Purpose By means of a multicentre retrospective study based on the failure of 418 aseptic unicondylar knee arthroplasties (UKA) our aims were to present the different types of revision procedure used in failed UKAs, to establish a clear operative strategy for each type of revision and to better define the indications for each type of revision.Methods Aseptic loosening was the principal cause of failure (n=184, 44 %) of which 99 cases were isolated tibial loosening (23.5 % of the whole series and 54 % of all loosening), 25 were isolated femoral loosening (six and 13.6 %) and 60 were both femoral and tibial loosening (14.3 and 32.6 %). The next most common causes of failure were progression of arthritis (n=56, 13.4 %), polyethylene wear (n=53, 12.7 %), implant positioning errors (n=26), technical difficulties (n=six) and implant failure (n=16, 3.8 % of cases). Data collection was performed online using OrthoWave™ software (Aria, Bruay Labuissiere, France), which allows collection of all details of the primary and revision surgery to be recorded. Results A total of 426 revisions were performed; 371 patients underwent revision to a total knee arthroplasty (TKA) (87 %), 33 patients (7.7 %) were revised to an ipsilateral UKA, 11 (2.6 %) patients underwent contralateral UKA (ten) or patellofemoral arthroplasty (one) and 11 patients (2.6 %) underwent revision without any change in implants. Conclusions Before considering a revision procedure it is important to establish a definite cause of failure in order to select the most appropriate revision strategy. Revision to a TKA is by far the most common strategy for revision of failed UKA but by no means the only available option. Partial revisions either to an alternative ipsilateral UKA or contralateral UKA are viable less invasive techniques, which in carefully selected patients and in experienced hands warrant consideration.