For at least a decade, posts and pundits have pushed the notion that personalization is a key — maybe “the” key — to email marketing. Personalization in digital channels seems like a logical, natural goal. People say they like being treated “more personally”.  So, marketing content that is more personal should deliver better results. 

Since digital content is built of data, the idea of changing content for each person is conceptually straightforward. A given piece of digital content — email template, Word document, PDF, or whatever — can be set up as a template, filled in and finished with specific-to-the-person bits. It’s not unique to the digital domain; form letters and carbon copies go back years, and some of us remember “Print Merge.” But in theory, digital “stuff” is infinitely malleable and more personalizable.

It’s a really compelling theory, but in practice — is it working?

There are at least four issues with personalization practices du jour:

1. There’s no clear definition of personalization.
2. The business motives for personalizing are in conflict with the human needs that make personalization appealing.
3. The data and systems available for personalization are, in general, too myopic and limited to do the job.
4. The tension between privacy and personalization is bad and will get worse.

Personalization Is Undefined

Some months back I attended an online webinar featuring a panel of top-tier email marketers talking about personalization. In the Q&A, I asked if any of them could point to definitions for the term — preferably definitions with some detail and structure. The answer across the panel was “No.” 

An Amazon search for books about marketing and personalization currently comes up with only a handful of titles, written (for the most part) by marketers for marketers. A search for the term “personalized” yields no non-fiction, expert-level works — not a single result in business or science. By contrast, there are 71 pages of results on “obscenity”, and dozens on “ERP”.

I’d be delighted to be wrong, and to be informed that there’s a widely-accepted, research-based model that defines “personalization” and “personalized” in clear, workable, structured detail.  

A sample base of 4 experts is narrow, to be sure, and Amazon searches aren’t the end-all of knowledge. But to have a term that’s purportedly key to millions of dollars of martech spending and consulting — and marketing careers — just hang in space, as if it were self-defining and self-supporting, is cause for pause.

Without definition and structure, there’s no way to measure relative progress of personalization efforts. A statement like “our marketing is more personalized” is meaningless if “more” and “less” aren’t measurable. 

Doing more of an undefined (or ill-defined) thing is a measure of the effort, but not of the execution. Using someone’s name in conversation is “more personal,” but anyone who’s ever been annoyed by a salesperson using their name repeatedly knows that “more” isn’t “better.” The digital equivalent — increasing the number of first-name field merges in an email campaign — probably wouldn’t improve the campaign results.

No marketer with taste would pepper “[Firstname]” mail-merges in every other sentence, but that’s a matter of the marketer’s taste (personal/cultural knowledge), not a decision based on a clear rubric defining or recommending those limits. 

So if we don’t know what “it” is…how do we know if we’re doing it right, or better?

Business Motives vs. Personal Motives

The lead line from a recent McKinsey report on personalization reads:

“This Next in Personalization 2021 Report reveals that companies who excel at demonstrating customer intimacy generate faster rates of revenue growth than their peers. And the closer organizations get to the consumer, the bigger the gains.”

Peel back the business-verbiage wrapper and there’s something appalling about that passage. The phrase “demonstrating customer intimacy” is especially tin-eared; imagine saying that in the context of an actual relationship:

   “But, honey, I’ve been demonstrating intimacy.”  [exit right, door slams]. 

That passage embodies the conflict between the business motives for “using personalization,” and the root word “person” in the term. Is your business personalizing for the benefit of the person at the other end of that digital pipe? Or are you ‘demonstrating intimacy,’ personalizing solely for persuasion and business gain?

To be fair, every consumer who ever answered a survey about personalization is at least partly to blame for this conflict of motives. The earlier-cited McKinsey report says, a few paragraphs later, “Consumers don’t just want personalization, they demand it.”  No doubt there’s a citation somewhere in that report of a survey — XX% of respondents who were asked if they preferred personalized treatment said, why, yes, of course. 

What a consumer hears when asked about “personalization,” and what marketing orgs do when they ‘personalize,’ aren’t the same thing. 

If that person’s friend says “you’ll like my optometrist, great personal service,” the friend is not saying that the optometrist uses their name in every sentence. They’re saying the optometrist  remembers personal details that matter in the service they provide, e.g. that the friend scuba-dives and reads a lot, which bears on their prescription. 

In other words, to a consumer, “personalized” means “actually about me.” Marketing personalization, with rare exceptions, demonstrates (or fakes) ‘about me’ in order to engage or persuade. 

The Blind Alley of First-Party Data

Personalization, almost by definition, can only be as good as the data available. Consider that against the somewhat predictable sequence of data acquisition for a company’s email marketing program.

First, some person signs up and opts in — usually on a Web site. The data that interaction yields is pretty slim; email address, name, and (perhaps) one or two additional fields. Additional contextual data from the browser session that could be appended to the record, and used for personalization, is either discarded, or saved in some other system like website analytics. In most situations, the email marketing system, the website platform, and website analytics are separate systems. Full data from that fleeting interaction is split up and parceled out — that’s just “how it’s done.”

What the marketer “knows” about the new person in their list, as captured by the email platform, doesn’t amount to much — far less than what a savvy in-person salesperson “knows” from first impressions and observations.

Sure, that’s just the first moment of signup; the working theory of “first-party data” is that more data about each person will accumulate over time. 

That sounds sensible, but in reality, it’s usually a fragmented mess. 

Email itself returns very little data about messages sent to individuals. The “spy pixels” hyped by Apple and others (technically, actually HTTP, not email) were more commonly used for aggregate statistics such as open rates. Apple’s approach to masking that signal by “false opening” all emails has reduced interest still further. 

Even if “spy pixels” returned detailed data, with an average open rate of around 20% — only 1 person out of 5 opens each message.

Where the business objective of email marketing is to “drive clicks” — website visits — marketers hope that consumer click behavior would provide a carrier signal to build out first-party data. But that, too, is challenging in actual practice. Click-through rates are a fraction of the fraction of opens — 2% - 5%, according to Campaign Monitor. 2 to 5 people of every 100 happen to click a link; that’s hardly the basis for rich 1st-party data.

Two other factors make it even worse. One is pure math: churn. Sources vary, but “list churn rates” between 25% and 33% per year are commonly cited. That rapid turnover means that an individual is likely to stay on a list for less than two years. Munge all that math together with a hypothetical once-a-week mailing: the email marketer will have about 100 shots to score the single, ambiguous bulls-eye of a website click.

The biggest issue with the first-party data theory, though, is myopia. When the only measurements available are related to the offers and products of the business, so-called “first party data” is inherently narrow and slanted. The consumer that happens to click the link for shoes may only buy shoes once a decade. Interpreting that click as a signal to push them into “recent opens” and “shoes cohort” is an inherently narrow, constrained decision. 

Some businesses have made tighter data-links between websites, ecommerce and email, but that doesn’t exempt them from the curse of myopia. I’ve been a customer and subscriber of a major outdoor/sporting-goods retailer for over a decade, and have spent thousands of dollars with them. I get the same generic offers — usually what they have for sale — that everyone else does (Same experience with text-message marketing…).

Watching speeding traffic through a chink in the fence is a poor basis for “a relationship” with the drivers. 

Privacy and Personalization — End of The (1st) Party?

There’s always been a tension between privacy and personalization. The digital world ran out ahead for a couple of decades, accumulating data before we realized the degree of psychological leverage it would afford. We’re catching up (slowly, in the US), and starting to rein that in. 

In theory, a customer who really, really wants to give you a lot of information so you can do a better job meeting their needs can still do so. In practice — I think we’re all too tired of filling in forms and boxes to do much of that.

The friction on data acquisition is only going to get higher; the pace and accuracy of data accumulation will get worse. Can companies realistically expect to systematically accumulate and manage data meaningful enough to the customer to enable effective, systematic personalization of content?

No doubt there are organizations that have this puzzle solved. If that’s you — congrats, mind letting the rest of us in on the secret? 

Solutions Or Alternatives?

Viewed systematically — technologies, content and (critically) data — I’m not sure that the vision of mass personalization can be successfully and continuously executed by many companies. Data is simultaneously the weakest and most important leg on that stool. Privacy and behavior constraints, weak data-linkages between systems, and the basic scattershot inconsistency of consumer engagement are crippling. 

I’m not suggesting that personalization isn’t a viable tactic, when the components and cost-justifications are there. One of the better email-personalization experiences I’ve had in the past year was executed by a small company who leveraged (I think) tight links between Shopify shopping cart and their Klaviyo email platform. It didn’t actually work – I didn’t go back and buy the shopping-cart item — but it was a successful, tight personalization-in-action example (It was also handled tastefully — one extra abandoned-cart-item nag message might have pushed me to unsubscribe.)

The lesson to me from this look at the current landscape is to be considerably more skeptical of personalization as a unifying strategic approach for email marketing. Is a many-zeros CDP investment (for example) justifiable? Is the personalization feature-set in that tempting new ESP really going to be used?

As to alternatives…several thoughts. An actual research-based, human-centered definition/model of personalization will eventually show up (There’s adjacent academic research in areas like UX.).

What people want — “treat me like you know me” — and how they feel when they perceive that you’re achieving that is too powerful to ignore forever. I just don’t think that the linear 1:1 1st-party data approach to personalization is working that well, for all the reasons mentioned. It does seem like a big, fat target for an AI approach, teasing the personalization patterns out of that perpetually messy data seems like a promising pursuit. Next post, perhaps!

Editor's Note: Matthew will be leading a discussion on this blog post during our weekly OI-members-only Zoom Discussion on Thursday, March 30, 2023, from 12:00 Noon to 1:00 PM ET.

OI members, watch the discussion list for your link. Not a member? Join today and be invited to our weekly members-only Zoom Discussions as well as other benefits. Just $200 per year

Midjourney Image  - ‘human figure partially built with legos’