Machine Learning Reply helps companies automate Management Reporting processes.
It is well known that the reporting, monitoring and management control needs of a service company become exponentially more complex as the company grows in size.
As the number of active jobs increases, it becomes increasingly difficult to identify issues in a timely manner, so that they can be effectively addressed and before their impact becomes irreversible. Once an issue has been identified, it is even more difficult to dissect its causes and take the action needed get the project back on track.
If we raise the bar in the area of Business Management goals, a greater impact would be seen by using a tool capable of anticipating the onset of these anomalies or predicting how they will be generated, thus minimizing – and, in some cases, eliminating – any negative effects.
Machine Learning Reply responded to these needs by developing the Smart Digital Controller, a tool capable of monitoring and analyzing all of a company's projects and highlighting – individually or as a whole – potentially critical scenarios for each job through the analysis of specific KPIs differentiated by area of interest or by business specificity.
Thanks to this approach, it is possible to isolate the positions that require attention from all the active projects, and thus show users the area(s) for which a warning has been raised. The user will then be able to take the best corrective action to get the process back on track or notify (and therefore teach) the system of valid reasons why the warning need not be raised.
Through notes entered in natural language, it is, in fact, possible to give feedback on the warnings raised, specifying which positions should not be considered to be at risk despite abnormal KPIs, or, vice versa, which jobs should be closely monitored despite pure data analysis not flagging any issues.
Artificial Intelligence predictive models also make it possible to predict which positions are most likely to raise certain warnings in the near future, drawing on past performance and comparison with positions that have similarities on various parameters. This generates value in the ability to anticipate – or at least be timely in responding to – potential critical issues and to identify repetitive patterns considered to be at risk.
A systematic approach that ensures that all the steps required are carried out correctly in order to define a solution adapted to the individual client
Analysis with the business team
The methodology developed starts with the analysis of internal processes to be carried out in collaboration with the business team, i.e., the end users of the tool. This is where the KPIs that will make up the reports are discussed in more detail and possibly edited or added, along with any useful outline information to describe the positions that might be highlighted by the Digital Controller Reporting (DCR).
Defining the assessment KPIs
The second phase involves more technical aspects of the client's as-is situation. It involves identifying and analyzing available pre-existing data sources to understand which KPIs can be monitored and analyzed. In this phase, the database to be analyzed is defined in order to flag critical situations and establish the metrics to be used.
KPI development and implementation
Once the framework has been defined, the actual development of the reporting system takes place. This involves data ingestion processes, calculation of indicators, and training the Machine Learning models. At the same time, an initial version of an interactive dashboard is also implemented. Here, the above-mentioned KPIs and information can be displayed, and user notes can be introduced.
Feedback gathering and improvement
The final phase of industrialization involves a cycle of user acceptance tests aimed at validating the correct functioning of the models and the user-friendliness of the dashboard, until a suitable version tailored to the individual client is achieved.
Anomaly Detection, Forecasting and Natural Language Processing models can be used to identify recurring patterns and define threshold values based on available history
The key element of Digital Controller Reporting, after defining the KPIs described above, is to determine the KPI values that require a warning light that allows the user to focus their attention on where there is a real problem to be managed. Through an Anomaly Detection model, threshold values are identified for each KPI. If exceeded, the specific job can be identified as abnormal and requiring attention.
To predict situations in advance on specific jobs that could result in exceeding a threshold, forecasting models have been integrated that focus on the historical job series at monthly granularity and provide a projection of the monitored KPIs over the months that follow. If the projection shows any anomalies, warnings are raised to allow the client to act in time and correct the problems that are causing these deviations.
To address the need of clients to transfer feedback to the system with regard to the warnings raised, it is possible to link notes explaining the current situation and any anomalies. A Natural Language Processing model processes the notes and is able to identify their intent, as well as any entities referred to in the notes themselves. Through the notes, it is, for instance, possible to tell the tool:
Which positions should not be highlighted despite the presence of abnormal KPIs.
Which positions should be monitored carefully even though their KPIs are within the normal range.
Time periods during which certain warnings can be ignored.
References to other jobs.
Amounts and values to be changed in the data extracted from the tool.
The DCR has often proven to be of huge benefit in several ways. First of all, it is a tool that can save time during control activities since problems that are present or might occur within a project are immediately and automatically highlighted. The choice of operations to be carried out once the problem has been identified is also facilitated by the large number of indicators that allow a 360-degree analysis of the project. The machine learning component also makes the tool highly adaptive to the individual application on which it is installed while also allowing constant interaction with the system through the natural language interpretation of notes.
While our methodology has thus far been applied to monitoring the management and improving the control of service companies, the same methodology is becoming available for more general business processes involving multiple items of data that need correlating, analyzing and monitoring. Examples include the human resources departments of any company or procurement management for the retail industry.
Machine Learning Reply is the Reply Group company specialised in providing artificial intelligence services and solutions, guiding its customers towards a process of digitisation and helping them become more competitive and data-driven thanks to the adoption of Smart Analytics, Machine Learning and Artificial Intelligence technologies. With extensive experience in deep learning, artificial vision, natural language processing and predictive modelling, the company helps its customers to enhance their business, putting highly experienced development teams at their disposal. These dedicated teams provide customised artificial intelligence solutions designed to reduce costs, maximise productivity and improve accuracy and speed. By collaborating with prestigious technological and institutional partners in the sector, Machine Learning Reply is able to keep pace with the latest industry trends and, consequently, to offer products and services in line with the highest market standards.