bpmNEXT – Day 2
Blog: Collaborative Planning & Social Business
Here are my notes from the second day of bpmNEXT on March 31, 2015. Note: I spoke on day 3, so was too busy, so these conclude my notes of the event.
Michael Grohs, Sapiens DECISION – How to manage Decision Logic
Decision aware business models are simpler, easier to maintain. Less ambiguity than natural language descriptions. Rule content can be managed by users because it is clearly separate from the program. Communities make their own vocabularies. Decision management can produce rules that run in several different rules engines.
Decision starts with an octagon, then some squares with top corners chopped off which represents a rule family linked together with arrows. Each rule family declares what values (facts) it generates. In a typical flat rule set the inferential information is lost, and once lost hard to maintain. Opening the rule family it looks like a table, essentially a decision table. Each row is OR with the earlier lines. Not sure if it is first line that matches, or last line that matches.
Example was a rules base, and then a specialized rule base for a particular state: Florida. You can see a side-by-side window, with the differences highlighted. There are logs of all changes.
Q: What about knowledge workers? Can they use it, and can they have their own rules?
A: Right now focusing on the automation, the 80/20 part. Everything is about faster and cheaper. In the future we may think about more elaborate rules. Agility and closed loop with knowledge worker is the next step.
Gero Decker, Signavio – Business Decision Management
BPM addresses 50% of the questions. The rest is making decisions. New standard Decision Modeling Notation (DMN 1.0) Signavio Decision Manager concentrates on the modeling and governance.
Drew up a quick BPMN diagram. Used a “business rule” task. Open that up and see a decisions tree. Open one node in the tree and it looks like a decision table. Created a quick example decision table. The decision node in the tree can have “sub-decisions” which are more decision tables. Product does some decision checking. There is a rule testing capability as well.
DMN is pretty powerful, it covers rules as well as predictive analytics – not supported yet by Signavio.
Export to DROOLS code, which is pretty flat view of all the rules. The hierarchy is not apparent, but coded into the rules. There is a declare statement at the top for run-time binding to the execution environment.
As for execution of the rules in the DMN standard, there are differing hit rules: first hit, multiple hits, some sort of weighting, last hit. He demoed ‘exclusive’ which means that you can not have overlapping rules, and in the case of exclusive it automatically shows you when rules overlap.
John Reynolds – Kofax – Digital World
Used to be that BPM ignored the physical world, and BPEL is the best example of that. Now we need to engage the customers int he real world. One rule: don’t force users to gather information that is already out there — use a robot instead. Some information is there in paper form. Claims that a lot of people still print their PDF files. Scanning is what Kofax does, and Kapow was awarded last year for creating BPM processes. Now, SignDoc for signing applications.
For the demo, he got out a utility bill and a driver’s license. Processes from the past too often assumed that all the information was already there. Holds the smart phone over the document,a nd captures the document from the video stream, so it processes and cleans up the image specifically for optical character recognition. Processed on the phone. The image has the coffee stain removed, and made it black and white. This cleanup is a kind of “compression” which is important for mobile and storage.
The documents are scanned in using some libraries for scaning that they make available to put into custom apps. The user does not have to re-enter anything … just take pictures of the documents.
These document “teach” the transformation servers. There is not any coding, but instead teaching. Characters are recognized, and then the fields on the document are recognized as well. There is a manual correction that feeds back to improve the recognition.
Mike Marin – Mobile Case Management and Mobile
Mobile is no longer Optional. Use case is an insurance company that has unhappy customers and decides to implement a mobile app to make customer experience better. Will show content, capture, and case — all on the mobile device.
Again, with the phone, took a picture of the document. After taking the picture, there are some options for cleaning up the document and submit it to the process / case. Can review the contents.
Robert Shapiro – Process Discovery
Take the event logs, and mine them in order to work out a BPMN process model that will have the same statistical behaviors as the event log. Example demo will be on stat orders. Looking at analytics, we see that we are not meeting the objective KPI. First figure out all the paths, and analyze all the variants, and find a critical path. We can see that one step is causing a lot of the delays. Propose two different strategies: one to reduce time, another to reduce costs. After optimization, we meet the delay time reduction criteria, and we see that the second case has better cost benefit values.
Started the demo by opening an event log. This creates a BPMN serialization. He has used the idea of “strict BPMN” to enforce BPMN semantics on the model. It found a top model and two sub models. It created events and gateways which never were in the log — they had to be figured out. He showed a series of different models that had been mined. One had found 3 parallel tasks. The models also look at the data, and find correlations between different data items and paths. This can be used to mine the branch conditions. Can discover a 1hour timer event task as well.
Can even detect manipulations on data if the event log has data captured in the event log.
Q: (Bruce) Impressive to figure this all out from the logs. What happens if 90% of the time data causes a branch, but not 100%.
A: It never requires 100% There are statistical assessments of the conditions. Not 100% precise.
Q: Is simulation the key different from Disco and the others
A: you need to have a complete, executable model if you want to make changes and improve the model.
Tim Stephenson – Omny Link – Toward Zero Code
Looked at a bunch of self-coding but found that things didn’t work too well. Going to focus today on decisions. WordPress claimed that they ran 1/4 of the web. “Firm Gains” is a business to sell your business. First step is to build a form. Then a decision table.
Demo started by logging into wordpress. Edited what looked like a blog post, but it was actually a form. Standard list of fields without an programming. Went to another page, inserted a square-bracket-style wordpress tag to include the form, and it appeared. Very easy, very simple, workflow processes.
Scott Francis – BP3 – Sleep at Night Again
How to automate static analysis for BPM. Your design team is not always experienced. Once they select technology, they implement, and many times it is over engineered and hard to maintain, and sometimes has to be throws out. In reality, the kruft does not get added all at once, instead it is incremental. Neches is a tool to analyze the code, and find problems early in the iterations, and keep them from building up. Does a complexity measurement on the application.
Q: How complicated is it to create new rules?
A: Not too hard. Today this is not exposed, but internally we find it easy to do, and believe once this is exposed people will find it easy enough.
Linus Chow- Oracle – Rapid Process Excellence
Showed a web console for starting and interacting with BPM applications. Mobile interface as well.
And that is it. On day 3 I had a presentation, and was too busy to focus on taking good notes, and then for the entire session before they sequester the laptop. Overall bpmNEXT remains a place for very forward discussion of new directions, a place that is helpful for me to stay on top of things. The new venue — Santa Barbara — is likely to remain the choice for next year. I am looking forward to it already.