Over and over again we get asked if an adaption of personalization rules is possible and does make sense. Especially the point of this meaningfulness is not answerable across-the-board and strongly depends on the context.

In this article you can find some reasons pro and con the adaption of personalization rules.

Pros

Flexibility

The adaption of personalization rules gives you the necessary flexibility to manually react extremely fast to specific constellations. If you need e.g. additional revenue at the end of a month through a sales activity, then you can influence the personalized recommendations to explicitly suggest these specific sale articles.

So a sellout ist faster possible and you knowingly resign on the more reasonable personalized recommendations to push this sellout.

Constellations Resigning Automation

Personalized recommendations are actually context independent. This means, if the recommendations are displayed e.g. on a landing page which displays a specific product type, and if it is absolutely necessary to recommend also fitting products compared to the content of this landing page, then completely automated personalized recommendations make no sense here. Because these recommendations would also diplay completely personalized recommendations in this case. And in the worst case, these recommendations don’t have to do anything with the actual content of the landing page. Therefore in this case it could make sense to adapt the recommendation algorithms.

Modified Data Pool

An adaption, which can also make sense, is the change of the data pool depending on a specific context. A very simple case could e.g. be, that recommendations on product pages of an online shop are generated based on the current product and not real personalized recommendations fitting for the current user are displayed. This so-called cross-sells are only working in the context of a product page. The selected product is needed to recommend other fitting products.

The online shop owner can decide in this case to influence the recommendation algorithm manually and to display cross-sells instead of personalized recommendations as data pool.

Cons

You Know Better Than the Personalization Engine Which Products Your Customers Like

Many of our customers claim over and over again, that they know better which products their customers like. However, in 95 % of the cases a personalization engine like Recolize generates better product recommendations than the shop owner. With nearly guaranteed reliability. In average, manual recommendations perform always worse than automated recommendations. In average at least around 15 %.

Manual Interventions are Limited

If you manually change the personalization rules, then you’ll always have the risk, that these changes are limited in time. You can’t foresee how the behavior of your customers changes, e.g. because of:

• current weather changes,
• current political changes
• or birthdays or deaths of celebrities.

Especially these unforeseen events can always be handled by an automated solution and this solution can react optimally, fast and without manual interventions. Every manual adaption of a personalization solution is therefore limited in time.

Limited Analysis Options

The adaption of personalization logics can only be analyzed after or during a specific timeframe because of the mentioned time limitation. You have to evaluate yourself, if the statistical significance for a specific personalization adaption is really reached, and if the results are really comparable with the non-adapted logics.

The calculation of the statistical significance of such an A/B test can be done like described here.

Should I or Shouldn’t I?

Flexible personalization solutions like Recolize give you the freedom to freely decide, if you want to adapt the personalization logics. Especially Recolize has brought this manual interventions to perfection by don’t adapting the underlying personalization algorithm, but filtering the data pool which was determined for the current user. So they combine the power of real personalization with manual interventions.

Therefore, from the view of a website owner, both goals get reached perfectly: a personalized experience for the website visitors in combination with the adaption of personalization rules on top for additional sellout logics during sales activities.