🆕 Beta · Available as a paid add-on. Compass builds on your marketing mix modeling and incrementality results. Contact your Triple Whale account team to enable it.

TL;DR

Compass brings first-party attribution, incrementality, and marketing mix modeling into one measurement system. This article focuses on what happens after those signals come together: how Compass scores how much confidence you can place in each channel's read, and what it does when two signals disagree. For what Compass is and the signals it combines, start with Compass 101: What is Compass.

Overview

Compass 101: What is Compass covers what Compass is, the signals it brings together, and why it exists. This article picks up from there.

Once Compass has combined your signals, two questions still matter for a budget decision: how much can you trust the read on a given channel, and what should you do when your signals point in different directions. Compass is designed to help with both. This is most useful for operators deciding where to invest, where to pull back, and which signal to trust for the decision in front of them.

Key terms

Measurement signal: A source of performance information, such as platform reporting, first-party attribution, incrementality testing, or marketing mix modeling.
Calibration: Using one signal to cross-check and adjust another, so the combined read reflects more than a single method.
Confidence score: An indication of how well-supported a channel's efficiency estimate is, based on how much measurement history and validation that channel has.
Signal disagreement: When two measurement methods tell different stories about the same channel.
Marketing mix modeling, or MMM: A measurement method that estimates how marketing channels contribute to business outcomes over time.
Incrementality: A measurement method that helps estimate whether marketing activity caused results that would not have happened otherwise.

How it works

Bringing signals together

Compass combines first-party attribution, incrementality, and marketing mix modeling into one view of channel performance. MMM provides the primary read for channels that have not been incrementality tested, incrementality results help calibrate and cross-check those estimates, and your first-party order data grounds the calculations in actual business outcomes. Compass 101: What is Compass covers each of these signals in detail.

The rest of this article focuses on the two things Compass adds on top of that combined view: confidence scoring and signal disagreement.

Confidence scoring

Not every channel's read is equally reliable. A channel you have spent on for two years and tested incrementally several times is better understood than one you launched six months ago and have never tested. Compass surfaces this difference so you know how much weight to put on each read.

Confidence	What it may indicate
Higher	The channel has substantial MMM history and has been validated by at least one incrementality test. The estimate is well-supported.
Moderate	The channel has reasonable MMM history but has not been incrementality tested, or was tested once with moderate confidence. Use the estimate directionally.
Lower	The channel has limited MMM history and no incrementality results. The estimate is model-based only. Running an incrementality test may improve it.

Confidence is meant to improve as you accumulate more MMM history and run more tests. Confidence improves with measurement, not with spend. The most direct way to strengthen a channel's read is to run an incrementality test on it.

How compass handles signal disagreement

Sometimes your MMM model and your incrementality test tell different stories about the same channel. Compass does not silently average these. It surfaces the disagreement so your team can investigate.

When signals diverge meaningfully, it is often a sign that something is worth a closer look. The channel's efficiency may have changed between the MMM training period and the test period, or the test design may have had a confounder worth examining. The practical approach is to act where your signals agree and investigate where they diverge. Compass is designed to support both.

When to use compass

Use this part of Compass when the question is not only how a channel is performing, but how much to trust that read before you move budget. It is most useful when:

Platform-reported numbers alone do not give you enough confidence to make the next budget decision.
Two of your measurement signals disagree about the same channel and you need to decide which to act on.
You want to know which channels have stronger or weaker measurement support before increasing spend.

To review and act on this in product, see Compass: Command Center.

How Compass scores confidence and handles signal disagreement