The key word here is ‘event’. An event is a record of a single instance of behaviour taking place on a website or app. It’s tied to a user’s identity and useful information (attributes) can be stored inside it.

This practice is sometimes referred to as trigger marketing, event tracking, event driven marketing or event streaming. The sheer power and potential of this new type of marketing is driving massive advances in the way digital marketers, growth hackers and product marketers work.

There are two components: managing identity and recording behaviour.

Identity

When a user visits a website or mobile app they are automatically assigned an anonymous identity. The anonymous identity is used to store the record of their behaviour until, and if, the user becomes ‘identified’. Identified users are known users who already have a record in the user database. At this point, the anonymous identity and the known identity are joined together so that all past behaviour recorded against the anonymous user is now stored directly against the known user.

Without a user’s identity we would not be able to assign the stored behaviour to a particular person. This would limit our ability to communicate with the user (we can’t send them an email or push notification if we don’t know who they are). Instead, we would receive general data about the way the whole service is being used by all users (think Google Analytics). Identity based marketing also allows us to track users across different devices more easily.

Basic Recipe of an Identify Call

analytics.identify([userId], [traits], [options], [callback]);

Behaviour

Behaviour is information that describes how the user interacts with the website or app. It is recorded as an ‘event’. An event is a single instance of measured behaviour; a single event in the journey of the customer. The event could describe a users explicit action such as ‘Clicked on Playlist’ or an implicit action such as ‘Moused Over Playlist’.

Basic Recipe of an Event

analytics.track(event, [properties], [options], [callback]);

Attributes can be parsed in to the event allowing for unique, and highly relevant knowledge to be stored inside the event itself. For example, the playlist name, the playlist ID, the duration, and the description could all be stored in the event ‘Clicked on Playlist’.

analytics.track("Clicked on Playlist", {
playlistName: "Australian Top 50",
durationMins: 132,
playlistID: 2388383820022992,
playlistDescription: "The latest and greatest hits topping the charts down under."
});

By measuring behaviour in this way and storing key information about the behaviour inside the event, we can send the user a much more contextual message or prompt when trying to engage them.

Origins

Event stream processing has long been used by software engineers to build event-driven systems. An event would be use to record a significant change in the state of the object. By measuring behaviour in such incremental detail we are able to more clearly identify exactly where the user stops in a given flow or process. Algorithmic trading desks are an example of such systems. In this example each change in the price of a stock is measured as a single change in state, down to the cent. By monitoring all movement in the stock price, an algorithm can treat each event as a new state, and better predict the next movement based on the current and past states. Such granular measurement allows the algorithm to react quickly and accurately. Contrast this to simply knowing the starting price and the current price of a stock at any given time of the day (the way humans typically monitor stocks).

Event-based marketing is literally the marketers equivalent of this practice. By recording much more information about a user’s behaviour—each change in the user’s state—we know exactly where a user is in their journey and can construct a highly relevant and personalised message to deliver to them.

The word ‘event’ might ring some bells for us old-school marketers. Google Analytics has long offered the functionality of recording and storing custom events about how users interact with the product. This was great, and certainly offered us some insights into those more granular and unique behaviours, but there was always one problem… identity. Who are these people?

Example

Let’s use a typical abandoned cart campaign as a way to compare event-based marketing to traditional behavioural campaigns.

Traditional Behavioural Campaigns

Typically when implementing an abandoned cart campaigns we track two stages:
1) when the user starts the checkout process and
2) when the user successfully checks out.

More advanced implementations might track the step in the checkout process the user was at. This allows them to send the user back to the exact step they were up to. In total, most abandoned cart campaigns are measuring two or three stages. More often than not, this is done by querying the clickstream data (specifically the URL path) for users who have viewed the ‘checkout page’, but not viewed the ‘payment confirmation’ page. Because of the vagueness around this method of tracking, we typically wait 2-3 hours before sending the abandoned cart campaign.

Event-based Campaigns

When implementing an event-based abandoned cart campaign, we can record many more events:
1) User adds item to cart
2) User starts checkout process
3) User confirms shipping details
4) User confirms billing details
5) Payment successful
6) Payment unsuccessful
7) User successfully checks out

Each time a user triggers the next event in the sequence their state changes. This helps clearly indicate the exact point in the flow that they are up to. The moment events stop being received (the user abandons the checkout process) we need only wait a short period of time before sending the abandoned cart email (20 minutes for example). If the user stops at step 4, where they have confirmed their billing details but no result has been received for payment, we need simply send them an email encouraging them to finish paying.

On top of this, we are able to store information about the progression through the cart inside the event itself. This can be parsed through from the email to the target (website or app) by attaching the information in parameters on the link.

How to Implement Event-based Marketing

Event-based marketing significantly improves the capability, speed and quality of marketing. The technology is rapidly advancing and new tools are emerging designed specifically around these capabilities. This comprehensive guide to event-based marketing is designed to help digital marketers utilise javascript “events” to trigger marketing, product, analytics, advertising and communication campaigns based on user behaviour.

To be clear, this is not a guide about promoting physical events. This is a guide about using javascript based events to automatically trigger marketing.

In the context of digital marketing, “events” record a single instance of a behaviour occurring on your website or app. They are called events because:

1. they record a change in state, and
2. Javascript technology uses events to trigger functions.

I’ve worked with event-based marketing since 2013 and in this article, I’m going to teach you everything I know about this field of marketing.

Does event-based marketing work?

Yes, event-based marketing works and it’s probably the single biggest opportunity for growth in almost every business today.

I’ve been using event-based marketing for around six years now. I’ve implemented it successfully in my own business, and have helped many other companies implement it. In every single case, those businesses saw an immediate company-wide increase in every metric that matters. From revenue to email deliverability. I’ve literally seen 400% to 600% increases in conversion metrics overnight. Event-based marketing absolutely works.

The performance of every single marketing channel and activity is enhanced when event-based marketing is added. Even interstitial popups convert well when enhanced with event-based marketing.

Yet, despite all of this… very few people know how to do it.

This makes sense though because it’s virtually impossible to learn. There is literally zero content online about it. Not a single guide or course… a few articles here and there, but that’s about it. So, being the content-savvy marketer I like to think I am, I started writing about it (and teaching it in real life). This was a great decision because it’s helped me refine all of the content I’m about to share with you. The diagrams, the explanations, the metaphors, the strategies and most importantly the application of the concepts are all better because of the real-life testing that has taken place in the classroom. Let’s begin…

Technical background to event-based marketing

It wasn’t until the early 90s that marketers first started tracking people visiting their websites. Unlike the way most marketing professionals approached their job at the time (think Don Draper), this new breed of marketer was fascinated by measuring whether their marketing activities actually worked. These data-driven-pioneers were the first ‘digital marketers’ because of one thing: data.

Access to data was the turning point because it closed the feedback loop of marketing. Immediately every single marketing function became measurable, which in turn creates an opportunity to optimise our marketing. Data also speeds the entire marketing process up; if you’ve ever tried to optimise a website with no traffic you’ll know what I’m talking about.

From this moment on, data gave us the ability to measure our performance as marketers and compared to the incumbent alternative of little to no measurement, it probably felt like it was all happening in real time. Indeed, this was the birth of digital marketing, and the data source we came to rely on was server logs.

Server logs

Initially, the ability to actually measure the traffic to your website and the pages people were interested in relied on the most basic analytics technology we have: server logs. Server logs still work to this day. They show us the documents being requested by people who visit our website. When PHP, XML and HTML were the default web languages this worked well because every change in the state of a page (and the information contained upon it) relied on a call being made to a server. The typical user path on a website would look something like this:

The server log would tell us which documents were being requested on the server, and in which order. Beyond this, marketers couldn’t do much more… but neither could websites, so the data still felt pretty powerful. However, far more meaningful metrics would become available if we moved the tracking scripts to the client side. Enter ‘click-stream analytics’.

Click-stream analytics

Although they were not the first with this technology, the release of Google Analytics in 2005 democratised online analytics by introducing a free analytics package that would enable anyone with a website to begin collecting in-depth analytics about site usage.

The technical innovation was based on the repositioning of the data collection away from a server-side implementation to a client-side implementation. Rather than tracking document requests on the server, we would monitor requests from the front end of the website itself using a piece of javascript. In other words, we tracked the “click” side of a document request, rather than the response side of a document request. Clicks are an important evolution because they also represent a shift towards intention based metrics.

This repositioning of tracking allowed us to obtain an ever increasing number of new data points. Things like bounce rate, time on page, session durations along with new browser information such as screen size suddenly all became possible. The click-stream still recorded the movement of a user between documents, but it leveraged the user’s browser to achieve it (rather than the server).

This was a major breakthrough but it was appropriate given the changing purposes of websites. Rather than websites being brochures online, websites were becoming shops… and applications (such as forums and trading sites like eBay). This change in the way we used the Internet also spawned the development of a new series of technologies designed to facilitate a more lightweight transfer of data between a user (client) and the server. Not only would this speed browsing experiences up, but it would also take an enormous computational burden off the servers themselves (simply by reducing the size of the requests).

The things we cared about, such as a reply to our message or the change of an item’s price on eBay represented a very small portion of the overall HTML document. So instead of needing to reload the entire document, we were increasingly only reloading the portions of the document that had changed (often just a single line or value). AJAX and Javascript languages offered a way for websites to achieve this, by transferring small packets of data within the page itself.

Google Analytics quickly added support for these in-page changes of state using a new report called “events”, but the fact that user data is not allowed inside Google Analytics largely renders this function useless. Events in GA can be used for goal tracking and simple reporting, but not much more. This was not a problem for click-stream analytics at first, until Javascript took over most of the web in the form of Single Page Applications (SPAs). Yep, I’m talking about Angular and React, along with back-end frameworks like node.js.

Suddenly a single document became an entire website.

Event-streaming analytics

As websites quickly evolved into web applications a document based tracking method was no longer adequate. Users could spend 10 minutes on a single page clicking, liking, saving, posting and otherwise interacting with the page despite the document path not changing. Think of Facebook.com’s home page news feed for example, as a user scrolls through their uniquely personalised news feed liking and commenting on posts, they are creating an enormous amount of data, but the document path is not changing. In this scenario, a click-stream based analytics tool like Google Analytics would simply report time on page of say 10 minutes with a 0% bounce rate.

The fact that all of these sites run on Javascript technology means that all of these interactions are already being recorded using javascript events. Amongst other things, these js events are used by the application to change what the user sees (the DOM) and how they interact with content whilst avoiding reloading the entire site itself. These events trigger client side or server side functions and can transfer data… again, all without requiring the page to be reloaded.

For example, imagine a user finding a product on an e-commerce website, adding the item to their cart, and commencing the checkout process. The event-stream would look something like this. Notice how it’s sitting on top of the document (server logs) and click-stream.

You can see from the diagram above that the click-stream is really not that useful at all. So, now we arrive at the present day.

Event-based marketing today

Today, javascript events are running most of the front-end and back-end of websites and applications, but marketers… and product people… and even engineers sometimes just aren’t using them. Or are they?

In 2017 Facebook announced the transition away from the 8 default ‘conversion pixels’ to a new ‘smart pixel’. You’ve probably guessed it, but the old conversion pixel was a server log based analytics method, much like the 1px by 1px image contained within emails to record email opens. The old Facebook pixel would trigger a request to a document on Facebooks server, located at a unique path where it could then be mapped as a conversion. The conversion count was simply the number of times the document had been requested on the server. The ‘smart pixel’, again you’ve probably already guessed it, is a new javascript based event tracking script. Javascript events would fire, sending a record of an event having occurred back to the Facebook conversion server. The number of conversions would be equal to the number of times the event had fired. This matters because marketers are interested in who triggered the event, not just the fact that an event was triggered.

So, that leads us to the present day. Any marketer using Facebook advertising is already using ‘events’ to measure conversion (they just don’t know it) and the worlds number one analytics platform doesn’t support personalised event tracking. Can you see the opportunity?

What is an “event” in the context of event-based marketing?

Definition

An event is a record of a single instance of behaviour taking place on a website or app. It’s pinned to a single identity (known or unknown) and we can store useful information inside it.

Think of an event like a box that you can store useful information inside of. It gets automatically created when a user performs an action or a task (a behaviour) you are interested in observing and measuring. This box can then be stored, and sent anywhere you desire. If you want your email marketing platform to know about the event, send the event there. If you want to serve the user an advert in Facebook based on this behaviour, send the event there. If you want to create a report of the number of times this event was fired, you might send a copy into Amplitude to help generate the report. If you want to send a push notification a week after the user triggers the event, send it to Vero.

Anatomy of an event

There are three parts to every event:

1. An identity (even if it’s anonymous)
2. A unique event name
3. Attributes stored inside the event.

Again, an event is a single instance of behaviour with information stored inside it. People trigger events, which means we can attach an identity to the event. As marketers, this means we know what happened, when it happened and who did it.

Why do we use the phrase ‘event stream’?

As users move around our website or app, clicking and interacting with things, they create a stream of data. This stream describes the user’s actions in chronological order. The stream really matters because it helps us perform what I can only describe as ‘reactive marketing’. We can react to users actions in real time by creating conditional rules. For example: if a user triggers, “add an item to cart” and does not trigger “started checkout” you may wish to commence display advertising containing the products the user just added to their cart. The fact that events happen in a particular order means we can more easily build conditional sequences around our user’s behaviour.

Stateful and stateless data

Lastly, it’s beneficial to know about the differences between stateful and stateless data. This is a very technical concept, but it fundamentally underpins event-based marketing.

Stateful data allow us to create, store and read the memory of the data generated in the lead up to a stateless change in our database (like filling in a form).

Stateless data holds no memory of the past, it merely contains the most recent version of a record (such as the information submitted by the form).

Most marketing platforms were built for stateless data streams. This means that marketers can only react to stateless changes in the database, such as a user creating an account, adding an item to their cart or making a purchase. However, to build a truly customised event-based marketing program, we must be able to trigger marketing automation on the smallest changes, and these can be easily measured using stateful data points.

Right person, right message, right time

The post What is Event-based Marketing? appeared first on Will Egan.

Let’s explore data collection methods for the effective measurement of how our products and campaigns perform.

This is a huge topic, and very few people truly understand it.

All online analytics packages rely on two types of data collection:

1. People
2. Behaviour

To track people online, we use cookies, sessions or other stored identity data points. To track behaviour, we use two primary methods of data collection: click-stream data, and event-stream data. The latter (event-streaming) is poorly understood yet offers the greatest value.

In an age where the word ‘personalisation’ is thrown around like glitter at Sydney’s Mardi Gras, event-based marketing is the backbone of this technology. Very few marketers understand the event stream, let alone realise that it probably already running on their very own website or app. For example, Facebook’s marketing platform relies on the event-stream.

Tracking User Identities

In order for data to be of value to the marketing and product development processes, we first need to be able to identify the unique people (both known users and anonymous users) on our app or website.

There are two common methods to track and identify a user: Cookies or Identity.

• Cookie-based identities which allows us to target a web browser or device.
• Identity-based identities which allows us to target a person.

Most websites use both. But from a technical perspective, identity-based marketing is far superior.

Cookie Based Identities

When you visit a website, that website will store a number of cookies in your browser’s memory or storage.

A cookie is a small file, which contains information (key pairs of data) served by the web server and stored in your web browser’s memory as a text-like file. This file contains useful information that allows the website (or third-party script on the website) to remember who you are, your preferences, or other behaviour for future reference should you visit again at some stage in the future.

Cookie-based marketing also allows a browser user who has been on your site to be marketed to by an advertising service. If you have a third-party script like Google Analytics’ analytics.js installed on your website, you can set it up to also talk to Google Adwords. Google Adwords will then know if this user hasn’t converted in a way that you’d like them to, and will continue to market to that user across its network.

When visiting the AFL.com.au website above, over 30 cookies were dropped on my browser (many of which were third party).

So, cookie-based marketing is useful for tracking users on your own website, with the intent of re-marketing to that user across an advertising partner’s ad network.

The problem with cookie-based marketing is that you are limited to tracking activity on one browser and on one device only. There are also huge limitations around cross-domain cookies authorisation, and the threat of the user clearing their browser’s cache (and your cookies along with it) at any moment.

Identity Based

Identity-based marketing however is built on the presumption that you, or someone else, knows who the user is (i.e. the user has a unique user ID). No matter where the user logs in and on what device, you know who they are. They’re more than a static web browser.

In your own product you have a system of keying user IDs (if your users can log in, you will), you will be able to identify this user uniquely and begin tracking every time they ‘do something’. We’ll get more into how that works (Javascript event-triggers) in the next section.

Identity-based marketing will always be superior in targeted reach. Firstly, because we can be much more confident about exactly who we’re tracking. Secondly, because all of the behavioural data being collected is being linked to them directly, not a browser.

In fact, cookie-based marketing if often associated with the idea of ‘anonymous IDs’, as although the browser needs a string of characters to identify it in code, there’s a strong chance that the user is still anonymous.

So what exactly is being tracked?

Identity Based Advertising

Because we’re always logged in to services like Facebook and Google, they provide a simple example of identity-based marketing. Have you ever noticed when you visit a website that Facebook is able to tell you how many of your friends have ‘liked this page’? It’s because you’re always logged in, they know the website you are on, and they use their javascript to determine (‘identify’) that you’re there.

Google takes this one step further by getting users to log into the Google Chrome web browser itself. This enables them to track movement and behaviour across every website you visit regardless of whether there is any Google javascript sitting on the websites you visit themselves. Whether Google does this or not is unclear, but technically it’s possible.

Most of the events that you will track against browsers (cookie-based) will be clickstream events, with some select event stream events (if at all). Cookie-based collection is often more clickstream-oriented purely because you are more interested in metrics about a user’s session, rather than the user’s profile. Identity-based marketing platforms will often put more effort into tracking the event stream because they can record engagement to a specific user, and engagement with specific features on a page far more easily.

For instance, Google Analytics largely tracks clickstream events, as it is primarily interested in statistics about your session or your client (by the way, “client” means the environment you are on: web, iOS, Windows 10, Android etc). Although it does also allow you to track event stream events in a limited way, you can’t feed a User-ID into GA along with those events because GA is not set up on the basis of user ‘profiles’. It merely groups events by User-ID for some better accuracy, with these user ids actually being browser ids stored in cookies.

In the same way that Facebook and Google can track users as they move across the web, you can use identity-based marketing to track your users as they move across your app or your website.

Now that we know how to track our users, let’s take a look at how we track what they do (their behaviour).

The Event-stream

Let’s start by defining what an event is.

An event is a packet of data that is sent from one location to another when ‘something happens’. It represents a single instance of behaviour taking place on a website or app, it is usually pinned to an identity (even anonymous identities), and the packet of data can contain whatever specific or contextual information you would like to include.

The Click-stream

The clickstream is the recording of a user’s interaction with a website and or other software application in the course of their browsing. The click stream is simply a record of ‘clicks’ or request between documents on a server. Therefore, the most commonly tracked clickstream data is page views.

The event stream is far broader in its application, and far more granular. Where clickstream events are generally collected in the normal browsing of a website, event stream data is specifically defined by the marketer or product manager based on what they would like to track. An example would be the event: ‘User signed up’. This is not browsing activity, so it has to be sent differently using Javascript.

Put simply: the clickstream tracks changes between two pages and the event stream tracks changes in state on a single page.

Choosing your preferred method of tracking all comes down to what you want to track.

Identity-based event tracking uses the event stream because the events are often reflective of a user’s behaviour across a long period of time. With an identity system, you can log these events to a unique user’s profile.

A simpler example of an identity platform is Intercom. If you wish too, you can install the Intercom tracking script on your website. Intercom will begin tracking clickstream events of users, but also event stream events from the same client or other clients. This isn’t cookie-based remember – the script on your website is sending the information directly to Intercom, and if you have a User-ID system, Intercom will record all of the events it receives against the user ID or user profile that triggered them.

In summary, cookie-based identities are designed to assist with click stream behaviour tracking. Whereas identity based javascript identities are designed for event-stream behaviour tracking.

So, how is all of this useful? Behavioural marketing.

The Right Message, to the Right Person at the Right Time

Okay, so now that we have a general understanding of the difference between cookie-based data collection and event-based data collection, it’s time to explain how all of this becomes useful.

To do this, let’s extend the concept of behavioural marketing.

There are many platforms, tools, technologies and people out there who offer up versions of a definition of behavioural marketing. But what is it exactly?

In our opinion, behavioural marketing is best described as:

1. Getting the right message
2. In front of the right person
3. At the right time.

Tracking user identities gives you the right person, tracking their events gives you the right time, and your own creativity (and the context) will help you craft the right message.

If we know who a user is, and we know what actions they are taking on our website or app, we can communicate with them in a much more meaningful way.

Let’s take a moment to explore a quick case study.

Case study: Spotify

Imagine you are responsible for the marketing of the Spotify music application. Spotify is about to release an app for the Apple Watch in an attempt to increase retention on the iOS platform. Your job, as the product marketer, is to announce this new service to your customers.

Let’s use the ‘right person’, ‘right message’, ‘right time’ framework to solve this problem.

First, we need to determine the right person; exactly who should we advertise the apple Watch App to (let’s build a segment, rather than promoting this to all of our users).

The good thing is that you have been tracking users using direct javascript identites along with eventstream events. At spotify, a generic ‘New Session’ event is triggered every time a user starts a new session on Spotify. One of the attributes sent in this event is the device type and operating system. It looks something like this:

event.newSession {
	deviceType: "mobile",
	operatingSystem: "iOS"
}

When you call the identify method and pass through these attributes, the user object is updated with new ‘traits’. Traits are the descriptors of the user, the things about them that make them unique.

Let’s head over to our event-based marketing platform, Vero, and run a simple query agains the user base to help us produce generate the segment. We can do this by selecting all users who have triggered the event ‘newSession’ where operatingSystem was equal to “iOS” and device was equal to “Watch”. Great, we have found ‘the right people’.

Next, let’s consider how we might construct the ‘right message’. The definition of “right message” is the one that converts the highest. The trick here is that because we can be confident about exactly who the audience is, we can be far more targeted with our message.

A message like “Spotify now available on smart watches” is unnecessarily vague for this audience. Whereas, ‘Install Spotify on Your Apple Watch’ is better, and we can do because of a higher sense of confidence that no one will ask about other watches, seeing as these are iOS users. However, this message still lacks the punch required to drive conversion (more on this later).

Seeing as we know who this person is, we also know their playlists and what song they just listened to. We can now incorporate this into the message. For example, “Listen to Avici Levels on your Apple Watch.”

As you can see, a whole new level of personalisation is available when tracking users in this way, and marketing to them accordingly.

Lastly, we need to decide upon the right time to show this add.

For every communication (email, text, display, push) there are two ‘right times’, one you’re in control of as the sender and one you’re not (when it’s opened/received by the recipient). But with event streaming, we can actually control both. Here’s how.

We can use an event to fire this communication at the right time. For example, we could use the ‘newSession’ event to fire an in-app pop-up message containing our advert. In this scenario, the user would open the iOS app on their smartphone and see the pop-up advert instantly.

Or, we could fire this message the first time a user triggers the event ‘newSession’ after having triggered the event ‘playsSong’ where the song’s artist was equal to ‘Coldplay’ and the time of day is between 5pm and 9pm in the user’s timezone.

This is the level of personalisation event-streaming provides behavioural marketing campaigns, and this is why you should build event-based marketing into your Growth Machine.

Getting Started with Event Streaming

Yes, setting up event stream events and tracking users will certainly require technical expertise. But, the good news is that this technology has been around a while and there are lots of tools that will make this easier for you, so long as you understand the concepts and can navigate your way around your app or website’s codebase (message from the future: learn to code).

The post Data and Marketing appeared first on Will Egan.