There is no single personalization technology category — and no single way to think about personalization in commerce. In fact, we need to reframe the concept of personalization as an adjective or a verb — not a noun.

Consumers want relevant experiences …

In commerce search, for example, we personalize product listing pages (PLPs) to prioritize products based on what we know about the customer (verb). The subsequent search results show a personalized product assortment (adjective).

But the act of personalizing isn’t a switch we flip. In fact, if we’re doing a really good job, the customer doesn’t realize we’re anticipating their needs and serving products based on their intent in the moment. Ultimately, we have many (potentially competing) influences on the order in which we display products to customers. It’s crucial that we understand all the ways we impact that experience, however invisible our controls are to the customer.

Previously, I’ve addressed the spectrum, from keyword matching in search results to deeply personalized, intent-based product search results. My colleague, Jessica Liu, has explained how invisible experiences will help us create more individualized moments, appropriate to each customer.

… and companies want to optimize experiences to help their bottom line.

More and more, I’m hearing a new request from my brand and retailer clients. Specifically, they want to serve products based on their own needs — not those of their customers. To be clear, this is not personalization!

What they’re asking for is optimization — ah, wait! Similarly, “optimization” isn’t a term we should use as a blanket “thing,” either.

Personalized product landing pages prioritize the needs or preferences of the customer, such as:

  • Products they’ve bought before and that are due to reorder.
  • Items that complement, or are compatible with, previous purchases.
  • Products with attributes like others they’ve viewed in their shopping session or that match their demonstrated intent.
  • Products they’re most likely to purchase, based on what we know about them and other customers like them.
  • Whatever will get them to complete a purchase as soon as possible.

Optimized product search results might prioritize the needs of the business, including:

  • The highest-margin products.
  • Overstocked items.
  • Inventory located physically closest to the customer..
  • Products nearing their end-of-season, expiration date, or similar.


Instead, balance personalized and optimized assortments

In a perfect world, we find balance between personalized and optimized product assortments, but this balance is difficult to strike. The systems we use may overindex on some strategies or even enable users to create contradictory rules. For example:

  • The search solution’s algorithm prioritizes products most likely to convert. Merchandisers then “boost” new products so they display higher in search results. The new products likely will take time to become strong sellers — if they ever do. However, they occupy crucial real estate and push high-converting items lower in the results.
  • Some search solutions use both AI and manual search tuning, which can work in opposition to each other. Merchandisers can easily — and inadvertently — tune results in ways that negate the automated work.
  • Some solutions force all boosted products to the top of results, “wallpapering” instead of “peppering” them in. This practice, too, diminishes the success of algorithms with a goal-based metric.

Finding the balance between optimized and personalized PLPs

  1. Prioritize zero-party data for personalization, above other metrics. When customers directly and knowingly tell us what they want, we should believe them and shorten their path to those products. Example: If they completed a quiz showing that the needs for their facial moisturizer are unique — such as wrinkles, plus acne — surface those few specialized solutions that they’re most likely to want over the highest-converting ones.
  2. Use optimization to “boost” within the personalized result set. This is a matter of philosophy, but it’s best to cater to customers’ needs and then fine-tune from there. It’s also easier to do so because most search algorithms are more focused on personalizing to customers than optimizing for businesses. So begin with a highly personalized result set, and then gently boost or bury within it to tweak the PLP to business priorities.
  3. Configure algorithms so that optimization doesn’t negate personalization. Many commerce search solutions enable degrees of boosting and burying, using numeric scales or sliders. Ensure that optimized factors are weighted less heavily than personalized factors.
  4. And, of course, always test, test, test! If your tech doesn’t include detailed A/B-testing based on search configuration, be a patient scientist. Change one variable at a time, run with the configuration for long enough to gather significant data, and compare the results against your goals.

When you’re ready to talk more about leveraging commerce search to improve metrics, selecting a commerce search vendor, or how to think about personalized vs. optimized product discovery, get in touch with me!