Interesting, I've used it in the past to analyze amounts distribution and not the invoice number (since they're generated automatically and usually don't correlate with BL).
I think that the main challenge is to identify relevant insights once there's a deviation.
+ some other variables like time of transaction, IP, location etc.
you can actually score each transaction (with a random forest model) with a suspicion score...
Which transactions look shady most, prioritize your efforts… 🙂
I've built a series of workflows that performs compliance audit against our indirect booking policies. Our goal after the audits are completed is to use predictive analytics to minimize the amount of infractions identified. Context: the less infractions, the better, since it means travel agencies are complying with our booking policies.
You'd be surprised at how very little variety of tools are needed. Most will be found in the Preparation, Join, Parse, and Summarize categories.
Happy to connect in PM if you'd like to speak more.
Alteryx Transportation and Montreal User Group Co-Leader
Alteryx Core Certified
Business Analyst, Distribution and Payments | Air Canada
Not sure why, but the messages would not let me add attachments. "maximum number of attachments allowed is: (0)"...
I've been doing internal audit analytics with Alteryx for the last 5 years (and are still fully in love with all it can do) in 2 companies (very different industries). Right now I have 2 risk-ranking tests, one for payroll (33 risk factors - i.e. "only paid by checks", "short tenure" etc.) and one for Procure to Pay (17 risk factors - geographic distance between vendors and employees, fuzzy on names, no phisical address, etc.). At my prior company I had some good tests on T&E, both for FCPA risk and for actual misappropriation by the employees. On predictive I had a decision tree set that would identify expense reports that were the approval was "rubberstamped".
Since every industry and system is different, I guess I would hope to use this as a sound board for ideas on what to test, more than sharing any specific workflow that would be tied to the company and its production systems/tables/variables. I would also hope we can bring-up issues when we are somewhat stuck (i.e. how to "safelist" a keyword search, or how would you follow-up when the test trigger is more abstract "(i.e. "why did you pay this vendor significantly faster than any other?"
@aristiguieta : Thanks for the information. Maybe each thread has an attachment limit?
I would love to be able to share my work, but unfortunately it is proprietary and contains information that cannot be distributed. I'm working on a version that can eventually be shared since I would like to publish this as a use case.
I would be happy to engage in a further discussion since as I mentioned, the tools used are in the Data Prep categories. For our purposes, it was all about the logic and business rules that needed to be put into place.