Actual Confusion Doesn't Prove Likelihood of Confusion

Dear Rich: In yesterday's question, the writer said that another company stole their trademark and that the writer got a complaint from a customer of that company. Since they have proof that customers are really confused, why do they need to prove the "likelihood" of confusion? Can't they stop the other company just by showing the actual confusion? You'd think so, wouldn't you? After all the standard for trademark infringement is likelihood of confusion. So, if you've got the real thing, why do you have to prove the "likelihood?"
Like a lot of things in trademark law, it's complicated. Although proof of actual confusion can be very convincing evidence, it is not always as convincing as it appears. As this article explains, the proof may be vague, anecdotal, and perhaps even untrustworthy. More importantly, it may not reflect how consumers actually interact with the marks in the marketplace. For this reason, proof of actual confusion is only one of the factors weighed in a trademark dispute