ToBias: Using AI to nudge overconfident investors into better investment decisions.

ToBias API empowers brokerage firms to identify overconfidence as behavioral bias among their retail clients and nudge them away from making biased investment decisions during volatile stock markets. ToBias API uses client profile data from brokerage firms to run a machine learning model trained on experimental data to identify overconfident individuals. It also identifies stocks in the client's portfolio which are currently in a high trading period by using trading volume data of these stocks from a financial data provider. In case that the models identify both, an overconfident individual with stocks in a high trading period, ToBias API provides the brokerage firm with the information on which clients to inform on which stock to show a message which discourages these overconfident individuals from placing irrational orders on these stocks.

Our mission

We help brokers protect their clients from excessive trading due to overconfidence by providing educational messages to improve investment performance. 

The problem

Two out of ten US retail investors loses money in their stock portfolio because they trade excessively and irrationally by being overconfident and when stock markets get volatile, they get irrational - and no one is there to help them make a better decision. For this quintile of retail traders, this means an underperformance of 10.4% per year in comparison to the average market performance. Trading costs and making wrong investment decision by being overconfident are the main causes for the underperformance.

Impact

Our ToBias API can improve stock portfolio performance for retail investors by:

• helping brokers profile their clients' behavioral type using machine learning models trained on experimental data.
• providing brokers with advice specifically targeted at their clients‘ behavioral type and their individual stock portfolio.
• reason clients away from placing an irrational order during high trading periods and hence save them trading costs.

Our value proposition:

• We help your overconfident clients make better investment decisions when no one else is there to help and reduce trading costs.
• We help increase customer satisfaction for brokerage firms and help them retain more assets under management by reducing negative performance of their clients.
• We help open up a new revenue stream for brokerage firms to compensate for a decrease in trading volume, since our survey showed that participants would be willing to pay an annual fee for educational messages which nudge them away from bad investment decisions when they most need the advice.

How It Works

To build a visionary product like ToBias API we designed our four step approach which consists of gathering experimental data, engineering the features we use for our AI models and to train and test these models which we then provide to our clients via our API.

1. We conduct a survey to detect overconfidence and test messages to nudge retail investors into better decisions.
2. We perform state-of-the-art feature selection and engineering with the data we collect for our models.
3. We train and test our several models to give us the best results possible in predicting overconfident behavior. 
4. We evaluate our model performance and provide our recommendations to our clients via our API.

Disclaimers

This tool is intended for use by brokerage firms to identify overconfidence among their retail clients and provide them with some advice in form of a educational message on the negative impact of overconfidence on trading to discouraging them to place a trade which is most likely to be based on an overly confident view of ones own abilities and lead to a loss.

This tool is not a replacement for a financial advisor or intended to help brokerage clients to place a trade. It should only provide general educational advice on the overconfidence bias when the client is most prone to this erroneous behavior.

Given that we did not find any research study which confirms the results from the survey we ran, we acknowledge that other future studies may or may not confirm our findings.