Calculating the

ROI for Process

Automation
1
 Table of contents
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Determining your baseline . . . . . . . . . . . . . . . . . . . . . . 4

Accounting for costs of

interruptions and waiting . . . . . . . . . . . . . . . . . . . . . . . . 7

Measuring value of automation. . . . . . . . . . . . . . . . . . 8

Calculating value per

automation run. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Showing how automation improves

more complex workflows. . . . . . . . . . . . . . . . . . . . 10

Calculating value when automation

becomes impossibly good . . . . . . . . . . . . . . . . . . . 12

Computing value of
improved operations. . . . . . . . . . . . . . . . . . . . . . . . . . 13

Reporting on business value and ROI. . . . . . . . . . . . 15

Other ways to show the

value of an automation program. . . . . . . . . . . . . . . 16

Commissioning a PagerDuty
Business Value Assessment. . . . . . . . . . . . . . . . . . 17

Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2
 Introduction
It takes more than excellent ideas to secure funding for an IT automation project. To justify starting or
expanding a project in today’s economy, you must demonstrate both current and future value, and
how the project will support key business initiatives in service of your company’s goals. Fortunately,
automation projects with PagerDuty Process Automation deliver a significant Return on Investment
(ROI). This e-book provides expert tips and ideas for showcasing the business value of PagerDuty
Process Automation projects.

Often some creativity is needed to demonstrate the ROI of an IT automation project. The value of
automating a business function varies based on what is being automated. Some examples of per
run value of automation include:

• A
 Fortune 500 financial services firm that saves 1 hour of personnel time per incident when a
triggered remediation can resolve a problem.
• William Hill reduces time to patch by 95%

Sometimes, what needs to be measured isn’t just actual runs of automation, but the impact or
effect of many more runs than is possible with human-driven workflows. Here is an example of an
Australian Bank that increased deployments to 500 to 600 per month thanks to Process Automation.

In this paper, we’ll cover what baseline measures you should gather for your as-is business, from
KPIs for business processes to benefits of the workflows you are looking to automate. Then we’ll
discuss how to determine useful metrics showing the impact of automation. Finally, we’ll suggest
how to package these up so that you can show your ROI and business value of automation projects.

3
 Determining your
baseline
In their purest form, Process Automation projects that resolve incidents and service requests
can reduce labor costs and waiting time by as much as 95% to 99% for fully automated tasks as
compared to previously human processed tasks. The trick to showing actual value delivered is
understanding how to report this savings in the context of more complex business workflows.

First, you’ll want to create a baseline of your as-is state. Most organizations define Key Performance
Indicators (KPIs) for business processes and departmental functions. Ideally, you should track the
value generated by your automation project in relation to these KPIs. As a next step, collect more
detailed statistics about the specific workflows you intend to automate. For the purpose of this
discussion, let’s define a workflow as a sequence of human and computer-based activities. Here
are some examples of metrics you’ll want to compile from your ITSM systems and other sources.

For an IT operations team, key KPIs you will want to relate business value to include meeting
customer and internal SLOs, as well as responsiveness and costs. Detailed statistics you may wish
to capture for operations center request workflows include:

Metrics for IT Service Request Workflows

Metric Definition Units

MTTC (Mean Time Mean Time to Close or Complete a request. Lapsed time. If there’s a known cost
to Completion) This is average time spent waiting by for waiting time, use cost
requestors, inclusive of any idle time units instead.
when no work is being done.

Processing time Elapsed time it takes to do the work for Lapsed time
completing a request, not including idle
time when no work is being performed.

Cycle cost The total cost of performing the work of Currency

the request.

Workflow productivity The number of requests completed or sum Scaler number / time period or
of total benefit generated in a given time. total benefit / time period

Standardization % of total requests following standard Percentage

procedures or runbooks.

Resources needed Number and roles of personnel involved in Number of people by job role
processing request workflow from L1 to L3.

4
 For Incident resolution workflows, you might recognize these terms as:

Metrics for Incident Resolution Workflows

Metric Definition Units

MTTR (Mean Time Mean cycle time for completing incident Waiting time
to Resolve) resolution workflows to resolve incidents.

Processing time Time spent managing incident response, Elapsed time working, not including
triaging and diagnosing problems, and idle time where no work is
remediating issues. being performed

Processing cost Total cost in personnel time to resolve Personnel hours or currency
an incident.

Incident response Number of incidents resolved in a given Scaler number / time period or
productivity period of time. total benefit / time period

Error budget Cost in hours of downtime, or lost revenue or Currency or time / incident or
consumed business productivity due to downtime. total / time period

Standardized triage Percentage of incidents occurring on %

systems with runbooks standardizing
diagnostics & rule out procedures.

Recurring remediation Percentage of incidents that are known %

recurring problems resolved by standard
remediation in runbooks.

Escalations Percentage of incidents escalated to L2 or %

L3 support.

Additional metrics you may wish to capture include:

• I dle time: average time requests sit idle with no work occurring on them – useful for calculating
total wait time
• Throughput: a capacity metric for simultaneous processing of workflow cycles if available
capacity is a bottleneck in meeting SLAs or urgency requirements.
• Cycle benefit: when a per run value can be attributed to a workflow, useful when that benefit is
materially greater than the processing cost
• Quality: per cycle error ratio when errors related to the manual processing of the workflow are a
source of increasing cost or delay

Once you’ve captured these metrics, you can also compute the following benefits:

• P rocess Productivity: Total completed cycles, or total benefit produced by a workflow over a
given period of time. (Example: incidents closed / month, or $3.2M liability reduction / year)
• Process Efficiency: Total completed cycles, or total benefit produced by a workflow, per person,
over a given period of time. (Example: incidents closed / responder / month, or $3.2M liability
reduction / person / year)
5
 Comparing Process Productivity and Process Efficiency values are a great way to show before and
after effects of automation.

Here’s an example of calculating these metrics for a workflow based on a use case shared by
Brandon Kresge, SRE Manager at Fox Corporation in his session at PagerDuty Summit 2022. In this
before case, he shared a manual workflow his engineers used to set up new monitors in Datadog
and connect them to PagerDuty. In this case, between 15 to 30 minutes of an engineer’s time was
spent completing all the steps of the workflow, Processing Time, which we can also attribute to the
Processing Cost of running one cycle of the workflow.

Before (manual):

“Create Monitor”

Completion time

Request Build monitor Set Add title & Add tags Add priority
posted query thresholds message

Idle time Processing time: 15 - 30 minutes

Processing time: 30 minutes Cost: 30 minutes

In this use case, after applying human assisting automation for setting up Datadog monitoring
alerts, Fox was able to reduce Completion Time by up to 25 minutes, and reduce Processing Cost
by 83%.

After (assisting automation):

“Create Monitor”

Completion time

Request Build monitor Set Add title & Add tags Add priority
posted query thresholds message

Idle time Processing time: 5 minutes Processing time: seconds

Completion time: reduced 10 - 25 minutes Processing time: 5 minutes + X seconds Cost: 5 minutes

6
 Accounting for costs of interruptions and waiting
Humans are not machines. There’s an inherent overhead in not being able to focus on a subject
at hand. In human-driven workflows, there are two major sources of context-switching overhead:
being interrupted in what you are doing to work on something else urgently, and having to wait to
get something you need to finish a task.

Cost of interruptions: Even if your job is 100% focused on taking calls and various requests, it takes
time to reset your brain to be able to focus on a new, or resuming a prior task – context switching.
This includes safely putting down the prior work, and then ramping up to learn the context of the
next task, or catch up on a resumed task. Conservatively we estimate that changing tasks imparts
an additional 20% overhead on the time spent on a task. So the cost of human Processing Time
can be considered to be 120% of actual time spent doing work.

In the prior example, creating a Datadog to PagerDuty Monitor, when fully manual, could be
considered to have a Processing Cost of 18 to 36 minutes as a result of the additional 20%
overhead. With human assisted automation, this “fully loaded” Processing Time is reduced to 6
minutes, including 5 minutes of actual work being done by a human.

Cost of waiting: In theory, if you have to wait for something to be made available to you, or are
dependent on a task being completed for you, there’s overhead imparted. This comes from
switching to other work while you wait for the dependency to be met. We estimate 40% as an
average for “non-recoverable” wait time that can’t be used for other value adding activity due
to context switching costs, planning overhead, misaligned schedule windows, communication
inefficiency, request follow-up, or other reasons.

This could be added on to the Completion Time as an additional factor similar to Idle Time where
no work is being completed.

The longer the wait time or the expected SLA, the less this penalty might be. We use urgency and
SLA to scale this factor:
• Urgent, 24 hours or less SLA: 40%
• Medium, 1 week or less SLA: 25%
• Low, 2 weeks or less SLA: 10%

So, a QA engineer’s time waiting 24 hours to gain access to test deployed software for a problem
might be seen as costing 9.6 hours (or 3.2 hours if we’re only counting business hours). This would
be calculated as a Completion Time of 33.6 hours. However, a developer waiting 2 weeks to be
issued a new VM in a well established process with known SLA could be considered to have lost
only 1 day in waiting. This would be calculated as a Completion Time of 15.4 days, or 11 days if we’re
only counting business days.

Eliminating interruptions and waiting times is another way to reduce overhead in a company, and
automating tasks is a great way to achieve this.

7
 Measuring value of
automation
Let’s explore a number of scenarios about how you might measure and report the value of your
automation projects.

Calculating value per automation run

The simplest model for calculating the business value of automation assumes value is generated
whenever an automation runs. An excellent example is this use case shared by Robert Powers from
Brinks at PagerDuty Summit 2022. They counted labor savings per automation run for a recurring
data transfer job that previously took a staff member 5 to 10 hours to complete manually.

As-is Process
(Manual step-by-step)

Import to
other SaaS

Export Transform Encrypt Enterprise

data data data file transfer

Analytics /
DW

By automating the process with PagerDuty Process Automation, they turned this process from being
¼ of one person’s job every week into an automated task that takes zero human time.

To-be Process
(Fully automated in PagerDut y Process Automation)

SaaS
Retrieve
data
PagerDuty
Process
Automation
Enterprise
file transfer

Export
data Encrypt
data Review /
Analytics

8
 Cost – 5 - 10 FTE hours / week = 520 FTE hours / year
– Cost to automate: 20 FTE hours

Opportunity – 99% reduction in FTE hours

– Immediate payback
– Errors more easily traced & fixed, process runs more more frequently ensuring up
to date operational systems

Complexity – Moderate complexity: multiple steps, systems & departments

– Potential simplification by standardizing data transfer processes

Risk – As is: stale data in operational systems & human error

– To be: Programmatic errors that pose low risk
– Mitigation: immediate notification of automation failures

Maturity – Well documented manual process

Business benefit – Ops systems updated in less than a day, versus 1 week lag, IT resource savings

The table above offers some data to compute some further metrics. Depending on how you
might define the before-case Completion Time, you can compute the duration of a cycle from the
scheduled start time. Or, if availability of the human resource is a limiting factor, you might argue
the Completion Time was 7 days, being reduced to ½ a day considering the 2X per day automated
execution. Processing Time was reduced by more than 99%. Processing Cost fell from 10 hours to
less than .03 hrs per run, averaging in the time to automate the workflow.

Completion time: 10 hrs to 0 hrs Or: 7 days to ½ day

Processing time: 10 hrs to 0 hrs

Processing Cost: 10 hrs to 0.03 hrs

Throughput: Increased x 14

Quality: improved 

Productivity: 14 updates / wk Before: 1 update / wk

14X increase in productivity

Efficiency: 36 updates / human / wk Before: 4 updates / human / wk

9X increase in efficiency

We can also see how increasing frequency of execution increases workflow productivity by 14X,
while at the same time, replacing human effort improves efficiency by 9X.

Process Automation and Runbook Automation have a feature called the ROI Metric Data Plugin that
makes it easy to compile per run data for future analysis and reporting. To learn more about it, read
the blog Calculating the Business Value of Automation in PagerDuty Process Automation and check
out the Process Automation documentation.

9
 Showing how automation improves more complex workflows
Automated tasks do not exist in a vacuum. They are part of a business workflow, with inputs,
outputs, and potentially some human interaction. This means showing the business impact on a
business workflow requires computing composite metrics. Many of those metrics will be compiled
in the systems that initiate or end a workflow.

Let’s look at another example use case shared by Robert Powers from Brink’s at PagerDuty
Summit 2022.

Here is their as-is workflow for provisioning and deployment of virtual machines, which Robert
reported was typically a 2 week process upon request – mostly waiting for technical specialists to
pick up tasks in tickets.

As-is Process
(Automation orchestrated by IT Ops)

Build & valida te

Deploy ( per region)

User SCM

ITSM / ITSM / VM
ITOM VCSA
ITOM
Monitoring Validations /
tests
VM
IT Ops IT Ops

They utilized PagerDuty Process Automation to automate the vast majority of this workflow, leaving it
to end users to begin a cycle of the workflow as a self-service request.

To-beProcess
To-be Process
(Delega
(Delegated end-to-end
ted process
end-to-end orchestrated
process orchestrated
by PagerDut y Process Automation)
by PagerDut y Process Automation)
Region 1
Region 1

VCSA VM
Valida te
User VM
VCSA
Validate
User
SCM VM
ITSM / SCM
ITOM PagerDuty VM
Process
ITSM / Automation Region 2
PagerDuty Validations /
ITOM tests
Process
Monitoring Automation
system Validations / Region 2
“Automation tests
“Standa rdized VCSA VM
Monitoring as an API” & governed”
system
“Automation “Standa rdized VCSA VM
as an API” & governed” VM

10
VM
 While it should be possible to compute all the previously mentioned workflow metrics here, Brinks
shared that their Completion Time from request fell from 2 weeks to 3 minutes, resulting from
elimination of Idle Time and greatly reducing Processing Time and cost. This is one key benefit of
self-service, scheduled, or triggered automation – huge reductions in Completion Time and Idle
Time. Sure, a requester can work on all manner of back burner projects, but they are not able to
immediately complete the urgent task that caused them to log a ticket.

Not all workflows can be fully automated, because in many cases human interaction may still be
required. For example, in describing the benefits of automated incident resolution with PagerDuty,
we state that automatic triage can save 15 to 30 minutes off MTTR (Mean-Time-To-Resolve – the
Completion Time for resolving an IT incident in an incident response workflow). Even though
diagnostic automation likely speeds up specific tasks by 99% or better, it’s possible that a net
new or infrequent fault will require human investigation and remediation assisted by automation,
whereas a well known problem might be able to be fully remediated automatically 99% faster.

Other examples of improved workflow metrics:

• ResultsCX sees MTTR for network failovers reduced by 95%

• Trimble sees self-service IT requests reduce Cycle Times from 5 days to 5 minutes
• An Australian Bank sees an 85% reduction in maintenance time for software updates

11
 Calculating value when automation becomes impossibly good
Consider the first use case again where Brinks reduced the personnel cost of a weekly data transfer
process by 99%. This process doesn’t just save a person’s time, it also runs considerably faster. This
then makes it possible to run automation far more often than might occur for a human-powered
task. In the case of Brinks, their update frequency increased from weekly to more than once a day.
Freshness of data has an intrinsic business value that is beyond the scope of this paper. However,
the more frequent updates in this case could be compared to having more people manually
running this job.

10 human hours per cycle x 14 cycles per week = 140 hours total cost

What if we were to run the update process once every 5 minutes, 24 hours a day, 7 days a week?

10 human hours per iteration x 2016 iterations per week = 20,160 hours saved, which is an
impossibly good result. It would be nonsense to report labor savings of 500 employees per week in
this case.

As you can see, labor savings have diminished value the more often an automation runs.
One Process Automation user offered an alternative approach for such a scenario: don’t measure
each update execution, measure every instance of someone pulling that data as if a human had to
gather fresh data for them right then.

So while there might be 2016 updates per week, maybe a downstream user looks at the data twice
per day, 5 days per week. In such a case, the value you would report would be 100 hours of
cycle savings.

To track such a metric, you would need a log of whenever the resultant data store was accessed.

From that number, we can generate our workflow metrics.

Completion time: 10 hrs to 0 hrs Or: 7 days to 0 day

Processing time: 10 hrs to 0 hrs

Processing Cost: 10 hrs to 0.19 hrs

Throughput: Increased x 10

Quality: improved 

Productivity: 10 updates / wk Before: 1 update / wk

10X increase in productivity

Efficiency: 26 updates / human / wk Before: 4 updates / human / wk

26X increase in efficiency

While similar to the prior scenario’s metrics, it’s interesting to note that Cycle Time is 0 since the
data is available on demand, similar to self-service automation. Actual value is similar to the value of
a cache: how often is cached data hit vs. a miss.

12
 Computing value of improved operations
A lot of batch processes involve validation and verification of data or configurations. While it’s
possible to measure labor saved by not having a human make these checks, it’s not a satisfactory
measure of the potential business value provided. For example, if a validation process greatly
reduces liability caused by missed deadlines, it would be more appropriate to express the business
value delivered as that reduction in liability.

Example:
A government reporting deadline imposes a $1 million fine if a report is not filed within 3 days of a
certain class of incident. An automated process is created that checks through incident history on a
daily basis looking for unfiled reports of class member incidents, and generates an aging report. You
could report the daily hours saved by not having a human do the checking and reporting. Or, you
could report a reduction in liability by the number of incidents in the class, and any history of fines
being levied before automation. Perhaps before automation, out of 500 reportable incidents per year,
2 deadline violations occurred. One could then claim a potential liability of $4000 per incident and a
0.4% exception rate averaged over all 500 cases. If automation reduced this error rate to 0 even as
reportable incidents increased to 800, you could claim a liability reduction of $3.2 million.

Similarly, a verification process might also improve the quality of operations. Here are a couple of
examples provided by Brandon Kresge, SRE Manager at Fox Corporation in his session at PagerDuty
Summit 2022. His team utilizes Process Automation to verify and validate the configuration and
activity of PagerDuty users:

• To see a 20% improvement in users being set up for on-call readiness:

Manual
75% of on-call users ready

Generates
Auto- Report
Automated Readiness
95% of on-call users ready

Sends Slack
Notification

Notif y Users

13
 • T
 urning a weeks-long auditing process of looking for inactive PagerDuty users into a 10 minute
weekly report enabling them to repurpose inactive seats for new users.

Manual
10-20 minutes ENG
time per user
Notification
Rules Generates
Report
Automated
10 minutes ENG time
for ALL users

On-Call Shifts
Human
Review
Subscri ptions

Audit Trail

It wouldn’t take much to convert these metrics into downtime and end-user on-call reductions, as
well as cost savings for the company.

Another example of improved operations includes:

• Trimble is able to deploy an automated remediation for a new cause of incidents in half a day

14
 Reporting on business
value and ROI
If you’ve followed the advice so far, you should already have a variety of potential metrics to show
improvements brought by automation to targeted business workflows. You should also know how
you will compile them. While the detailed statistics may be interesting to some stakeholders, others
will just want to know the overall benefits.

To summarize this impact, you’ll need to know the relevance and contribution of your automated
workflows to the top-line KPIs of the business processes they help implement. This is highly specific
to the workflow, the business process, and your particular organization. Here are some guidelines.

1. U
 nderstand the nature of the business value of the workflow. Is it a core process related to
generating or fulfilling revenue, such as sales, e-commerce, delivery, or invoicing? Is it a
supporting process related to business expenses or cost control such as procurement, accounts
payable, or IT productivity? Or is it a management process concerned with governance, reducing
risk, and ensuring compliance such as IT security or HR?

2. W
 hat is the cycle value of the workflow you are automating, or the impact of its iterations? Per
the metrics we discussed above, it should be possible to show cost improvements, faster
processing times, increased capacity, improved quality, higher productivity & efficiency. You will
want to relate these to the business function KPIs of the process.

For example, perhaps departmental goals include reducing the cost of sales or speeding up
quoting, so reduced personnel costs of sales operations and shorter Completion Times for quote
generation fit these KPIs. Another example in the case of customer support, if SLAs are consistently
met with customers, improving profitability through higher productivity and efficiency of customer
support workflows may be an impactful automation.

15
 Finally, there is creating an analysis of the Return on Investment. For a monetary ROI calculation
you’ll need the following metrics:

1. R
 educed costs of workflow execution
2. Reduced cost from reduction in errors
3. Increase in business value generated by workflow through better operation such as faster cycle
times, higher throughput (example: higher data quality, faster software releases), ideally related to
KPIs of the relative business process
4. Opportunity opened by freeing people to do other higher value tasks – add further cost savings
or business value generated by additional tasks
5. Cost of your automation environment
6. Cost of time spent automating each workflow

Calculating the return on investment, or your net benefit is simply a matter of adding together
benefits and subtracting out investments:

Reduced Costs +
Business value generated +
Value of other opportunities opened +
Cost of the automation environment -
Cost of time spent automating each task -
===================
= Net benefit

You may wish to include qualitative improvements that do not translate so easily to costs such as
improved employee morale, improved customer satisfaction or net promoter scores.

Other ways to show the value of an automation program

In his talk from PagerDuty Summit 2022, Jamie Vernon of ResultsCX offers a few additional program-
level metrics you might want to compile:

Total automation runs and their results How automation runs were initiated

• S uccessful runs, • Senior / specialist resources

successful results • Junior / generalist resources
• Successful run, • Scheduled
unsuccessful results • Triggered
• Failed to run

This makes it possible to show the full scope of an automation program such as:
% Requests automated % Requests invoked by % Requests available as

Other / unknown Humans

not automated
Tickets
Standa rd requests
not automated

Standa rd requests APIs Self service

automated operations

16
 Commissioning a PagerDuty Business Value Assessment
While it’s certainly possible to create your own business value and ROI analysis, you might find it helpful to work
with the Value Engineering team at PagerDuty. This is especially true if you are setting up an ambitious program.
Our Value Engineers have experience with a wide variety of customer use cases incorporating the PagerDuty
Operations Cloud, and can help you create a detailed forecast of potential benefits specific to your business
needs. If you would like to have an assessment of potential business value of your automation program as
depicted below, please contact your Account Executive.

Sample Business Value Assessment

Annual benefits by value category

(Ramped benefits over 3 years)

Cost vs. Benefit=Return

Potential Impact
(Benefits over 3 years)

$1.6M $142.1K 354% ~12

3 Month cost ROI Payback period
Total benefit (Benefits vs. Cost)
of delay (Weeks)

Annual savings of $511,523 based on year one investment of $118,800.

17 *Estimated cost and savings based on model only, not a substitute for proposal.
 Conclusion
The key to demonstrating business value for a PagerDuty process automation project is
identifying the right metrics for the workflows you’re automating. It is also important to tie
them to KPIs and goals for the overarching business processes they support. Establishing
a baseline, creating your forecast, and tracking the project’s impact over time can help you
demonstrate the value of automation to stakeholders across your organization. Focus on
the benefits that matter most to your audience and tailor your approach to meet their
specific priorities.

Learn more about PagerDuty Process Automation.

Contact your Account Executive to ask about a
Business Value Assessment of your automation program.

About PagerDuty

PagerDuty, Inc. (NYSE:PD) is a leader in digital operations management. In an always-on world, organizations of all sizes trust
PagerDuty to help them deliver a better digital experience to their customers, every time. Teams use PagerDuty to identify issues
and opportunities in real time and bring together the right people to fix problems faster and prevent them in the future. Notable
customers include Cisco, DocuSign, Doordash, Electronic Arts, Genentech, Shopify, Zoom and more.

To learn more and try PagerDuty for free, visit www.pagerduty.com. Follow our blog and connect with us on Twitter, LinkedIn,
YouTube and Facebook.

18 2023 PagerDuty® // All Rights Reserved.