A BPM Prediction For 2012: Connecting The Dots

Reading the recent Harvard Business Review article from Tom Davenport et al., it occurred to me that next best offer (NBO) is actually a subset of what my colleague Jim Kobielus calls “next best action” (NBA). And when you couple that predictive thinking with advances in process mining (see Wil van der Aalst’s post and the Process Mining Manifesto), it clearly becomes possible to optimize operations dynamically on the fly. First of all, the organization could mine the existing system (the transaction logs of traditional systems or a newly implemented BPM/CRM system) to identify what happens today. This then enables you to identify the outcomes that are most interesting (or those you want to achieve) and then optimize the NBA accordingly.
We take for granted a process definition where the next action is predetermined by the arc of the process definition. But if we can do NBO in 200 milliseconds, we can also do NBA in a similar time frame. Directed arcs in process models and the business rules that go with them start to become a little redundant. This sort of combination (mining and NBA) enables wide-open goal-oriented optimization for all sorts of processes, not just those related to marketing and cross-sell/upsell ideas.
So to the prediction: I think we already have a few BPM/dynamic case management (DCM) platforms that can do NBA on the fly; so the next step, which I expect to see in 2012, is to see this predictive capability dynamically combined with process mining technologies. These platforms will also need to leverage complex event processing (CEP) and social technologies to really make them usable. Of course, all of that needs to be delivered in a compelling and accessible package.
We are starting to see the tip of this iceberg in the dynamic/adaptive case management area; no doubt, some of the usual suspects will claim they do it already. 
However, I think it will take a bit longer (say by 2015) before we see business models that really connect the dots to leverage this sort of capability. What seems to be generally missing is the management mindset to take advantage of this fundamentally new capability.When I posed notion this internally, our practice leader, Kyle McNabb, responded with this:
Here’s some food for thought based on recent discussions between our peers in the Customer Intelligence role team and folks at the Harvard Business Review, the New York Times, and the Financial Times. To summarize the frank discussions we had a few weeks ago with those editors about big data: All this technology is great, but organizations that don’t treat data literacy as a competency are doomed to be laggards. Western organizations have outsourced or simply removed too many important roles and skill sets over the past few years that would help address data literacy — business analysts, data analysts, and market researchers. Most organizations now dont have enough internal personnel who can objectively help their business identify which questions to ask, how data (including big data) can help answer those questions, and — more importantly — how analyzing data in different ways identifies new questions and answers that people are not now considering. These people have to be able to translate data into business terms and vice versa. All this technology without people skilled in data literacy (including the data scientists Jim Kobielus has blogged about) is akin to arming a baby with a machine gun. I’d be interested in understanding what firms like Nordstrom have done in terms of skills to make this work.


Process Mining – bring speed to your improvement efforts

Hi Derek:

Process Mining is absolutely the 2011 discovery and will be in 2012 a hot topic with low hype because contrary to many other trends works and brings value to companies that really want to go to the next step of business performance.

In these last quarter a lot of doubts and fears raised among the community, but I was successful to carry a project in a company that manages Airport infra-structures with success to a point where people never achieve before together with the excitement around how process mining works.

For those still with doubts, i leave two links:

Process Mining – bring speed to your improvement efforts - a key note on Process Mining, on 28th September at the British-Portuguese Chamber of Commerce.



A Process Mining Project - what is process mining all about


The Future is Here

Hi Derek,

A related thread on LinkedIn ACM Group - http://linkd.in/w3wcHZ

Happy New Year!

BPM or ACM Process Mining.

'Knowledge is between two ears and two ears only!' Peter Drucker

How about not mining processes and not talking to the people but simply real-time training the system as to what actions performers chose given a certain real-world case situation. It doesn't even require case logs as it collects the patterns from the live cases DURING execution and performs recommendations in less than a second.

We introduced the User-Trained Agent for the Papyrus Platform in 2007.

My discussion on the subject from a year ago:

Connecting The Dots


I think that mining existing process data (especially if it is stored in retrieveable manner) can have a significant impact on NBA. Especially for in those unstructured processes where the process path is not pre-determined and you want to add an element of dynamic process adjustment to it. A GPS for BPM that can highlight important "Points Of Interest" as you move through the process.

Thanks for the heads up on the HBS article, the Tesco example of beer with diapers is great.

Process Mining

Hi Derek,

interesting blog. Definitely more interest in process mining all round. Perhaps the Process Mining Manifesto had something to with it. There was some misunderstanding about what process mining actually is and hopefully the manfesto will clarify that.

Most of the comments are informed and the connections are being made with other disciplines which is a key element.

Process Mining and unstructured processes

I think there is a substantial misunderstanding what process mining could discover. Not only that there are a lot of illusions as to what process mining software can actually do but also what is required to actually utilize the gathered data in some way. Yes, there is always a lot of potential for future inventions, but large corporation resist even buying existing technology they could test RIGHT NOW if it is not already used by a share of the market at least. I also don't understand the analyst coverage as there is no proven business benefit in process mining albeit some potential.

For a start, mining unstructured processes is not possible, because which structure would you want to detect if there is none? It is possible that subsets of activities in cases or some of the activities that seem unstructured aren't and they can be identified by mining the process logs but the mining would not know what to actually do with it. Let's not forget that if you mine the process in virtually all cases the complexity is so high that there is no automatic identification of what the mined data actually tell. It always needs a process engineer to analyze the data and then create a process that will replace some of the previously unstructured work.

Process mining needs a consolidated customer care, content management, collaboration, case and process environment with backend MDM silo integration. If the process has so far been executed in another environment such as email or some collaboration tool, there is the question if the people involved will be even willing to switch to the much more restrictive BPM environment. And what if the process suddenly requires again more flexibility because of external changes? How do you identify that if you made the process already rigid? So the key will be to get people to switch to a collaborative environment first that will support both case work and well-structured processes in any mix. Otherwise the whole idea is destined for failure in any case. There would be no discernable simplification for the the business user and they would resist any change of what they do, no matter how much the process mining software says different.

And finally, Next Best Action IS CERTAINLY NOT to be found by identifying structured processes in cases. NBA is a data pattern that can happen at any point in time and (case data) space. Process Mining cannot identify the real-time external events that lead to user actions except if it would track everything that a user does with phones, emails, applications and office tools and analyze it from a semantic and data entity perspective. That would need a well defined business ontology and MDM. Apart from being technically difficult to achieve, at least in most of Europe such user tracking for employees is illegal. Unions restrict the identification of user work data that can be traced to a single person. They do not even allow the data collection much less the analysis. With the UTA we only discover patterns related to a particular user role and there is no date pointing to a person. With mining that is not possible because you have no predefined roles across multiple systems.

Finally process mining cannot simply identify, which actions that a user performs are meant for a particular work item related to a particular business outcome and to achieve which goals. Without a well defined business architecture what process mining discovers is not assignable. Process engineers would not even know where to assign what and how. It is an illusion that the process mining will in some way automatically and 'intelligently' improve processes. Maybe in a university test lab but not in the real world. Gartner sees process mining as a part of what they call iBPMS or intelligent BPM. It is the process designer who is intelligent and not the process mining.

Without the primary aspects of BPM being externalized and well defined as I propose that is necessary in my Strategic ACM approach, the whole idea that process mining will deliver grand insights is as illusionary as what we hear from Big Data mining. If you don't what the data collected means in your well defined business model then whatever data you collect will be nothing else than static noise of your IT systems!