As the AP departments have been introducing automated solutions in invoice receipt and their automatic processing, the differencial in time dedicated to a correct invoice and one which presents incidents is greater and greater. One of the most intensive tasks for personnel in the accounts payable department is the resolution of incidents (incorrect invoices).
It is common practice for invoice automation processes to focus on streamlining unproblematic invoices, leaving out all invoices that are incorrect and ultimately represent the greatest workload in the process.
The optimization of the incident process is decisive as it is where the most cost resides, as on average, for every 5700 invoices with incidences an an additional administrative assistant is needed in accounts payable.
Treating exceptions and implementation of metrics
With the focus on the incident flow procedure, we should also commit to actions that reduce the total process cycle: the longer the process time the greater the possibility of failures and thus errors and costs. Tasks like document management should be reviewed to employ as little time as is necessary, setting the objective in hours rather than days.
To improve the exception resolution procedure we can undertake two types of actions:
Actions orientated to improve the ratio of recognition and matching against purchase order
This is necessary as it happens that, as a result of the use of recognition programs, invoices indicated containing problems when there are none, due to an erroneous data extraction (The OCR programs are still far from perfect and require in many cases a manual review/validation process)
This is the most difficult to resolve internally. The difference between the 70% in a standard recognition/matching process and a first class one with more than 95% accuracy, is years of technology and software development.
The process should be equipped with adaptation and automatic error resolution mechanisms.
Some ideas on this respect:
- Automatic error correction
- Automatic localization of order/delivery numbers (for the cases of order not specified on invoice).
- Recurring process in balance matching with the possibility of auto-adaptation for provider.
- In line by line recognition, use an advance algorithm that includes the order line, not only that of the product..
- Setup the matching with all possibile combinations, not only one invoice for order, but X invoices to Y orders and Z deliveries, with different numbers of lines and details.
The metrics that can be implemented in this area to control the process are:
- - % recognition = invoices automatically recognised vs total invoices recognised (automatically and manually), excluding invoices returned.
- % recognised automatically specified by provider.
- % Exceptions recognised automatically by category, at least four: quantity, amount, delivery note failure, order failure.
- % electronic invoicing.
- % partial matching vs orders or delivery notes (depending on if the “matching” is 2WM or 3WM).
Actions orientated to accelerate the exception resolution process and its analysis.
- Exception classification: to identify correctly the error category of the invoice it helps if it is assigned to the correct person. Not all exceptions should be managed by AP. The exception should be solved by the person with the correct knowledge and necessary resolution capacity. Classifying in this way the invoices, the main source of the errors can be better analyzed and the ability to take action about it. The best indicator in this area is an aging report: total invoice by category vs age.
- Improving the task assignation process: the AP personnel should not auto-assign exceptionsto resolve. The system should assign the exceptions based on, at the very least, these four criteria: knowledge/experience, exception type, priority and age.
The metrics that can be implemented to monitor this area are:
- Number of invoices pending processing with a determined age (for example, those which are more than 3 days). This parameter should be reduced as the process starts to improve.
- P85 (Percentile 85) of the exception resolution cycle: How many days are necessary to resolve 85% of the exceptions? (The percentile depends on the state of the process; P95 should be the objective value).
- P85 for each exception category.




