Well, Actually ...
Just about anything, really.
The simultaneous lineup sucks, mainly because it's common for witnesses to just point out whoever looks the closest to what they remember, whether the actual perp is up there or not. The question the police want answered is "Do you see the guy who mugged you up there?" but the witness always hears it as "Tell us which of those six guys looks the most like the dude who mugged you. And it has to be one of them, or else you've wasted everyone's time."
"No one's out there stopping murders now, thanks to you."
And it's an even bigger problem when the officer standing there with the witness is the one who's working the case -- the officer is going to, consciously or unconsciously, subtly do everything he or she can to make that little old lady pick the guy the cops think did it (scientific experiments always have to control for this, otherwise the results are considered invalid).
There is an easy fix -- it's called a sequential lineup. The witness sees photos of people one at a time instead of all at once. The trick is not to tell the witnesses how many photos there are in total, and to only allow them to go through the pictures once. That's all there is to it -- it prevents witnesses from simply picking someone out because they're the best option in the bunch, it encourages them to keep looking until they see one who actually fits their memory, and it eliminates the chance for an officer to say things like "Now, did you look closely at No. 2? Look at him again, with his squinty, purse-stealing eyes ..."
"No, the guy who robbed me was painted more maroon than red."