Yearly electricity demand is as varied as the people and businesses that use it. This poses considerable challenges in confirming the plausibility of automatically generated invoices. But mistakes lead to considerable costs and unhappy customers. Can we improve quality by automated plausibility checks?
get a free data sample
Anonymised SAP data (in csv files) of past transactions and consumptions are available. Request a sample set now!
get the data USE CAse
Kolumbus serves tens of thousands of invoices to customers every year. In the case of normal households, not much changes every year and employees only check a small sample of invoices. Problems still arise because in some cases the wrong consumption or none at all is communicated to Kolumbus. In these cases it should be possible to see whether this year’s electricity demand is similar to last year’s. Devising a rule/model with meaningful thresholds would save a lot of money.
The problem becomes more difficult for invoices to large electricity consumers. Here, every invoice is checked manually but large customers have individual contracts and the process is difficult and time consuming. Still, a few % of invoices contain mistakes. A system that points out potentially wrong numbers would be of tremendous help.
A project duration of approximately 2 months is expected.
What you bring to the table Experience in accounting In-depth knowledge of anomaly detection and text mining. Expected Outcome Reliable identification of faulty invoices Baseline to beat: Reduce number of faulty invoices by half. Milestones
The project can be divided into three milestones:
The first milestone is reached when the available data has been cleaned and brought into the right format for further analysis. Visualizations and descriptive statistics have been used to understand the problem of identifying mistakes in invoices. The second milestone is reached when a model/machine has been developed that accurately identifies faulty invoices to households. The third milestone is reached when another algorithm is reliably identifying mistakes in invoices to large businesses. The result is presented to and accepted by Kolumbus. Let's talk! Patrick Majunke –Your personal consultant firstname.lastname@example.org GOT ALL THAT? APPLY FOR THE CHALLENGE!
If you are convinced that you can solve the problem together with us and you are up for an intense experience full of opportunities then directly apply for the challenge.
Let's talk! Patrick Majunke –Your personal consultant datahub@ gruenderallianz.ruhr
Close X GET YOUR CALLBACK
Oops something went wrong. Please make sure to fill out everything or try again later.