Personalisation was supposed to close the gap between what the customer wanted and what your site showed them. You implemented it. The gap is roughly where it was.
This is one of the most consistent findings in ecommerce businesses that have invested seriously in personalisation technology over the last three years and I want to name why it happens rather than just that it does.
Personalisation technology works on the data it is given. If the data architecture underneath it does not cleanly connect browsing behaviour to purchase history to CRM profile to real-time session context, the personalisation engine is making sophisticated decisions on an incomplete picture of the customer. It is like a tailor with excellent technique working from measurements taken three years ago on a different person. The craft is real. The fit is wrong.
The gap between what personalisation was supposed to produce and what it actually produced in your business is almost certainly not a technology problem. It is a data architecture and integration problem that the technology purchasing decision did not surface because the technology vendor was not incentivised to surface it.
Related reading