Historically, any process analysis began with a series of interviews, followed by whiteboarding sessions in conference rooms and the creation of ‘as-is’ process maps in Visio, based on the ‘most likely’ scenarios that you and your team members could remember.
The process was not only subjective in its very nature, but it was also not expected to capture every last detail of the process. These maps were then used as ‘points of departure' to create future state process flows involving new technologies, outsourced operations, or some combination thereof.
There was no easy way to measure time spent in process or between stages. People used everything from stopwatches, metric reporting, and best guesses to establish a baseline.
Now let’s cut back to the present.
What if you had an objective, data-based way to dive deep into your company’s business processes? A process that did not take up chunks of your SME’s workday whiteboarding process flows with a team of consultants.
What if you had access to a process mining tool?
Process mining is a way to infer a process model or process flow using event data from a system of record. To break that down, events require three basic data elements to be tracked, every time a user takes an action while performing their job:
- a unique transaction ID (e.g. loan number, application number, invoice number)
- a distinct event or status recorded in the system of record (e.g. application received, invoice exception triggered, application rejected, etc)
- a timestamp
(Note: additional data attributes only serve to enrich the process model and quality of inferences and recommendations).
Process mining tools use this event data to create visualizations of a process flow, identify Key Performance Indicators (KPIs) and create a process baseline.
Process Mining vs Task Mining
Now, process mining relies on historical, event data to create a process visualization. However, it is extremely rare to find a company where employees work out of a single system and capture each and every task that they perform within a single application or database. This leads to a lack of transparency between events.
The solution to this is task mining. Task mining is a way to record ongoing work being performed by team members and extrapolate that effort to fill in the visibility gaps. It is usually utilized more sparingly than process mining, and for shorter periods of time, given that it can be intrusive and results in a lot of data being captured via screen recordings.
Typically, it involves installing an agent on a user’s desktop that keeps track of their activities. Personal activities can be excluded from the analysis and the process can be anonymized, so that team members do not feel targeted by the process.
Process Mining in Real Life
Coforge recently had the opportunity to work on a process mining engagement with a commercial real estate lender. The objective of the exercise was to review their existing multi-family origination processes and identify opportunities to eliminate variability and increase the level of automation.
The project kicked off with a data request. In an ideal world, all companies will have all their data neatly stored in a single database ready for analysis. But as we all know, this is seldom the case. Companies grow through acquisitions. Different divisions use different applications and users don’t necessarily capture every process event in their system of record, as they should.
Step 1 – Data Wrangling
First off, we had to identify a unique identifier that allowed us to tie data back to a single transaction across multiple systems. Once that was done, data was ingested into a database and cleansed for personally identifiable data. It is important to note that personal data such as names and addresses of borrowers are not required as part of a process analysis and need not be collected in the first place. At best, if a geographic analysis is pertinent, zip codes or MSA (metropolitan statistical area) data may be retained within the dataset.
Step 2 – Process Mining in Action
The next step was to run the data through the process mining tool and generate the preliminary process flows. This first pass also yields initial KPI data that can form the basis of future analyses. At this stage, the client had an opportunity to gut check the process flow and the KPI calculations to determine if we were on the right path. For KPI data that did not make logical sense, a further deep dive was performed to understand the cause and a fix was put in place.
Step 3 – Acting on Insights
With the baseline model in place, the team had detailed workshops with the client leadership and SMEs to deep dive into processes and create and test various hypotheses. The team jointly identified root causes for the bottlenecks being presented visually through the process mining tool. Lastly, this facilitated discussions around opportunity identification for future improvements. Every business leader who attended a deep dive session walked away with a much clearer view of their processes and a way to objectively validate their gut feelings about their processes. Seeing their process data flow through the visualization on the screen almost always brought forward a slight gasp!
Lessons learned
Without getting into too many specifics of our clients’ business, we were able to validate things that most business owners know to be true but don’t have a way (or the data) to seriously drive change:
- Process rigor – Some teams (sales, especially) don’t do a good job of capturing deal information through the process. We saw data for initial stages missing in a significant number of deals. However, once a deal was qualified, the rigor improved. If that upfront data is important for your business, you must have controls and incentives in place to ensure that data capture is complete.
- Disparate systems – We observed some large gaps between events captured in the system of record. Upon further analysis, we noted that large parts of the process were being performed outside the system via email, Excel and Word. Here is where task mining helped to bridge the gaps and increase the level of transparency.
- Resource prioritization – We saw a significant number of deals come in, go dormant and be closed after months of lying in limbo. Upon further analysis, we understood that borrowers in this space tend to shop their loans around to multiple lenders. Lenders spend effort upfront to issue quotes on these deals and have their pipelines filled with ‘dead’ deals.
However, they do want to know what’s going on in the market and consider this a fair trade-off. Additional digging produced a list of counterparties that were more likely to consummate a deal than not. Our recommendation was to score new applications based on their propensity to close and prioritize efforts accordingly.
- Data-based business case for transformation – Process mining creates a process baseline, as we have discussed before. We now can go back again months after implementing new technology or process automation solutions to re-run the process mining analysis and show improvements (or otherwise) in KPIs. This leads to objective conversation with executives to showcase wins, ask for additional funding etc.
In a follow-up article, I will talk more about other use cases for process mining. If you would like to know more about how process mining can help you solve a business problem please get in touch with us at coforgebps@coforge.com and we would be happy to discuss.
