We recently changed the headline of our home page to, “Event Tagging is Dead,” in order to better reflect the game-changing nature of Convizit’s next-generation user behavior data capture. We love talking to customers and hearing their amazement when they “get” that the manual event tagging they have been suffering through for years is no longer necessary.
A Brief History of Website Event Tagging
For decades, companies have been trying to capture detailed information about what users are doing on their websites. An entire industry of tools and solutions have arisen to address this need – server log analytics, user behavior analytics, funnel analysis, product analytics, heat maps, session recording and so on.
Yet, after so many years and so many tools, it remains extremely difficult to simply get a complete list of every action every user performed on a website, in a properly structured and clearly named data stream, that can be used for analytics, testing, personalization and other business purposes. In fact, the state of the art today – for someone interested in more than just pageviews – is still manually selecting and naming a subset of specific on-page events to utilize. And except for the most advanced tools available, one still has to actually modify the code of webpage UI elements to be able to capture clicks on them (whether via a developer adding code to the page or by using a tag manager to define events).
This process is typically called “tagging events” or “instrumenting events” and it is an extremely difficult, time-consuming and error-prone undertaking. It also results in very partial data, as only a limited number of pre-selected events are tracked. Furthermore, because websites frequently change, this is a never-ending effort. And, worst of all, manual event tagging is fragile and frequently breaks.
This is no industry secret; we hear from customers and prospects time and again that implementing extensive event tracking on large websites is a huge challenge that initially takes months and is actually a difficult process that is never perfect and never ends. One of the largest website analytics tool vendors even mentions clearly on their site that, “maintaining intricate implementations on websites in a constant state of flux is a heavy burden” (source).
Isn’t it crazy that, in 2020, this is still how website event tracking is being done?
Self-Driving Automation is a Huge Challenge
Well, one could ask the same question about self-driving cars. After all, well-known companies have been telling us for years that full-self-driving cars are just around the corner. But here we are in 2020, and all of us still drive with our hands on the steering wheel (even in the latest, fully-loaded Teslas, this is a stated requirement!). Why? Because developing full-self-driving technology is a huge challenge; the world will get there, but there are still major challenges to overcome. (If you’re interested in this topic, you’ll likely appreciate this recent CNBC video about what’s taking Tesla so long to get there.)
OK, so I’m not trying to make the case that automatically capturing, naming and structuring every user action on a website is as difficult as developing fully self-driving cars! However, the truth is that the technological challenges involved with smart, “self-driving” website event capture are truly not trivial. Which is why, in 2020, achieving complete and granular user behavior capture is still such an expensive, difficult and fragile undertaking.
A Look at Two Previously Attempted Auto-Capture Solutions
Let’s take a look at two attempts by others to usher in a new generation of automatic website event tracking. One of them is a product analytics tool that offers an auto-capture feature that the company highlights as a major differentiating factor. Their website states that their tool, “is the only tool that automatically captures all user data on your site, from the moment of installation forward. A single snippet grabs every click, swipe, tap, pageview, and fill – forever. There’s no need for manual tracking. No need to choose what to measure. No need to wrangle engineers to write code.”
Without getting into a debate whether or not one can actually “wrangle an engineer,” the other two advantages they mention are at the heart of the matter: there’s no need to choose what to track, and there’s no need for manually implementing that event tracking. At first glance, this sounds like the breakthrough everyone has been waiting for: track and analyze everything, all the time, with no manual effort. Unfortunately, it’s what they don’t mention that is problematic with their approach.
It is true that they capture every user action automatically, without specific event coding. However, that data is not available for use until someone actually does choose to analyze a particular action – and then someone has to manually select the action and type in a name. In other words, this approach simply pushes the major effort to a later stage: it saves the manual coding effort, but still requires the manual tagging effort for each and every event. So, it’s a step forward, but it’s hardly a self-driving solution for being able to capture and utilize comprehensive user behavior data from websites.
Another serious shortcoming of this approach relates to tracking continuity – as websites change, so do the CSS selectors on which this approach relies to identify individual UI elements. When the CSS selector for an already selected and named UI element changes, event data for that element will no longer be available. And even if someone manually selects it and names it again, it will no longer be connected with its previous incarnation. The result is that many analyses will end up being incomplete, and much of the data out of date, disconnected and unusable.
Yet another downside of this auto-capture technology is the lack of event context (event properties) captured with each action. For example, it will capture a user clicking on the Remove from Cart button – and, if someone selected and named this event, it will appear in the analytics data. However, the system will not automatically associate with this event important details, such as the name of the product that was removed, the quantity that had been in the cart, its price or selected color. Again, one can add this information to events using manual processes, such as writing code to call their API, but it is hardly a self-driving system.
The second attempt was by another product analytics company. This company invested considerable resources in developing a feature within their analytics suite, which they named “Autotrack.” In an August 2016 press release announcing the feature, they described Autotrack as, “An innovative feature that allows customers to automatically collect every action on web and easily analyze that data without a developer.” They clearly highlighted the problem they were trying to solve by stating that, “Traditionally, companies have needed to plan out their entire analytics strategy up front and allocate a developer to manually write code for every action they want to track.”
However, about three years later they decided to abandon this feature, for a number of reasons that they outlined in a blog post. I will summarize the reasons they mentioned for reverting to the manual-only method of instrumenting event capture:
- Because there are multiple ways that a user can perform the same action, auto-capture data won’t accurately reflect what are essentially identical user actions. (What they mean is that, when manually tagging events, identical keywords in names can be used across events to allow them to be analyzed together.)
- Relevant on-screen context is not captured as event properties along with auto-captured user actions. (When manually coding event capture, this context can be programmatically added to events.)
- Minor changes to on-page elements (e.g., testing a change of text on a button) quietly breaks tracking continuity and requires lots of manual maintenance work. (This refers to the “tracking continuity” challenge I mentioned earlier.)
- Generated data is difficult to use, because it is not logically named, and will require software or data engineering resources to make it usable.
- By collecting everything, it is inevitable that the auto-track code will collect personal/private information and violate security/regulatory requirements.
Of course, some of these issues are much larger than others, and not all the issues will affect different data consumers in the same way. But, what if “self-driving” website event auto-capture technology could overcome all of these challenges?
Introducing the Next Generation of User Behavior Data Capture
When we boldly claim that “Event Tagging is Dead,” we mean that technology has advanced to the point that all of the above-mentioned challenges have now been successfully addressed! A team of some very smart people from the worlds of academia, military intelligence and the private sector joined forces three years ago to solve these formidable challenges. By incorporating cutting-edge machine learning technology, algorithmic heuristics and a number of other disciplines, our software essentially understands webpages in the same way that humans do –without relying on CSS selectors or human effort.
We are proud to announce that full-self-driving website user behavior data capture has now been achieved. Specifically, Convizit has developed the world’s first website event auto-capture solution that:
- Deploys in minutes and requires no further manual effort
- Automatically captures every user action in every webpage
- Automatically enriches every event with relevant on-page properties
- Automatically groups functionally identical on-page elements, even when there are minor differences between them in terms of appearance or code, and even when they appear across different pages
- Automatically ensures the continuity of element tracking over time, even when changes are made to elements or the page structure surrounding them
- Delivers a continuous stream of pre-named, pre-structured, ready-for-use user behavior data to any third-party tool, platform or data warehouse, to instantly enhance the business value of existing product analytics, BI, CDP, marketing automation and personalization tools
- Provides multiple layers of data privacy protection, to ensure that no PII is ever collected, transmitted or stored
- Has no impact on end-user experience or page load times
Convizit doesn’t bring to the table an analytics, BI, data warehouse, personalization, marketing automation or CDP platform; there are no shortage of such solutions on the market. Instead, we are singularly focused on automatically delivering complete, ready-for-action user behavior data to fuel and enrich all the tools and platforms that utilize this kind of data.
We are extremely excited about what this means for the industry, and hope you will contact us to try it out yourself.