Ad Hoc Workflow Analysis
Anything we do enough times, we develop patterns around. Hiring decisions, sales cycles, safety testing, handling complaints, and financial filings are a few examples.
Some organizations put a huge amount of effort into developing appropriate “processes” for these things, but they are hard to enforce. Employees will form their own procedures for any number of reasons, from not knowing the rules to not having enough time to outright attempting to circumvent controls. Finding a mismatch between approved and normative processes can be the first step in fixing unworkable processes, and finding anomalous workflow instances can help identify employees who are malicious, or simply disengaged. So how can we tell what is actually happening?
Workflows leave behind electronic artifacts. Emails to other team members, personal notes, phone calls to candidates or vendors, and entries in software systems all record pieces of what actually happened. The key is bringing them together. Fortunately, artifacts of workflows tend to be visible as Discussions™. After all, the data generated about a particular job candidate or prospective sale are causally related events and constitute a conversation. Linguistic methods such as ontologies can be used to find all workflows of a given type, and to tag the events within them. For example, for financial filings we might tag approval by team leaders, financial officers, and legal officers. For hiring decisions, we might look for evidence of a resume being received, an interview occurring, and a decision being made.
Once workflows have been collected and tagged, they can be analyzed. Clustering techniques can be used to identify major patterns, and rigorous statistics can be applied to identify biases.
Read about other Cataphora technologies:
