What a first-party data strategy actually involves.

Data a business collects directly from the people who deal with it: what they did on its site or app, what they bought, what its CRM holds about them, how they responded to its email, what they told its support team. The business collected it, under its own terms, and holds it. For an agency advising a client, that makes it the client's data, which is worth saying early, because it decides who has to agree to everything that follows.

What does not count is just as useful to know. A platform's report about a client's ads is the platform's data, shared on its terms. Audiences bought from somebody else are third-party. Both can change or disappear without the client's agreement.

Browsers block more tracking, consent decides more of what may be collected, and platforms fill more of their reporting with modelled estimates. The share of the picture you can observe directly through somebody else's tools keeps shrinking.

First-party data is the part you can see completely, because it never depended on a third party's cookie or interface to exist. It is also the part every other measurement can be checked against.

A strategy is mostly these four, done properly.

• An inventory

  Every system holding customer data, what it holds, who owns it, and which identifier it uses. Most teams find more systems than they expected and fewer shared identifiers.

• A home

  One controlled place where it all lands, which in practice is a data warehouse in a cloud account the client or its agency controls. Data spread across rented tools is collected, but it cannot yet be used together.

• Identity

  The keys that join one system to another: a customer ID, an order ID, an email address handled with care. A join is only as good as its keys, and the keys are usually the weakest part.

• Permission

  Consent recorded alongside the data it covers, with what it was given for. Data you hold but may not use for a purpose is a liability for that purpose rather than an asset.

Measurement first: checking what the platforms claim against the orders and leads you recorded yourself. Then segmentation, deciding who is valuable using everything you know rather than what one tool sees. Then activation, sending audiences and conversions back to the platforms. And modelling, such as mix models that need years of consistent history.

The order matters. Clients who start with activation send the platforms data nobody has checked, and the platforms optimise towards it all the same.

Collecting without a home, so the data exists in six tools and is joined in none. Keys that do not match across systems, so the join silently drops most customers. Consent stored in the banner tool and nowhere near the data. And no owner, so the inventory is accurate on the day it is written and wrong a quarter later. An agency is often well placed to fix that last one, because it already touches most of these systems to produce the reporting.

SightShark is the home and the plumbing into it: a client's CRM, store, analytics, email and ad platforms, from 200+ sources, landed and modelled in a data warehouse in your own cloud account and scoped to that client, so the data sits in an account you control rather than inside a rented tool.

It does not resolve identities beyond the keys your systems already share, and it does not push audiences back to the platforms. Both are separate work, and both are considerably easier on modelled tables than on raw exports.