Matching Personalization Strategies to Business Goals


Share on LinkedIn

Personalization is the heart of digital. As a direct channel, digital rewards personalization in almost every form, and there are no shortage of forms. I’ve been walking through how to select the right place to focus when you start building a personalization program. At a high-level, the basic types of personalization I described were these:

  • Personalizing a Marketing Drive: Optimizing a specific drive within a broader experience
  • Personalizing an Experience: Creating a unique creative environment for a visitor
  • Suggesting Relevant Content or Products: Optimizing User Selection inside a stable experience
  • Tuning an Offer: Changing the parameters, size or definition of an offer based on the user’s value, price sensitivity, existing relationship or interests

I went up to this level of abstraction because, after mapping out a baker’s dozen of personalization strategies, I thought it would be too hard to sort out how to map those strategies to the business. By going up to a higher level and reducing the number of types, I think it was easier to get at the principles (cardinality, journey multiplicity, and offer tuning) that can help drive a strategy. But now it’s time to go back down one level and show how each of the personalization strategies fits inside these four basic types.

For convenience sake, here’s a list of the various personalization strategies I cataloged:  

Content/Product Related Personalization

Content-History Strategies

Next Best Offer (NBO) Strategies

Trigger-based Strategies

Last Behavior Strategies

Threshold Strategies

Time & Event Strategies

Offer Matching Strategies

Filter-Based Strategies


People-matching and Nearest Audience Strategies

VoC Based Strategies

User-Directed Strategies




So let’s run through each type of personalization and see which strategies seem appropriate.

Personalizing a marketing drive is probably the most common and easiest starting point for most personalization programs outside of ecommerce and publishing. For this type of personalization, you’re trying to find the right creative or the right drive to add to an experience. As you might expect, Next-Best-Offer strategies are core to this type of personalization. Trigger-based, Last Behavior, Threshold and Time & Event strategies are also highly appropriate for personalization of marketing drives. You’d use trigger-based strategies to tie personalization drives to visitors hitting particular events on the Website. Last Behavior strategies might be even more common in this context. For sites with any significant amount of repeat behavior, one of the most common and effective personalization techniques is simply to re-surface a marketing drive around whatever their last visit focused on. Last Behavior is popular as a strategy because it’s easy to implement and it nearly always demonstrates at least some lift. Threshold strategies are much less common but are equally applicable. On Websites with lots of behavior (like media portals), doing a single behavior isn’t necessarily much of cue for personalization. Threshold strategies simply up the ante a bit by requiring a visitor to view X pages or spend X amount of time on a tool or content. Time and Event strategies focus on finding seasonal or time-based patterns (like birthdays) which tie naturally into trigger behaviors. Where applicable, they nearly always are used for personalizing marketing drives not complete experiences. All four of these methods are natural ways to create rules around more personalized marketing drives.

When you need to personalize an entire experience, you’ll more likely focus on content-history personalization. By weighting all of a visitor’s viewing history, these strategies create natural direction for larger, experience-based personalization. However, there’s a bit of a conundrum here; it’s often hard to personalize a large experience when you need lots of behavior inside the experience to decide how to personalize it! People-matching and VoC techniques are both deep enough to drive significant experience personalization and can (at least sometimes) be deployed before a visitor accumulates lots of behavior. User-directed strategies are also common in this type of personalization. For experiences that a user spends significant amounts of time in, letting them make the personalization decisions is often an extremely successful strategy.

Suggesting relevant content or products has been widely recognized as the key to effective ecommerce and publishing  – both being business models with many, many different offerings. Content personalization, content history personalization, and crowd-sourcing are all common strategies for whittling down a vast universe of offerings to the things a specific visitor might care about. Content personalization relies on the idea that if a visitor is interested in this product or this story, there are relationships between that product or story and other products and stories that make them more likely to be of interest. The range of alternatives here is broad: one might suggest directly competitive products (brand X vs. brand Y), complementary products (buns with hot dogs), related products (charcoal with hot dogs), or even products related by demographics (luxury car, luxury watch). The same is true for content. Stories about the same topic (the SF Giants), the broader topic area (baseball), related topic areas (football), or topic areas related by demographics (UCF, Energy Drinks) can be used to personalize drives. It’s important to understand that the range of possible dimensions to connect products or content is vast. The types of dimensions I’ve listed above (competitive, complementary, etc.) are merely suggestive. An article about baseball would certainly be categorized as Sports, but it might also be classified along demographic dimensions like male and older, along lifestyle dimensions like sedentary or fan, by associated activities like gambling, by affiliations like team, or by geographics like SF or local or US. The complexity and potential ambiguity of taxonomy has put a premium on analytics techniques in this area that don’t require any categorization. But while categorization may be complex, it can help cull out key relationships in the data that may be very hard to discover with purely statistical techniques.

I’d also put filter-based personalization strategies in this group. Most filtering and faceting strategies are around content or product selection, so using filter cues seems to fit most naturally in this category.

Which brings us to offer tuning. If you have lots of offers, offer-matching strategies are appropriate. The key to offer-matching is that offers are classified along similar dimensions as visitors or content. In other words, you might create matching taxonomies of luxury products and luxury shoppers, electronic shoppers and electronics products, or office workers and lunch restaurants. By scoring visitors and offers along identical taxonomies, you have a matching strategy that should be able to pair visitors up to the most appropriate offer. As with content, however, there’s a less taxonomic approach that uses people-matching. If you know that I’m most like previous visitors X, Y, and Z you can pick the offer that worked the best for that group and use if for my visit. There are advantages and disadvantages to this type of approach. On the one hand, it removes the necessity of constantly building taxonomies and mapping both visitors and offers to them. Big advantage. On the downside, it requires considerable volume and sophistication in rotating offers and testing on both the visitor side and the offer side.

I don’t want this to seem didactic (in the bad sense of the word). Almost any of the strategies could probably be used in at least some cases for any given type of personalization. All I really want to do is suggest the richness of potential personalization’s and provide some basic direction about the strategies that might fit best.

With all of this in hand, in my next post I’ll try to put all three elements together and map out the relationship from type of personalization, through strategy and down into specific analytics tactics. After that, I may try to tackle some of the high-level technology gaps in personalization and the choices you have to make around black-box, rule-based or model-based techniques.

Republished with author's permission from original post.

Gary Angel
Gary is the CEO of Digital Mortar. DM is the leading platform for in-store customer journey analytics. It provides near real-time reporting and analysis of how stores performed including full in-store funnel analysis, segmented customer journey analysis, staff evaluation and optimization, and compliance reporting. Prior to founding Digital Mortar, Gary led Ernst & Young's Digital Analytics practice. His previous company, Semphonic, was acquired by EY in 2013.


Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here