Criminals are occasionally recorded on video committing a crime. At trial, jurors may be shown images of the culprit to determine if they match the defendant. However, several sources of bias may influence juror matching decisions. Also, even with clear video, the accuracy of defendant‐culprit matching can be relatively poor. To reduce these problems, we propose that defendant‐culprit matching be viewed as a type of forensic test. If conducted as a forensic test, defendant‐culprit matching can be improved by adding fillers and testing mock witnesses rather than the actual jurors. A Bayesian analysis of data from two experiments shows that a filler‐control test can be highly diagnostic, even when the decisions of mock witnesses are far from unanimous. However, when viewing conditions are poor, a filler‐control test may not provide much new information about whether the defendant matches the culprit. Copyright © 2017 John Wiley & Sons, Ltd.