General vs Specialized
It is relatively easy to distinguish general and specialized analysis tools. Perfect example of general tool is MS Excel – you get the rows and columns and count, sum and average at the bottom. Plus, you get dozens of functions so with enough patience and skills you may answer more complex questions. The good example of specialized tool is Waze which gives the fastest driving route to any address you specify. Unlike Excel, it does one thing only, but it would be next to impossible to obtain this answer any other way.
The latter is the approach of TimelinePI. The platform is organized around 26 specialized types of analysis, each serving very specific purpose. In few cases, user could obtain the same answer from other platforms after long and hard work. In many case, it would be just impossible.
Let’s look at just three examples of such cases.
It’s hard to find a business running without any rules whether they call them protocols or use some other terms. In general, a protocol is a collection of rules starting from “If XXX happens” and describing what should or should not happen next and when.
The rules could be very strict: “Once A happens, you must have B in 2 hours and then C in 4 hours, then D”, or quite loose: “Once A happens, you must have B and C in any order as long as they happen within 2 hours from A and before D”.
Protocols present two unsolvable challenges for general purpose process platforms.
First, the non-strict rules like “Steps A and B could happen any time between steps C and D, as long as it happened within 3 hours from C” don’t fit into the sequential definition of process schema.
Second, general platforms cannot identify the specific violations of the rules. What exactly went wrong? Did you miss process step? Performed the steps in the wrong order? Broke the time limit? These violations have very different impact on the business and should be treated differently.
Solution: prebuilt analysis in TimelinePI. The product allows user to define the extensive collection of rules and then identifies and alerts on all individual violations.
In virtually all case management, ad-hoc processes, call centers, and similar environments, some decisions of the next step in the process are done by people or maybe even a machine. For example, a customer calls a phone company and explains his problem to the first-line support representative named John. John routes the call to Mary from billing.
Sometimes the process is misdirected. Mary was the wrong person to deal with it. So she returns the call back the first line support department, when a representative named Mark picks up, and then routes the call to Sam, who resolves the issue.
What we have is the process pattern:
PBX – Answer (John) – Route (John) – Return (Mary) – Route (Mark) – Resolve (Sam)
The questions are:
- How after does it happen?
- Who is the person who mis-routes the calls most often?
- Is there a specific department that mis-routes the calls most often?
- Could an automated answering service eliminate call routing and improve the customer experience?
The goal of the analysis is to find all patterns Route-Return-Route-Resolve and break it down by the person who did the first routing. To complicate the analysis, it doesn’t have to the same person who did later routings. Also, the second routing could also be mistaken and therefore be included into our analysis.
There’s simply no way to perform this kind of analysis in any general purpose tool unless it provides a powerful API and a customer has an army of available developers to code against it.
Solution: pre-built analysis in TimelinePI. You define the patter indicating mis-routing (such as Route-Return-Route-Resolve or Route-Re-route-Resolve) and the application find all occurrences of such patter and breaks them down by the guilty party.
Like protocols, workflow exists in most organization. A unit of work is transferred from employee or application to the next one, until it’s done. In general, the workflow process could be viewed as the network of queues. Each work item is taken from a queue for some operation and then placed into another queue.
Workflow process presents the unique analysis challenge. This is just few of the questions required to get a workflow under control:
- How many different queues a process goes through before completion?
- How often a process comes back to a queue where it already was?
- Break down the overall process time by the queues in which the work was done.
- Separate waiting time in the queue from the actual processing time
Once again, answering these questions in general purpose analysis platform is impossible. ABBYY Timeline includes the specialized Workflow analysis providing these answers with a single click.
About the author
Alex Elkin received an MS in Physics from Moscow Institute of Physics and Technology. He co-founded Intelliframe Corp, which was acquired by webMethods. Then he co-founded Altosoft, which was acquired by Kofax. In 2014 he co-founded ABBYY Timeline and serves as its CTO.