Your personalisation engine is making very sophisticated decisions on very incomplete data and the gap between those two things is where your investment in it is going.
Personalisation technology is not the constraint in your business. The data architecture feeding it is. The customer profile your personalisation engine is working from is a partial picture assembled from sessions that were not all attributed to the same user, purchase history that does not fully connect to browsing behaviour, and CRM data that was last enriched at a point in the customer relationship that may no longer reflect who the customer is now.
The engine is doing exactly what it was designed to do. It is making real-time decisions based on the data it receives. If the data is incomplete, the decisions are sophisticated interpretations of an incomplete picture. The personalisation looks right in the interface and is wrong in the commercial outcome it produces with a frequency that the click-through metrics on the personalised recommendations are not designed to surface.
I think about it as a personal shopper working from a client profile that was written by someone else, three years ago, about a different customer with a similar name. The technique is excellent. The fit is wrong.
Fix the data architecture before you invest more in the engine running on it.
Related reading