Process Automation: Automate Where Returns Are

Current Practices in Process Automation

To make their operations more efficient, organizations continuously look for process improvement and automation opportunities. Traditional approaches to process automation begin with conducting a due diligence project to build an understanding of the current as-is process through

Floor walks and
Capturing inputs from process owners/users/participants etc.

The cross functional nature of large processes has an impact on capturing the actual on ground processes in detail. It is difficult to assess true process performance inhibitors as each function generally cites a dependency or bottleneck on others and true ownership is a contentious issue. Teams conducting the due diligence find it rather difficult to resolve and get agreement from stakeholders.

Another challenge that often occurs is that End to end processes run supported by several systems (IT stacks) which at times are

Under transformation or
Migration or
Under Update

This adds to the complexity of process due diligence and subsequently lead to challenges in assessment for automation opportunities. Rarely do we focus on the true process performance and hence derive our automation implementation roadmap with a focus on performance.

Selection and prioritization of process automation opportunities

Two primary parameters are used to prioritize automation opportunities:

Ease of implementation and
Potential benefits and ROI

This approach identifies low hanging automation opportunities which are easier to implement and deliver valuable benefits post implementation.

Though the approach sounds reasonable, it is generally implemented via an analysis within a very narrow context i.e. an automation use case. The aforementioned two parameters are used for prioritizing automation opportunities without ascertaining the impact on the overall process performance.

This leads to a “local” optimization and appears to be beneficial superficially but may not result in any enterprise wide benefits.

There are three main challenges with the current approach of selecting and prioritizing information opportunities:

Real process performance is not understood/analysed properly in the due diligence phase
We do not select/prioritize the automation opportunities to make maximum impact on the overall process performance, rather the narrow focus is only to remove the manual efforts and address low hanging fruits/immediate outcomes/alignment with some ongoing transformation programs
Rarely do we think about re-engineering the process to focus on addressing the bottlenecks prior to automation

The ultimate goal for automation should be to increase throughput from the overall process rather than reducing processing time for one of the activities i.e. Optimization for the end-to-end process and chain of activities rather than the local optimization of one activity in the process.

Using a simulation based approach enables identification and prioritization of process automation opportunities. This approach may be further extended as a continuous process automation (CPA) framework.

Simulation and Analysis

To illustrate our analysis and findings, we will consider a simplified order processing business process as illustrated in figure 1. The hypothetical process consists of 4 sub-processes and A3 represents the bottleneck with an existing backlog. The simulation study is conducted under the assumption that all the orders processed are of a similar nature and relate to the same product/ service line. The order arrival is assumed to be linear and not stochastic for simplicity.

The processing time associated with each of the sub-processes that is used for simulation is also illustrated in figure 1.

Automate-Where-Returns-Are_fig1 — Figure 1: Order Processing Example

Process simulation prior to automation is illustrated in table 1. As A3 is the bottleneck with backlog, orders are queued before processing through A3. Note that the queue time before an order enters A3 and subsequently cycle time (time to process an order end-to-end) goes up with every order due to the bottleneck.

Automate-Where-Returns-Are_tab1 — Table 1: Order Processing and Cycle times prior to Automation

Next we will simulate scenarios considering automation for each of the steps in this simplified process and analyze the impact on queue time, cycle time and throughput the process.

Scenario 1: Sub-Process A1 is automated

For simulation purposes it is assumed that after the automation sub-process A1 takes only 2 minutes per order compared to 5 minutes per order prior to the automation. Table 2 illustrates Order processing and Cycle time simulation for this scenario.

Sub-process A1 is at the start of the process and automating it would result in orders reaching to A3 earlier than before and thus the automation will result in higher queue time and cycle time. While comparing simulation results with the baseline data in table 1, It should be observed that throughput of the end-to-end process remains the same in spite of automating steps in A1.

Automate-Where-Returns-Are_fig2 — Figure 2: Sub-Process A1 is automated

Automate-Where-Returns-Are_tab2 — Table 2: Order Processing and Cycle Times Simulation After Automating A1

Scenario 2: Sub-Process A2 is automated

We assume that post automation processing time of sub-process A2 is halved to 2 minutes. Sub-process A2 is a pre-cursor to the bottleneck activity A3 and its automation will have an effect similar to automation of A1 i.e. orders reaching to A3 earlier than before thus higher queue time.

Table 3 illustrates Order processing and Cycle time simulation for this scenario. It should be observed that throughput as well cycle time does not improve as a result of this automation. Thus automating sub-process A2 has no effect whatsoever.

Automate-Where-Returns-Are_fig3 — Figure 3: Sub-Process A2 is automated

Automate-Where-Returns-Are_tab3 — Table 3: Order Processing and Cycle Times Simulation After Automating A2

Scenario 3: Sub-Process A3 is automated

Sub-process A3 is the bottleneck. We assume that post automation the process takes 3 minutes as illustrated in figure 4. Processing and cycle time simulation results are detailed in table 4.

Automate-Where-Returns-Are_fig4 — Figure 4: Sub-Process A3 is automated

Automate-Where-Returns-Are_tab4 — Table 4: Order Processing and Cycle Times Simulation After Automating A3

Simulation results suggest that Order cycle time and queue time are reduced and process throughput improves significantly. It can be demonstrated by extending the simulation that as a result of the automation, 24 orders can now be processed in the same amount of time as it took to process 10 orders before automation. It can also be demonstrated that from order 12 forward, orders are processed as they arrive at A3 and queue time is eliminated.

Scenario 4: Sub-Process A4 is automated

In the last simulation scenario, we analysed the automation of sub-process A4 assuming processing time of A4 is halved to 1-minute post automation.

Automate-Where-Returns-Are_fig5 — Figure 5: Sub-Process A4 is automated

Simulation results, as shown in table 5, demonstrate a slight improvement in cycle time which is precisely equal to the processing time reduction in A4 as a result of automation. This also results in a slight improvement in the process throughput. There is no change in the queue time before bottleneck A3.

Automate-Where-Returns-Are_tab5 — Table 5: Order Processing and Cycle Times Simulation After Automating A4

Simulation Analysis:

In table 6, we illustrate the summary of our analysis and simulation results. It is clear that the automation interventions introduced in different sub-processes produce different outcomes for end-to-end process. Through scenarios based simulation and analysis we demonstrated that all these automation scenarios do not result in the desired impact on the overall process e.g. in Scenario 2 there is no impact on cycle time and throughput.

In scenario 1 and 2 there is no change in the process throughput in spite of automation. Such automation seems to be a local optimization at the sub-process level but will result in few business benefits and ROI when analysed from a broader perspective.

Scenario 1 also results in an increased cycle time and can thus have an adverse impact on backlog (or inventory in a manufacturing process) after processing time is optimized through automation.

Automate-Where-Returns-Are_tab6 — Table 6: Summary of Analysis and Simulation

Automate-Where-Returns-Are_fig6 — Figure 6: Impact of automation on Cycle time

In Scenario 3, we introduced automation intervention at the bottleneck, and it resulted in maximum benefits to local as well as global optimization.

The key takeaways from the simulation and the follow-up analysis are as follows:

The automation of the sub-process (i.e. local optimization) does not always lead to improvement in the end-to-end process (i.e. global optimization)
The benefits achieved through automation of the bottleneck sub process (in this case A3) illustrate that it is critical to first identify existing bottleneck(s) before automation and target identified bottleneck processes/sub-processes for introducing automation interventions.

Continuous Process Automation (CPA) framework:

Using simulation and associated analysis, we have arrived at a conclusion that to achieve maximum returns through automation, enterprise should follow the following approach:

Within the end-to-end process, identify the bottleneck sub process
Introduce automation or process improvement interventions within the bottleneck sub-process

Automate-Where-Returns-Are_fig7 — Figure 7: Continuous Process Automation (CPA) Framework

As illustrated in figure 7 as a Continuous Process Automation (CPA) framework, the above approach can result in an ongoing continuous analysis and improvement cycle since the bottleneck shifts to another sub-process which requires optimization.

References:

Eliyahu M. Goldratt, J. Cox, The goal: excellence in manufacturing, North River Press.
W. Edwards Deming, Out of the Crisis, The MIT Press.
Ravi Anupindi, Sunil Chopra, Sudhakar D. Deshmukh , Jan A. Van Mieghem , Eitan Zemel Managing Business Process Flows, Prentice Hall.
S. Limam Mansar, H.A. Reijers , Best practices in business process redesign: use and impact, Business Process Management Journal, Vol. 13 No. 2, pp. 193-213. https://doi.org/10.1108/14637150710740455
Manuel Laguna, Johan Marklund, Business Process Modeling, Simulation and Design, CRC Press.
Sheldon M. Ross, Simulation, Academic Press.