We have all been exposed to various automated recommendations for content and product consumption. These online widget powered recommendations are everywhere these days, and many marketers, eCommerce merchandisers and publishers are leveraging these units for improving recirculation, increasing monetization and reducing exit rates. However, behind these powerful widgets are algorithmic decisions, dictating various recommendation “approaches”, also known as recommendation “strategies”.
As you can guess, not all recommendation strategies are created equal. In fact, misapplication of strategies can cause not only the loss of readership or revenue otherwise gained by a successful recommendation, but also loss of faith by the customer in the value of your recommendations altogether – and as a result in your ability to deliver a personal and engaging customer experience. With customer experiences slowly taking over as the key differentiator between online entities, it is not only a shame to miss out on the opportunity of recommendation engines, but choosing the wrong one may actually constitute a contributing factor to an overall failure.
While contemplating the most appropriate strategies to apply, let’s not forget that the purpose of a strategy is always to promote a goal. In essence, the marketer must first decide which goals he/she is trying to achieve and only then can a suitable strategy be selected and applied. Additionally, with the ubiquity of data collection, aggregation and segmentation technologies available to marketers, different recommendation strategies can be delivered to different users and segments which presents yet another fantastic opportunity to connect with visitors on a personal level.
Let’s delve into the various recommendation strategies provided by the top personalized recommendation engine providers, and understand how and when they should and should not be used:
One popular/effective recommendation strategy is to display ‘top items’, or items ranked as “highest” or “best” based on custom criteria such as popularity, views, purchases, trends, exposures, additions to cart and special promotions.
As this strategy usually does not contain personalization data, it is recommended for use when little-to-nothing is known about a user, or when a user displays general “browsing” interest in your site. The “Top Items” strategy is great for promoting your “hottest” items and standing out against your competition. Use “Top Items” to assist with product/article/reference discovery and to market your brand by promoting your top merchandise, content or referrals, as applicable.
However, if you DO have in depth personal data fairly obtained from a visitor, you may want to consider the following strategy as your best option:
Recommended for You
Recommended for you is primarily based on perceived or “implied” user affinities and not on explicit knowledge. This strategy first obtains a visitors’ perceived preferences and characterization which the recommendation engine collects over time, and then presents items which the individual user is most likely to identify with.
Using personalized recommendations is in essence a “swing for the fences”. If the recommendation is indeed appropriate it can serve as an excellent way of creating a lasting bond between the site and the visitor. However, 2 main obstacles may make this connection slightly difficult:
- Not enough data – not being able to construct a logical recommendation will result in a superficial offering masquerading as a personally tailored recommendation. This may cause the visitor to feel unappreciated and misunderstood, and lower his/her engagement with the site going forward.
- Too much data – providing recommendations that are spot-on but are based on little volunteered information can be perceived as “creepy”. Visitors may wonder how the site is so familiar with their affinities without the visitors having shared them, and decrease their engagement for fear of data and privacy violations.
Using the “Recommended for You” strategy can yield terrific benefits- as long as you have enough data about a user to logically suggest affinity-based recommendations, and as long as such data is perceived to have been voluntarily provided by the visitor.
As its name would imply, this strategy dictates that the recommendation widget display items (of products, listings or content as applicable) which are of a similar nature to a particular item or group of items. Complex algorithms are designed to ascertain the metrics resulting in a similarity “score” for each item, and those with the highest similarity scores are displayed to the user.
While there are various reasons to use this strategy, we have found that the most efficient way to use it would be when reaching out to users who have viewed a particular item for a while, and have not “converted” (whether by reading it in its entirety, completing its purchase, clicking on a referral etc., as applicable). It is safe to deduce that these users are seriously looking to consume a particular item but are not “hooked” on the specific item they are viewing. Offering similar items lets the users know you understand what they are looking for and are not merely trying to assist their pocket emptying or time wasting endeavors.
Focusing visitors on items they are actually looking for provides real value to the user, and increases your chances as a marketer to drive a successful sale.
Bought Together (for eCommerce)
Under this widespread strategy, visitors see items which have been purchased together with the item they are currently viewing . If your goal is to pad a visitors’ cart with additional products, it may be logical to recommend items others have bought with the viewed item, as these may often represent upselling and cross-selling opportunities (such as a car charger with a smartphone, sandals with a pair of shorts, subscription to an auto magazine with a new sports car etc.).
While this strategy is a little less “complex” from a technological standpoint— as it basically only requires the onboarding and continuous tracking of customers’ purchase history— it is somewhat flawed in that it may provide additional products which are not at all in line with what the user is looking for. Offering products others have purchased which are not at all tailored to the individual visitor may look opportunistic and even cynical to some visitors. Make sure that when using this strategy, sufficient rules are put in place to override any illogical recommendations.
Despite the grain of salt attached to this strategy, it is still widely used and highly effective. To gain the most out of the “Bought Together” strategy, make sure to use it when a visitor has already added an item to the shopping cart- otherwise you will be signaling to your visitors that their business is taken for granted. However, once items are added to the cart- this strategy is fairly simple to apply and is usually perceived positively by the visitor (provided, as aforementioned, that the recommendations are not completely illogical).
While the “Bought Together” strategy is most common among eCommerce marketers, it can be quite easily applied to other industries as well. For example, publishers can recommend content which was consumed together with converted or currently viewed content. The “Bought/Consumed Together” strategy is very useful across industries as long as it is applied logically.
In conclusion, recommendation strategies vary greatly from one another, and selecting a recommendation engine and strategy depends greatly on your goals and what you are looking to achieve. Make sure to look for a recommendation engine which allows for flexible strategy setting and testing, as well as strategy optimization and personalization (employing different strategies for different visitors/segments). By optimizing these strategies, you’re guaranteed to extract the highest value from your recommendation engines and realize your highest ROIs.