21 Mar 2026 · Identifiers

Click IDs and the install handshake

A campaign-to-install claim is strongest when a click identifier is minted at tap, stored, and still present when the app first opens. That is the handshake. Everything else — device graphs, IP-plus-UA residue, “likely the same person” — is a weaker sentence. Both sentences can appear in an MMP. Only one of them should be presented as proof.

Students in the intensive are asked to find the handshake in a redacted export before they discuss CPI. If the click ID is there, ownership is a configuration question: which lookback, which deduplication, whether a later click steals the install. If the click ID is gone, ownership is a humility question: will you claim probabilistic residue, or will you mark unknown?

Where the handshake breaks

Store redirects that strip parameters, delayed opens past the identifier’s lifetime, and privacy changes on iOS all break the handshake. LINE-to-app paths popular in Thailand sometimes survive; sometimes they do not, depending on how the deep link was wrapped. Teams that assume every chat click carries a durable ID are surprised when the MMP fills the gap with modelled matches.

Modelled matches are not automatically dishonest. They become dishonest when they are formatted like click-ID rows in the same column. Vault Nodecore’s rule is typographic as much as analytical: residue gets a different label in the install graph. Finance can then decide how much weight to give it. They cannot decide if you have already laundered residue into “attributed installs.”

Click flooding as a fake handshake

Fraud desks know that a click ID can be minted without a meaningful human tap. A flood of identifiers shortly before organic opens will look like a perfect handshake in a naive export. The fraud studio block treats volume, timing, and device repetition as reasons to doubt the handshake even when the ID is present. Presence is necessary for a strong claim. It is not sufficient.

If you only remember one practice from this note: when you present campaign-to-install numbers, say whether the week’s ownership is mostly handshake or mostly residue. That single sentence prevents a lot of theatre. The Install Attribution Intensive makes it week two, before anyone is allowed to argue about media mix.