Consider three guys, Bob, Joe, and Fred. All have four cards. For all three guys all cards are reporting zero.
In Bob's case, that's because he has never used his cards for the last two years.
In Joe's case, it's because he recently paid all cards to zero, but usually he has 1-2 cards with a small balance (different cards each month usually).
In Fred's case he uses all four cards each month but always pays the card to zero before the statement prints.
Historically FICO has never been able to tell the difference between the three guys. All FICO could see is that their cards were (currently) at zero. The three guys represent different levels of risk: Joe and Fred are demonstrating that they know how to use credit and neverthless never miss a payment and have very low debt. In Bob's case, the fact that he hasn't been late is much less meaningful. As far as FICO knows, the only reason he's never late is because he hasn't used the cards. If he ever needs to use them, he might run up debt or have trouble paying on time.
The key issue for FICO was somehow trying to detect whether the person was actually a user of his credit cards. And for most of FICO's history, the only way to do that was to see if at least one card was reporting a positive balance.
In future models (e.g. FICO 10) it is possible that FICO might leverage a comparatively new thing called Trended Data, in which case FICO will be able to determine if you use your cards even if they all currently show $0. None of the big FICO models now (including FICO 9) use TD. Vantage 4 does use TD, so perhaps FICO 10 will attempt to do that.
