Zero or negative lift means that the system couldn't optimize your ads effectively. In this situation, ads end up being delivered more or less randomly between the control and optimized groups, and it's possible that this will result in a lift that's lower than expected for the optimized group. There are several reasons why this can happen:
If it's been only a few days since optimization was enabled for your network, there may not yet be enough data to provide an effective optimization. The following situations may result in insufficient data:
The daily ad impression volume was too low.
The campaign period was not long enough.
Optimization will always respect delivery goals, and the scheduled needs of the ads may have been more important. In this case, ad delivery will tend to be more random and closer to normal DFP delivery, and it's possible that the results for the optimized group will be lower than for the control group.
If the delivery goal is very close to the total amount of available inventory, essentially all of the inventory will go to either the control or optimized groups. When the available inventory is larger than the delivery goal, the optimization algorithm can be more selective and pick the inventory that will result in the highest CTR.
Similarly, overbooked (or close to booked) ads are less likely to be optimized, because there is little flexibility in whether they can be selected for optimization.