Connect Everything Or Trust Nothing

The short version

A half-connected reporting setup is worse than an obviously incomplete one, because it produces numbers that look complete and are not. This post covers what to connect before you start trusting a blended figure, and why the disagreements between your sources are the most useful thing you have.

The problem with partial connection

Say you connect Meta and leave Google out because it is a smaller line item. Every blended number you calculate from that point is wrong in a specific and invisible direction: it credits Meta with conversions that Google contributed to, and it understates your true customer acquisition cost because a chunk of spend is missing from the denominator.

Nobody notices, because the dashboard renders fine. A missing platform does not show up as a gap. It shows up as a plausible number.

This is why "connect the platforms carrying meaningful spend" is not a setup checkbox. It is the precondition for any cross-platform figure meaning anything at all.

Three kinds of connection, three different jobs

Ad accounts

These supply campaign structure, spend, and delivery data, and they are what allow a change to be made rather than only recommended.

The current set covers Meta, Google, TikTok, OpenAI, and Amazon. Connect every one carrying meaningful spend, including the ones you think are too small to matter — those are exactly the ones that quietly distort a blended average.

Analytics

Google Analytics 4 brings site-side behavior alongside platform-reported numbers.

People often connect analytics expecting it to confirm the ad platforms, then get alarmed when it does not. That reaction is backwards. Platform-reported and site-reported figures are measuring different things through different attribution models, and they are supposed to disagree. The size and direction of the gap is diagnostic information, not an error to reconcile away.

Mobile measurement partners

If you advertise a mobile app, an MMP supplies attribution, retention, and cohort data that ad platforms do not report on their own. Airbridge and Adjust both cover this, with Adjust also carrying SKAN data.

Without one, cohort retention, LTV curves, and ROAS curves have no source. They are not thin or approximate; they are empty. Any app advertiser reading platform-reported ROAS alone is reading a first-week number and calling it a business result.

What order to do it in

Every ad platform with real spend. Nothing blended means anything until this is complete.
An MMP, if you advertise an app. This is not optional for app advertisers, it is the measurement layer.
Analytics, when you want a second reference point on platform claims.

The common mistake is doing this in reverse: connecting analytics first because it feels like the neutral source, then building reports on an incomplete spend picture.

What connecting actually grants

A connection grants read access, and for ad accounts, the ability to propose changes. Changes that spend money or alter the optimization signal wait for approval rather than applying directly.

That distinction matters when you are deciding how nervous to be about connecting a live account. Access and authority are separate: reaching an account is not the same as being allowed to act on it unattended.

Practical rule

Connect every platform carrying spend before you calculate a single blended number. Then treat the disagreements between platform, analytics, and MMP data as findings rather than as a reconciliation chore.