feel better.
Fig. 03 · Model Card
2616 × 3230 px
Model Card
The graded exam no vendor shows you.
Hidden true values, planted in advance, checked against what the model recovered. A twelve-week blindfold forecast, scored on weeks the model never saw. Sampler diagnostics translated out of jargon. Most vendors ask you to trust a score; this is the exam that produced it.
03 / Privacy
Your data never leaves your Mac.
Pricing
No subscription. No seats. No upsells inside the app.
Private beta running now. Want in early?
Learn
MMM 101, the visual cheat sheet
A one-page primer on baselines, adstock, saturation, and marginal cost.
The Graded Exam, our measurement standard (coming soon)
How the parameter-recovery and holdout tests work, written out in full.
Frequently asked
Does it fit a model on MY data yet?
Not in this build. Today the app ships a fully worked sample model end to end and validates your own CSV exports against a strict data contract with specific fixes. In-app fitting on your own data is the next milestone.
What data do I need?
Weekly spend and outcomes by channel. Two files required, three optional. The validator tells you exactly what is missing and whether your history is deep enough to model honestly.
Why offline?
Client spend data is confidential. Software that phones home is a procurement problem and a trust problem. SignalRig has no network capability at all.