There are a lot of ways to do the legal system right. Doing it justice-wise is just one of them.

A modern digital bail system has three pretty good design features that should go hand in hand.

First, it's supposed to streamline the judicial process. Current bail isn't necessarily there because a judge thinks it's necessary. (In a few circumstances, bail may be warranted, but a perfectly good alternative to it is available. And in most situations, bail serves as an incentive: If someone has a cash bail bond to post, then her bail is replaced by the required resources, and her release is effectively preordained.) But the bail process is a monumental paperwork nightmare. The state of Florida alone, in a whopping 1,326 cases in 2015, spent $9.9 million on bail bonds attorneys.

Second, digital bail systems make it easier to avoid pretrial detention, with the added security that someone responsible for the clunky old practices won't be sitting behind bars.

Third, in an effort to mitigate the risks posed by poor risk assessments, digital bail systems are supposed to provide a risk-adjusted risk profile that helps each person and family choose the safest option. An algorithmic algorithm, like one likely used by most counties and courts in the U.S., would predict each person's riskiness based on an oversimplified array of financial and criminal records and behavior.

For all these reasons, many states and municipalities have made bail reform a priority. As early as 2015, the American Bar Association developed a checklist for states to use to help them create their own digital bail systems.

But, in reality, when technology and data are applied to the bail system, the outcome hasn't always been what we hoped it would be.

In the legal term, the entire "procedure" is called court entry, so when it comes to bail reform, court-entry systems are by definition imperfect. More focused systems, like criminal charges, are often more accurate, but the "process" is the thing that matters. So because the digital process for the bail system isn't yet perfect, it wasn't possible for it to solve the bail-bail problem.

We launched a research project last year to study how accurate that process really is. Here's what we learned: