In Part 1 of this two-part series, you stepped back into cookie history … from their invention at Netscape in 1996 to when the EU’s GDPR legislation signed their death warrant in 2016. In Part 2, ahead, we’ll get past the problems – including those that persist in the new server-side data collection and tracking model – and focus more on solutions. Solutions you can adopt today that turn the diversity of data into an asset for your business.
(Which means Part 2 is more salesy, since one such solution is our product. We’re all about full disclosure here at Tilores.)
What you’ll discover is that connecting the different snippets of data users leave behind as they wander across the web needs more than just understanding data types and formats. It’s not just a change in technology – but a change in thinking.
The tyranny of the Tag Manager
Back in the Ye Olde Worlde of Cookies, tech-savvy marketers used TMSs – Tag Management Software – to pass data between their sites and the ad-serving technologies, analytics applications, and social media platforms they’d formed partnerships with.
The “tags” we’re talking about here are not HTML tags, or meta tags you write for search engines. A tag is a tiny chunk of code, often a single pixel “called” when an end user interacts with the website – enabling the site to keep track of that user as he or she journeys through the site and beyond. Those long strings of random characters that magically appear after a simple URL? They’re players in the tagging game.
For a while, tag managers were awesome. They gave marketing professionals a very useful web facility without requiring them to be coders; they’re incredibly versatile, handling tags of countless varieties from simple counters to bloodhound-like tags that follow users across devices. And as they developed, standards and conventions arose, so their community of marketing people could adopt a Tag Manager easily or switch to a different one without too much pain.
But to collect data, Tag Managers used methods that looked a lot like – guess what? Cookies. So they rose and fell on the same curve. But Tag Managers created a bigger issue: across the martech world, many people couldn’t imagine doing customer tracking with anything but tags. Tags had become the main, even the only, way to follow the breadcrumb trail of user behavior across pages and sites.
Cookies and tags had become accepted wisdom. So when browsers and laws started limiting data collected through them, marketers were all at sea, looking for alternative technologies that did the same thing. They were looking for evolution, when what they really needed was a revolution. Or rather, a *resolution*.
The resolution will not be televised
Consider two conceptual frameworks for building a data profile of an individual. We’ll dub the first “structured” and the second “unstructured”.
The structured view is that used in both cookie and tag manager paradigms. It looks at what links your different bits of tracking data, and expects those chunks of data to follow consistent, repeatable formats: an email address here, an IP address there, perhaps a Firstname-plus-Lastname or telephone number.
These are all good data points. And if you’ve got one common piece of uniquely identifiable data across all the datasets you’ve gathered – like an email address – it’s a pretty good basis for linking those datasets, and assuming they all relate to the same person. But for them to work smoothly, that key piece of data has to be what you expect, where you expect it. A partial email address, or a strangely formatted ZIP code, can fox it.
And the trouble is, you’ll write off a lot of data this way.
Because people can have more than one email; they may write their names in more than one way; the things they do on their owned devices may be completely different to the things they do on their work ones.
So the structured, consistent approach can lead to a partial profile that looks more complete than it is. It’s also a bit of a sledgehammer-to-kill-an-ant approach, since it needs a lot of computing crunch to scour endless gigabytes of records.
The second approach is less structured – and deliberately so. It builds on the structured approach by adding other datasets that aren’t normally used to build a detailed customer profile – using intelligent software to make educated guesses on how likely two datasets are to be related.
What kind of other datasets? Geographical data is one such, and an exceptionally useful one. Because knowing where an interaction takes place adds a rich backdrop of information to a customer touchpoint.
This alternative approach – expanding the universe of relevant data, even if that data tends to be messier and less structured – is the one we use at Tilores. It’s called entity resolution.
How inconsistent data can deliver consistent results
To understand entity resolution’s usefulness, think of the confusions in most tracking data. Your PC in the den contains a wealth of tracking information … but it’s largely valueless to marketers, because that data represents your family of five using it, each with very different habits.
Looking for consistent connections in that mass of data (like the way they share IP addresses) will certainly give you plenty of information … and it’ll almost certainly be useless.
Say the audience you’re building gains fourteen women named Jones. Traditional marketers might be overjoyed at increasing their prospect list by 14. But are there really that many individuals? You’ve got an E Jones. An E P Jones. An Ellen Jones. And a couple of Ellies. In reality, they’re all the same person.
The entity resolution approach – comparing that data against other data, in an intelligent way – can shake the real Ellen out of the mix.
Maybe eight of those entries share a geographical location. Or six of them visit the same website. Or two have the same birthday. By looking for what supports and what opposes the notion all 14 are the same person, the singular entity – Ms Ellen-Paige Jones of West Valley City, UT – can be resolved with a high degree of certainty.
Yes, using inconsistent, “unstructured” approach means more nuances. Not every question has a strict right/wrong or yes/no answer. There are shades of gray. But if you take the right approach, those shades of gray can be critically assessed for their level of confidence
This is entity resolution. And it’s ideally suited for cookieless marketing. Because with those datasets moving onto the web, entity resolution provides an ideal way to sift and sort the data that isn’t computationally expensive.
(As the size of the dataset increases, entity resolution methods require a linear rise in processing power – not an exponential one. Which means resolving an entity takes the same effort each time. Which means you can do it in real time, a huge advantage by itself.)
That’s why we believe at Tilores that entity resolution – rather than an evolution of Tag Management Software – is the best bet for large-scale marketers seeking to make sense of ambiguous tracking data.
Entity resolution lets you build high-certainty, real-time, self-cleaning data profiles about your customers. And the move of tracking data to server-side, where it can be compared and contrasted with other datasets, is a huge opportunity to get to know them even better.
CONCLUSION: entity resolution makes cookieless an opportunity
The web has undergone countless paradigm shifts in its short history. We believe the move to server-side tracking data is another one – but not enough marketers are doing it right yet.
Cookieless tracking data is a near-perfect scenario for entity resolution – and because it scales proportionally, it maintains performance and cost-effectiveness even as your customer list grows from thousands to millions.
If that sounds of interest to you, come talk to us.
