An Intellyx BrainBlog for iGrafx
In our previous chapter on automated decisioning, we discussed the importance of bringing together all of the real-time and historical data necessary for a human expert to make better decisions in the moment.
As we shift our gaze from the present to the future, we uncover another critical factor in continuous process optimization: predictive analytics.
At the time of any strategic or tactical decisionpoint, what if we could already predict the most likely best path forward?
With so many variables in play, teams may not be able to accurately observe the success or potential failure of a process in flight—and in the real world, there will often be varying degrees of success or failure rather than a clear binary outcome.
A predictive analytics approach looks at how close we can get to the intended outcome in order to guide our actions.
How will we know when we’re done?
I didn’t get to spend enough time with the legendary optimization guru Ken Sharma in my early supply chain days at i2, but I do remember his signature line upon starting any meeting: “How will we know when we are done?”
Indeed, if there wasn’t a clear intention to accomplish something at the beginning of a meeting, even with all the bluster and strategizing in the world, it’s unlikely that thing would ever happen.
Using predictive analytics, we can start with a timing goal for a desired outcome, and work backwards from there.
Take for instance a mortgage approval process, a complex flow which contains many tasks. These include market evaluation, appraisals, risk assessment, fair housing, insurance, and more. Manual forms, checks and approvals are conducted by agents on behalf of the lender, buyers and sellers.
There is a best case or ideal timing goal we want to try to meet in order to provide a great customer experience for buyers and sellers, which could be a week, or even a day, if everything flows smoothly.
There is also a worst-case or minimum drop-dead date for process completion. It means an SLA violation or compliance issue is triggered, or in the case of our home mortgage, the contract period expires and the whole deal becomes null and void.
Being able to accurately predict when a process will be complete is absolutely fundamental to any enterprise and its customers, and it is the primary reason why leaders might intervene to correct or expedite that process.
Digging deeper than process knowledge
The art of pairing process mining with analytics has come a long way over the last decade.
Highly competitive firms consider it a closely guarded proprietary secret for how they do business. While the exact mix of tools may vary, this practice involves extracting and meta tagging event data from systems of process and record, and then correlating that data within data warehouses, in order to make better decisions using advanced analytics.
Data mining could be much more useful than a post-mortem archaeological documentation of events and performance that contributes to the trend lines within an analytics dashboard.
What would it mean to the enterprise if we could turn our gaze forward, and use this collected body of knowledge and current state events to predict when processes in flight will be complete, and take action to guide them to more likely positive outcomes?
Sort of like an auto-complete for processes … but it needs to be way more intelligent than that.
From archaeology to sociology
There is an interesting thru-line from process mining to predictive analytics.
Ever notice how archaeologists will dig up, categorize and study artifacts, then pass them to anthropologists? The anthropologists reflect on them to understand the motives and behaviors of past cultures. Then, sociologists try to predict how this collected knowledge will impact human group behavior and society in the future?
Translating this metaphor into business terms takes more than just setting rules, heuristics and statistical models. A predictive analytics system will need to learn from collected knowledge, in order to anticipate impacts in the real world.
Supply chain optimization offers a great venue for seeing predictive analytics in action. An advanced high-tech manufacturer relies on a galaxy of suppliers. These suppliers source raw materials, produce parts and semiconductors, sub-assemble parts, assemble goods, and ship finished inventory.
In addition, the high-tech brand at the hub of this network is also constantly watching customer orders coming in. The ‘demand signal’ side of the equation results in additional orders and network manufacturing workload.
But why should they even wait for the orders? Predictive analytics can anticipate seasonal retail and business buying trends, but it can also provide insight into outliers and other leading indicators. A new Fortune 100 corporate headquarters or a spike in merger activity could signal an increased workload. This may cause the brand to start finding additional supplier capacity ahead of pending orders.
The preventative side of predictive analytics
Most process mining solutions can identify possible areas for realizing efficiency and performance improvements after the fact. Existing process models can be adjusted or replaced based on results.
But what if we could start remediating customer complaints and compliance issues while they are happening, by analyzing processes in flight that are not proceeding to plan?
A major international airport has strong SLA agreements with the many major companies relying on their hub. It also has very strict data privacy and sovereignty requirements from the government regimes its passengers come from.
After refreshing its datacenter infrastructure, one of the airlines started noticing that some passenger records were becoming inaccessible to desk agents. Before a privacy compliance report emerged or flights got canceled, the airport’s predictive analytics kicked off a disaster recovery workflow. It restored the infrastructure to its previous good state and prevented a possible data exfiltration or ransomware attack, or just an unknown bug hidden somewhere within the systems.
The Intellyx Take
There is no reason why enterprises should solely rely on process data dug up from past events to understand the future.
Predictive analytics is a practice, not a product. It builds upon the advancement of process mining and machine learning-driven analytics. It also allows the massive real-time data of processes in flight to inform a more proactive approach to improving performance.
Fortunately, it is not just an aspirational goal. Leading process optimization vendors like iGrafx are already incorporating predictive analytics to proactively spot an earlier remediation of process failures. This enables enterprises and customers can breathe easier.
©2023, Intellyx LLC. Intellyx retains sole control over the content of this document, and none of the copy was written by AI. At the time of writing, iGrafx is an Intellyx customer.