"Customer 360" is one of those phrases that gets used so freely it stops meaning anything. In plain terms, it's one record per customer that pulls together everything you know about them, from every system: what they bought, what they browsed, how they paid, whether they contacted support, whether they open your emails.
The appeal is clear. Right now that picture is probably spread over five places. Your website analytics sees visitors, your order system sees buyers, your email tool sees subscribers, your support tool sees ticket-raisers. Often they're the same person and nobody can tell.
Why it matters in practice
When the pieces are joined, simple things become possible that are awkward or impossible otherwise.
You can stop sending a discount email to someone who bought at full price yesterday. You can see that the customers who complain once and get a fast reply spend more over the next year. You can count how many unique customers you really have. You can find out which first purchase tends to lead to a loyal customer, and make more of that.
You may not need a "platform"
Vendors sell customer data platforms, and they can be the right answer, particularly for large businesses. But for many companies a Customer 360 is just a well-built table in an ordinary data warehouse, filled by a few scheduled jobs. That's cheaper, and you control it.
How to build one
1. List your sources. Write down every system that holds something about customers: store or web orders, app events, CRM, email platform, support tool, loyalty programme, payment gateway. For each, note what identifies a person (email, phone number, a customer id, a device id) and how often it updates.
2. Choose the grain. One row per customer. That sentence sounds obvious until you ask: what is a customer? A person? A household? A company with several users? Decide, and write it down.
3. Get the data into one place. Load each source into your warehouse with a scheduled pipeline. Keep the raw copies untouched.
4. Match the records. This is the hard part and deserves its own section below.
5. Build the customer table. One row per matched customer, with columns such as first seen, first purchase, latest purchase, number of orders, total spend, preferred channel, products bought, last support contact, marketing consent. Document each column.
6. Use it. Build the first uses straight away: segments like recency and frequency groups, lists for campaigns, a dashboard of customer counts and retention. A table nobody queries is wasted effort.
The matching problem
Deciding that "R. Sharma" in the email list, "rohit.sharma@..." in the orders and a phone number in the support tool are the same person is called identity resolution. It's trickier than it sounds.
Start with exact matches on reliable identifiers: the same email address, the same verified phone number. Normalise them first (lower case, trimmed, standard phone format), because a large share of mismatches are just formatting.
Add looser rules carefully. Matching on name and postcode will catch more, and will also wrongly merge two different people. For each rule, decide how much you trust it, and keep the uncertain matches flagged instead of silently merged.
Measure the quality. Take a sample, check by hand whether the matches are correct, and report the rate. Perfect matching doesn't exist, but measured matching is something you can improve.
Things that go wrong
Starting too big. Trying to connect every system at once. Begin with two or three that matter most, usually orders, website and email, and add more later.
No owner. A customer table needs someone responsible for it, or it decays.
Ignoring privacy. Joining data creates a much more detailed picture of a person than any single source did. Be sure your privacy policy and consents allow what you do, limit access, and avoid including sensitive details you don't need. Our guide to data ethics covers the principles.
Building it and never using it. Pick one or two clear uses before you start.
A good Customer 360 doesn't have to be elaborate. A clean customer table, matched sensibly and used every week, is worth more than a grand platform that no one understands.