Operational Business Intelligence
Optimizing banking operations with business intelligence.
Following EAI’s FORCE methodology, the second module right after the Foundation is the Observation. This is where we cover the awareness and monitoring of the metrics that Stakeholders previously defined on the Foundation module, with what’s commonly referred to as Business Intelligence.
What is Operational Business Intelligence?
Operational Business Intelligence, or OBI, is the process of collecting and analyzing data from operational business processes or activities within an organization, to enable better strategical and tactical business decisions.
It allows businesses to generate key insights such as what type of customers buy one specific product, when are they buying it, why, and where.
These analyses are commonly deployed on static or interactive dashboards, in order to promote awareness across stakeholders at all levels (the entire organization, one department, or one team) of their particular KPIs and OKRs.
What are KPIs and OKRs, and why are they important?
KPI stands for Key Performance Indicators, and they measure certain operational outputs or business results over time. They can be defined on different levels: across the entire organization, for departments, teams, or even individuals.
The OKRs differ from KPIs in that they provide a concrete business objective and a direction and actions to improve said objective, while KPIs track day-to-day operations but won’t really tell you what to do. KPIs are more permanent, but OKRs will be constantly evolving depending on what the company should be focusing on at any given time.
Together with a target value, they provide a direction for the people responsible for them, help identify the impact of any given initiative, and help distinguish between good and bad-performing departments, teams, or employees.
One way to look at it is to consider that you are piloting a plane from Los Angeles to New York. Your KPIs will be the speed, air pressure, altitude, gas level, temperature, etc. These measure the conditions under which you are flying. The OKRs, on the other hand, will act more as a roadmap to your destination, for example by dividing the entire trip into certain milestones. If you only have KPIs you might be flying well, but it won’t guarantee you reach the destination you wanted while flying only with OKRs would be like flying blindly and most likely would crash.
For example, “Monthly Revenue” is an important financial KPI for any company, but this alone won’t be helpful for the business in order to improve it. You need to couple this KPI with an Objective and Key Result, which in turn leads to an initiative, such as:
Objective: Increase Monthly Revenue
Key Result: Increase personal loan sales by 20% from the same period last year.
Initiative: Promote a lead-scoring campaign for personal loans.
This is now something your sales and marketing departments can work on.
For your KPIs and OKRs to be of any use, they first need to be tracked properly, whether this means having someone track them manually or collecting them automatically through an API at a later stage.
Afterward, they need to be visible to all those involved in making sure they reach the targets or don’t fall below certain thresholds. By doing this, we are creating awareness across the organization of those KPIs or OKRs and everyone will be able to see the impact their initiatives and actions are having on their personal performance or on the organization.
For this, you need to have dashboards, like an airplane’s cockpit, but with only the most important metrics for a specific department or person. Too much noise will just turn the dashboard into an information radiator, and people will eventually stop paying the attention it deserves.
Which branches are outperforming or underperforming?
With KPIs defined across branches, you can monitor and compare the performance of different branches across geographic locations and figure out which ones are having a stellar performance, selling personal loans or mortgages, for example, and which ones are falling below expectations.
Armed with these insights, it is crucial to study these branches thoroughly and take away the best practices they are implementing, for example, using different sales scripts or offering different conditions. You could also realize that geography alone is the reason they are outperforming, which could suggest opening more branches or targeting these particular regions for marketing campaigns.
Monitoring operational risks
An operational risk is a risk that will cause your company to lose money or fail to meet its business goals. It includes the risks of not following systems or processes properly, such as the failure to identify anti-money laundering operations or identify the riskiest applicants for mortgages.
As such, these processes will need to be monitored through KPIs, and these KPIs need to be visible to the teams working on the processes at all times. This requires one or several dashboards so that the team or employees responsible for the process can be aware and react in time before the problem becomes an operational risk.
Anomaly detection
These KPIs need to have not only targets for good performance, but also some threshold values we do not want them to reach. This allows the creation of alerts that will automatically fire as soon as we get near or pass the threshold values.
An occurrence that causes the KPI to get near those threshold values is typically an anomaly. One can use this data later to find out which events ended up creating the anomaly. For the process operators, this means going through the data, documents, or logs to figure out any discrepancies between the usual day-to-day operations. Depending on the volume of data, it can become a costly endeavor, much like finding a needle in a haystack.
While large volumes of data are bad news for human operators to search manually, they provide a great training ground for a machine learning algorithm.
Our Experience
For one of our clients in the banking industry, we’ve built dashboards to monitor the status of key jobs in the data engineering pipeline. These were crucial to identify problems such as job errors from the logs, missing variables, or variables with statistics that would change a lot over time, all of which would potentially impact the performance of the models already deployed to production, that used batch and near real-time data for predictions.
The model performance was also measured using custom reports in Power BI that featured common KPIs in machine learning, such as ROC AUC, Precision, Recall, F1-score, Lift curve, and other important metrics defined by the business depending on the model and use case.
You can read about other Operational Business Intelligence use cases here.
Want to learn more?
Miguel Cabrita
Senior Data Scientist, Co-founder
Miguel has helped various companies in banking and finance implement lead scoring and AI solutions.
Having a strong technological background and understanding of business processes and the banking industry helps him detect specific needs and offer the necessary AI solutions for each of them.
