Skip to main content

Home / The Buddy System / Measure Buddy

Measure Buddy

Measure Buddy: which dollars sold cars, and how sure we are.

It reads tracked media spend against retail sales, market by market. It fits a mix model, designs geo holdout tests and reads them out with a confidence range. Every number carries a label: measured, modelled or projection. When the data cannot prove a channel works, it says so.

Measure Buddy Sandbox mix model: simulated markets with known true effects, channel contributions and a budget shift, labelled Sandbox

How it works

From spend to a test you can trust.

Five steps, each one honest about what it can and cannot tell you.

Bring in spend and salesTracked media spend by market, month and channel, next to retail sales. It checks that each month is complete before it uses it.
Fit a mix modelAdstock and saturation, fitted across markets and months, then tested on markets it has not seen. A bootstrap shows how wide the range really is.
Design a holdoutPick the markets or let it pick them. It builds a matched control and tells you the smallest lift the test can reliably see before anything changes.
Read it outA lift with a 90 percent confidence range. A result is labelled measured only after a real change in the test markets was recorded.
Shift the budgetSame total, each channel moved within a limit you set. It becomes a draft Media Buddy plan, and a person approves every dollar.

What it answers

Six questions every media budget raises.

Mix model

Contribution and response curves for each channel, and what the next million dollars would buy. Labelled modelled, with the range beside every number.

Geo holdout lift tests

Design, power and readout in one place, including an A/A check on past data that should read close to zero when the design is sound.

Offer lift

Did richer offers move sales? It compares markets inside the same period and gives a plain verdict, including “leaning positive, not reliable” when that is the truth.

Attribution

A modelled split between base sales and media, and site sessions against sales within each market. Where there is no lead data, it shows zero rows instead of an estimate.

Hands off to the team

A budget shift becomes a draft Media Buddy plan. A winning test starts a Campaign Buddy campaign. Any buddy chat can answer “what is working?”

Labels you can check

Measured, Modelled, Projection, A/A check and Sandbox labels on every screen. Simulated numbers are never shown as results.

Inside the product

Real screens, real product.

Captured from the working product in the demo environment. These screens use the Sandbox: simulated markets with known true effects, labelled Sandbox on screen. They are not client results.

Measure Buddy Sandbox mix model with channel contributions, ranges and a budget shift panel — The Sandbox mix modelSimulated markets with known true effects show what a full read looks like, and prove the engine recovers the truth.

Measure Buddy lift test readout for a streaming video holdout, labelled Sandbox, with the confidence interval — A lift test readoutLift against a matched control with a 90 percent range, next to the test design. This readout is simulated from the design.

Measure Buddy budget shift from the Sandbox model, with today and proposed spend by channel and a Send to Media Buddy panel — From a budget shift to a draft planSame total budget, channels moved within a limit, sent to Media Buddy as a draft a person approves.

What it found on real data, and why we show it. Run on more than a year of tracked media spend and retail sales for one automotive brand, the mix model could not separate any channel’s effect from zero. Measure Buddy reports that as a finding and points to the next step, a geo holdout, instead of ranking channels on noise. The Sandbox, with simulated markets and known true effects, shows what a full read looks like: every simulated effect was recovered within 8 percent.

Where it stands today. Measure Buddy is built and running inside The Buddy System on real tracked spend and retail sales. Lift is a projection until a holdout has run in market. Nothing is spent without a person’s approval.

Measure Buddy closes the loop.

It measures the plans Media Buddy makes, sends budget shifts back as drafts, and starts a campaign in Campaign Buddy when a test wins.

Meet The Buddy System →

The Buddy System by Onit Digital

See Buddy work.

Give it the goal. Get the campaign. A person still presses send.