Personalization has always been one of the major talking points in Web Content Management – with an ever escalating arms race between vendors who have put massive investment in making personalization more complex, more powerful, more rules, simpler, faster, better informed, “now with even more customer data!”, etc. etc.

It could be argued that this one area has taken the majority of the cycles in product investment in the last few years. Adobe integrated AEM with Audience Manager (a DMP via the Demdex acquisition) and parts of the Omniture suite. Sitecore built the Online Marketing Suite (OMS), which was replaced by the Digital Marketing Suite (DMS), which was replaced by the Experience Database (xDB). Acquia did the same with their acquisition of TruCentric and additional iterations of Lift. Episerver bought and integrated Peerius, and now Idio – all in an effort to try to outflank opponents on that all-consuming personalization front.

So, Gartner must have sent some waves into that space recently when they boldly proclaimed “80% of Marketers Will Abandon Personalization Efforts by 2025“; certainly at least one DXP vendor CEO bristled publicly at the claim. We also see Personalization Engines near the nadir in the “Trough of Disillusionment” in the Gartner Hype Cycle for Digital Marketing and Advertising, 2019.

Here are the elements I agree with in the Gartner report, and those I disagree with. You’ll notice I asked if the hype was dead, not whether personalization is dead – because crafting an customer experience aligned to their tasks is still important, but many of the superficial reasons for implementing personalization simply don’t resonate with either audiences or marketing teams trying to service them.

It is a pretty well-known fact at this point that most customers have massively struggled to implement personalization on DXP platforms. This is due to a number of causes. Tony Byrne of Real Story Group gives an excellent example which we can use to parse through the issues:

…when an agency executive declared, “I want the owner of a car wash in Rhode Island to only see content tailored to them,” it fell to a plucky underling to explain, “Uh, we actually don’t have any content for that person.”

Tony Byrne, Real Story Group

So let’s break down the main reasons how we got into this state of things:

Lack of business justification

In the example given by Tony, what would be the value of trying to only serve content relevant to [business owners], [car wash] in [Rhode Island]? Are they trying to minimize distractions of other non-relevant content? Does that particular micro-segment have content needs that are unique to the exclusion of other content? (Likely not) Also, that persona also probably implicitly has other intersections that might be actively excluded if you personalize down to that initial, primary, micro-segment; [personal taxpayer], [family provider needs], [health care needs], [citizen needs such as fire, water services], etc.

By trying to segment too granularly, you risk alienating your target audience by aggressively de-prioritizing other content and providing too little content actually aligned to the underlying (in this case, unspoken and poorly articulated) business need or customer success metric.

Similarly, with this increasing granularity and trying to capture and tag every interaction, personalization systems in DXP vendors are often extremely expensive to run. Tony Byrne has an informative graphic outlining the labyrinth of various databases, indexes and processing services needed to run the simplest of scenarios in Sitecore.

Doing the wrong thing

There is a special case of the above example, where you not only miss the mark and do something irrelevant, but you use personalization in such a poor way that it actually is off-putting to your target customer. This always results from an “inside-out” view and not putting your customer first. These happen so regularly, there are countless examples of “personalization gone wrong” or even “creepy marketing”. 

I saw a great example recently; Kristina Halvorson from Brain Traffic (a well-regarded content strategy agency) posted this tweet:

In this case, Delta Airlines has taken her company name (using a 3rd party service, mapping her ISP network address against an Account-Based Marketing database) and put it right in the hero call-to-action. However, if you look at the underlying decision that went into this gratuitous use of personalization, you have to ask: “What was the Delta marketing team trying to accomplish? Would using the company name increase engagement, and to what end?” If they took a customer-centric view, it may be that using information known about her business travel (i.e. from their known purchase pattern data) they could craft offers specific to her company usage patterns – but in this case, it was lazy, not aligned to making her experience better, and not actually personalizing anything (as you can easily see any business name being pasted into the call-to-action) which isn’t actually personalization at all. It was not designed to provide any particular context or service of helping Kristina achieve her goals in any way and thus, falls into the category of doing the wrong thing.

Lack of content (and appropriate processes)

This is another glaringly obvious issue in that scenario – often organizations simply don’t realize the scale to create that content needed for this segmentation, so it simply remains undone.

Any personalization effort means a lot more process than you might think. In some cases, it means having a framework in place to actually understand your content strategy – which is always a good idea, but usually a personalization effort also reveals gaps in your overall persona or customer experience approach and these need to be addressed first. Nick Barber from Forrester has a great report on how your content foundation is key.

Second, often many personalization systems may lack API extension points and rely on “side-loading” content outside of a CMS, which may add additional logistical overhead and maintenance as you have to maintain and copy content across systems.

Lack of scale

By scale, I mean having such a large corpus of unique audience segments and content as a starting point. If you are in the rare position of actually starting from a point of having too much content (from a problem of having too large a mandate or overlapping personas) then personalization may be effective as you have scale, content and business justification. Examples would include sectors such as healthcare, insurance and large government institutions or “mega-sites” where you need to quickly determine intent, and minimise non-relevant content. For example, someone looking for personal car insurance does not need to see corporate or re-insurance information in their browsing experience.

But the opposite is also true, if you are having to “invent” relevant content to meet the needs of personalization, that is a sign that you have an issue of scale, and in this case, personalization is probably not a business necessity.

More restrictions in the use of customer data

We live in an era where free services have necessitated the clandestine monetization of customer data and preferences to be used to drive other experiences, always keeping in mind the importance of good data management (visit this page to get all the details). According to the Yale Law School Privacy Lab, “More than 75% of the 300+ apps … contain the signatures of trackers”. As a result of this, and other practices, both governments and vendors have started to recently make privacy more of a priority. The European Union and California have instituted measures such as the GDPR and CCPA, which govern usage and transparency around customer data. Similarly, increasing privacy controls in iOS 13 as well as all browser vendors phasing out 3rd party cookies make it far more difficult to obtain the data which was previously used to infer customer interests and location-based intent. Visit https://www.venyu.com/office-365/ and get more information.

All this adds greatly to the cost of using 3rd party customer data to personalize – old approaches no longer work, and those that still do require far more oversight and compliance. Oracle is one DXP vendor that has shuttered an adtech acquisition entirely (and removed its DMP data augmentation capabilities) because of this changing environment.

Lack of ROI

Sometimes, you may actually meet all of the above criteria; in that you’re a product experience company, with an endless supply of content with great customer relationships and data – Netflix, for example. Even then, sometimes the effort involved in personalization is not worth the cost. Netflix famously ran one of the earliest data-scientist contests in order to crowdsource content suggestion algorithms. They paid $1 million dollars for something where ultimately “the additional accuracy gains that we measured did not seem to justify the engineering effort needed to bring them into a production environment.

Now, as the market for Netflix changed from DVDs to streaming, the ROI equation changed; the product was no longer logistics; categorization, queue management and shipping – it changed to the digital interface and experience itself. Patterns in the data and audiences became easier to maintain, but even Netflix found that more data was not necessarily better, and a simple “thumbs-up, thumbs-down” approach to rating was more effective for the ROI vs. a 5-star ranking.

In the case of WCM, often the personalization aspects of a DXP are more expensive (from a licensing perspective) than the core content creation features, and rarely takes into account the significant run costs and overall marketing spend to properly staff and run the program.

Where is the hype actually justified?

While personalization is falling flat in a marketing sense, personalization has been used extremely effectively in a product management sense – if your customer experience is a product, or being run as a digital product (with product managers, roadmaps, etc.) then you probably have many of the organizational structures and skillsets in place to be successful. You’ll note the article said “80% of marketers…” so, product and actual customer-centric teams (or any marketer with product and CX-like responsibility) don’t fall into the same category of those not getting the value from personalization. In other words, personalization is effective only when it is core to the experience across channels and is not just using gimmicky tricks to sell to a customer.

Fortunately, implementing product-based approaches is often far easier as you typically have far more implicit data from information such as opt-in, logged-in state, transaction history, etc. – in these cases, personalization is based on an actual transactional relationship, not the use of data inferred via page visit analysis or bought via other data services.

I am optimistic that WCM vendors and customers will start to realize that effective personalization, while looking fantastic in the demo, is no longer a differentiating factor. Ideally they take a step back and realize the irony that is effective content is not the targeting of unique micro-experiences but for brands to think broader, to tell effective, consistent, narratives. And in order to do so, they need to really look at tools around content strategy, creation and curation.

Again – the “Trough of Disillusionment” doesn’t mean companies aren’t getting value from those technologies (including personalization), it simply means that you need to have a clear people, process and technology strategy to take advantage of those elements and avoid falling for the hype – and there is no magic bullet (and that includes “AI” which is worthy of a blog post itself).


1 Comment

Delivering on the Promise of Personalization - Gene De Libero · February 24, 2020 at 9:18 pm

[…] his latest blog post, my buddy Mark Demeny, Director of Content Management Strategy at Contentful, discusses […]

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.