“Let’s personalise our customers’ communication” – said every product manager and marketer at least once in their life. I love ambition! But oftentimes, it’s not a trivial task to get there.

Getting started with Customer Data Platforms

First and foremost, they’ll need to plan what data will drive this personalization. Also personalising your customer journey means omni channel communication: SMS, Mail, Mobile app, In-app messaging and even ads.

Customer Data Platforms (or CDPs) support you in centralising data collection and governing where each data point gets delivered to. To name a few benefits of having one, it simplifies the work for your software development team, it gives you more control over your data, and how it’s used for activation.

If you’re still reading, it’s probably because you’re starting your Customer Data Platform journey or are just curious to know more about how to optimise your current setup.

I was a Senior Solutions Architect at Segment, the leading CDP, and I’ve seen many implementations fail or succeed. So let me share a few tips with you so you don’t fall into the most common traps.

Tip #1: Don’t track too many events

Less is more. Especially if you’re just getting started. You’d be surprised by how much value can be inferred by one single event.

To give you an example, I wrote an article when I was at Segment around building a completely automated RFM (Recency, Frequency, Monetary) Value model using one single event: “Order Completed”.

Many companies start with events like “Page Viewed” and “Button Clicked” because it gives a (false) impression of control.

There’s no harm in implementing a simple “Page Viewed” event, but chances are that the value that data point provides is not as meaningful – unless you’re in the content business sponsored by advertising and you would have to have a LOT of traffic to achieve statistical significance.

Additionally, you’ll need someone to own the respective dashboard, be responsible for coming up with an action from their findings and reporting those findings and actions to the team or management.

The key question to take into account is: what are the 2 or 3 user actions that help you move your most important metrics? Signed Up? Order Completed?

Related: Learn about Demystifying Events Data and how to Track like a Pro with Kritika Jalan

Tip #2: Start with one source of data

That could be your website or your mobile app but it’s unlikely that you need both (or all) at once.

Starting with your most important source helps you and your technical team to understand
how your CDP works and achieve value faster.

What is your main source of traffic? Which one drives more revenue or subscriptions? Or if you have a centralised backend for all touch points, that may be the best place to start since this is where important transactions happen.

Tip #3: Don’t underestimate your identification strategy

That can drastically affect costs, the availability of functionality downstream and how you grow your customer data strategy.

“How can it affect costs?” you may ask. Segment and some downstream systems charge you per number of MTU (Monthly Tracked Users) or MAU (Monthly Active Users).

So if you don’t make sure you tell them that customer A is a returning customer this month, you’ll pay for a second user (or more).

I’ve seen a few customers forgetting to assign a unique identifier to their customers so every visitor was a new MTU/MAU. It was a very expensive mistake for some of them.

Also, since we’re talking about unique identifiers, this is very important: Generate your own id. That could be for example, a unique identifier (UUID) generated from your database.

Some databases use incremental numbers for their “customer” table. For technical reasons, that’s not a good practice and complicates migration, future M&A (Mergers & Acquisitions), among other things.

Finally, try not to use “email” as the unique identifier. What if your customer changes their email? Or have more than one email? E.g. work and personal.

Tip #4: Be consistent with your event naming convention

Consistency is at the core of any good process. Implementing a predictive and clear naming pattern for your events would help your customer data strategy to be adopted by developers and data consumers (marketing, product, analytics, scientists etc).

The “Object Action” framework was created by Segment to help you achieve that. It’s simple, effective and scalable. For example: “Button Clicked”, “Page Viewed”, “Order Completed” – where the first part (noun) is the “Object” and the second part (verb) is the “Action”.

I’ve seen this convention successfully adopted by companies of any size and products of many levels of complexity.

Some companies fall into the temptation of creating too specific events like “Sign Up Button Clicked”. Once you do that, you open for exceptions to the rule.

In those cases, you would likely prefer to have a property called “label” within the event that makes it more specific. So “sign_up”, “cart”, “sign_in” etc are just possible values for the property “label” of the “Button Clicked” event.


These tips are not exhaustive and certainly they’re not a blanket rule. They are patterns that solve 90% of the cases.

If you’re interested in knowing more about Segment or just having a chat about your startup technology, please find me on GrowthMentor and stay tuned for more posts like this.

Or listen to the podcast episode that inspired this blog post: