Lesson learned: our benefits preference survey told us what people say, not what they use

A cautionary tale. Two years ago we surveyed our 3,800 employees at an Australian energy company about which new benefits they wanted. The clear winner, chosen by 64%, was subsidised childcare. We spent a year and a lot of goodwill negotiating places with two providers near our main sites.

Take-up after twelve months is 41 families. It turned out most parents already had arrangements near home, not near work, and were not going to move a settled three-year-old.

What I should have done was ask people to trade benefits against each other and against cost, and test intent with something closer to a real decision. Has anyone found a method that predicts actual use?