How statistics saved over $5 million worth of parking tickets


  • Author: Carlos Grajales
  • Date: 15 Sep 2017
  • Copyright: Screenshot of Do Not Pay (11th July 2017)

I rent an apartment in the west of Mexico City and I’ve been having some problems with it lately. Rain has been unusually intense in the city the past couple of weeks and some leaks are starting to pop-up in the parking lot. My landlord has abandoned me, so I thought it would be nice to complain about the lack of repairs, so I decided to seek a lawyer. I met him a few days ago. His office was rather minimalist, though I personally found the decorations nice. We had a nice chat. He was very direct at moments but helped me through the process with ease. After a few questions, he helped me prepare a legal document which I was able to send to the local officials to file a formal complaint. I have to say my experience with the lawyer went remarkably well, especially considering that my lawyer wasn’t a human.

thumbnail image: How statistics saved over $5 million worth of parking tickets

Well, to be honest, my landlord is not that bad and my building has been fairly resistant to the recent storms. Still, this was the best idea I got to tell you about what one of the most novel applications of statistical models in recent years; a mix of language recognition, pattern detection and statistical machine learning, all masked within the now loose term “Artificial Intelligence”. The lawyer I chatted with is an algorithm that provides free legal counsel for commonly required, basic legal needs, such as contesting parking tickets. It only takes an e-mail and a two minute login to get an audience with this lawyer: you can visit him in the website DoNotPay [1].

The original project, the website DoNotPay, was first launched in 2015, promoted as the first “Robot Lawyer” of the world. The story of the algorithm is interesting in itself, as it all originated thanks to the poor driving skills of Joshua Browder, the author of the app [2] [3] [4]. While living in UK, he got a high number of parking tickets. His parents supported him up to the fourth one, a moment when he had to start looking for a way to deal with his copious fines himself. Being a 19-year-old teenager, he had to find some original solutions that didn’t include spending too much money on his own fines, so he became interested in the legal process of appealing parking tickets. Once he learned about the process, he processed a few appeals as hiring a lawyer for such a task might cost between $400 and $900, more than the ticket itself [4]. After submitting a few successful appeals, Joshua became quite good with this particular legal process, so it didn’t take too much time before friends and family asked him for advice: he became an appealing consultant, sort of. Due to this repetitive consultant task, Joshua realized most of the drafts he wrote for his friends were fairly similar, and included much of the same information. He thought the process could be automated to make it more efficient. DoNotPay would be born.

Browder used data obtained from the Freedom of Information Act in the UK, where he got the Top 12 reasons for which parking tickets are dismissed. He then consulted some lawyers and web sources to create 12 legal templates of appeal letters for each of these reasons. The first version of the DoNotPay website consisted of a brief questionnaire where the user completed a couple of questions regarding the causes behind the ticket, along with some required details to fill the information for a draft of a legal appeal, such as name, the street where the incidence occurred and the number on the fine. With that information, the system created the full text of a valid appeal draft. It then suggested the user to attach additional evidence to the appeal, such as photographs or witness reports. A user could then easily send the appeal to the court.

The current version of DoNotPay is a bit more elaborated, since it now includes a Chatbot, which generated the conversation I described at the beginning. Using language processing, the Robot Lawyer understands English conversations to identify which of the 12 reasons for appealing you are describing. At first, Browder used a Bayesian Classification algorithm to identify the situation users were describing in the chat. When the project gained attention, IBM provided assistance for the bot, which now uses IBM Watson Natural Language Classification, a decision tree based algorithm, to match the user’s conversation to one of the appeal letters loaded in the system [3]. The author describes the whole process as “A game of 20 questions but for Parking Tickets” [3].

Thanks to natural language technology, the app has been updated to handle some other legal requests, such as helping with complaints over delayed or cancelLed flights and payment-protection insurance (PPI) claims. For its latest update, the app is now allowing users to enforce legal actions against landlords who don’t repair a property or who haven’t provided the required minimal benefits to their properties. Most of the legal background uploaded is pertinent to UK law, though some American cities now enjoy the benefits of the DoNotPay service, such as New York.

As you might imagine, some cities haven’t taken the DoNotPay service with the same excitement the technological community has regarded it. It is for that reason remarkable that some city officials are actually working with Browder to produce money saving apps for some of their own processes [3]. Alliances with governments and non-profit organizations have allowed Browder to apply his technology in a couple of different projects: helping refugees and homeless people. Let’s talk about that second one first.

In 2016, Joshua worked with the UK government to use his bot to provide free legal aid to people facing homelessness [6]. In that country, the process to help homeless people involved paying a lawyer to file an application in order to be eligible for government housing. With the assistance of a Top Housing Charity in the U.K., which provided Browder with the legal assistance required to program both the application documents and the questions required to verify the eligibility of the candidates, the Robot Lawyer technology allowed homeless peoples to prepare the housing application for themselves, without the cost of the legal fees required by a real, human lawyer. A few thousand people have used this functionality so far. [3]

The other major project the Robot Lawyer has tackled is related with the refugee crisis. A new version of the bot, this time using Facebook Messenger, can now help refugees fill in an immigration application in the US or Canada. For those in the UK, it helps them apply for asylum support [6]. This version of the app works in quite a similar way to previous iterations. The user is asked a serious of questions which are used to determine which application the refugee needs and whether he or she is eligible for asylum protection [7]. The app is available in English but is currently being translated into Arabic.

Even though Joshua Browder has become the most recognizable face of the statistical disruption happening within the legal processes, some other researchers and companies are working in ways of using models and technology to aid the legal profession [8]. TypeLaw is a commercial platform very similar to the DoNotPay original site. Just like Browder’s app automatically prepares a legal document to present in court, TypeLaw’s platform allows users to present more sophisticated legal briefs, by preparing its layout, identifying citations and formatting tables. You upload an ugly document and get a totally formatted legal brief, ready to deliver. The main difference is that, due to the complexity of the documents, this app is designed for a more professional user, unlike Browder’s systems, which are developed for the everyday man. As such, TypeLaw allows law professionals to prepare and format their documents easily. But statistics are used beyond the format and edition of legal documents, some game changers for the legal profession rely heavily on statistics for broader purposes.

Lex Machina is a platform devoted to “Legal Analytics”, my new sixth favourite term into the pool of the Big-Data/Analytics catchy slogans. Lex Machina provides a series of tools specially devised to bring valuable data-driven insights to law-firms and lawyers [9]. One of the analyses they advertise is a historic analysis of judges’ verdicts, detecting trends and patterns that might explain their resolutions. With it, a lawyer can prepare a speech or documentation by taking into account the factors the judge has historically considered as relevant in previous cases. The user can even use the model to forecast a judge’s verdict, given the evidence and documentation presented so far in the case. In fact, modeling court verdicts isn’t something exactly new, as some efforts on the topic have been around since the last century, when discriminant analyses signaled how the race and gender of the perpetrator, as well as his/her relationship with the victim could be powerful predictors for verdicts in domestic violence cases. Other factors such as previous criminal history and availability of photographic evidence proved to be also important [10]. What the company does is just add another variable in those models, the judge itself. Some authors have argued that nonverbal communications between the judge and trial participants might have a non-negligible impact on trial outcomes [11]. Even legal factors can be weighted differently by different judges. One might give more weight to photographs whilst the other is only interested in genetic evidence. As such, considering the judge in the model might improve the accuracy of their predictions.

Besides studying judges, the company also offers models to predict the length and timeline of a legal case [9]. Usually, legal affairs can be very time consuming, as some processes tend to be delayed for years. By using “Legal Analytics”, the system promises to forecast the times and durations of the different periods of the whole litigation, like time for a trial or time to termination, all these based on the characteristics of the case and even the judge involved.

The marriage of statistics and technology has provided a whole new professional platform for the legal profession. It is worth remembering that many of the most astonishing applications of artificial intelligence rely on complex statistical models, simply dressed up within a technologic platform. As such, we as statisticians hold the key to some of the biggest technological breakthroughs of the 21st century. By analysing data we can change a profession, disrupt new markets and even save over 5 million dollars worth of parking tickets [3].


[1] Browder, Joshua. DoNotPay Website (2015)

[2] The world’s first bot lawyer. Medium Website.

[3] Robot Lawyers by Joshua Browder. Partially Derivative Podcast (Jan, 2017)

[4] Garfield, Leanna. A 19-year-old made a free robot lawyer that has appealed $3 million in parking tickets. Tech Insider Website (Feb, 2016)

[5] McGoogan, Cara. 19-year-old's 'robot lawyer' overturns 160,000 parking tickets. The Telegraph Website (June, 2016)

[6] Cresci, Elena. Creator of chatbot that beat 160,000 parking fines now tackling homelessness. The Guardian Website (August, 2016)

[7] Cresci, Elena. Chatbot that overturned 160,000 parking fines now helping refugees claim asylum. The Guardian Website (March, 2017)

[8] Type Law. TypeLaw Official Website

[9] Lex Machina. Lex Machina Official Website

[10] Cramer, Elizabeth. Variables That Predict Verdicts in Domestic Violence Cases. Journal of Interpersonal Violence (November 1, 1999).

[11] Blanck, Peter. Calibrating the scales of justice: studying judges' behavior in bench trials. Indiana Law Journal Fall (1993)

Related Topics

Related Publications

Related Content

Site Footer


This website is provided by John Wiley & Sons Limited, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ (Company No: 00641132, VAT No: 376766987)

Published features on are checked for statistical accuracy by a panel from the European Network for Business and Industrial Statistics (ENBIS)   to whom Wiley and express their gratitude. This panel are: Ron Kenett, David Steinberg, Shirley Coleman, Irena Ograjenšek, Fabrizio Ruggeri, Rainer Göb, Philippe Castagliola, Xavier Tort-Martorell, Bart De Ketelaere, Antonio Pievatolo, Martina Vandebroek, Lance Mitchell, Gilbert Saporta, Helmut Waldl and Stelios Psarakis.