ROI: Are You Really Reducing Costs – or Just Reallocating?

Cartoon of Man Carrying a Large Bag of MoneyBefore jumping in, I should say that I am an unabashed fan of technology in both the personal and business contexts. My back is thankful for garage doors that open with the click of a button, and the survival of grass in any recognizable form outside is a testament to scheduled, programmed watering. I’ve benefited in my work life from analysis and visualization tools, grammar and spelling checks, and all manner of calendars, task-trackers, and reminders.

With this technophilia in mind, I can’t help but think that we’re doing a disservice to these technologies when we become overly – and inaccurately – focused on cost savings. Specifically, I’m thinking about ROI cases for AP automation premised on per-invoice cost savings. We’ve all seen the numbers, usually some variation on $2 for top performers, $7 for the middle-of-the-road, and $15 for those with some work to do. I’ve published them myself and agree that they can provide a helpful shorthand for improvement potential.

But these numbers aren’t perfect, and they may be affirmatively harmful if you attempt to extrapolate from these per-invoice figures up to your normal processing volumes. Here’s what I mean:

Imagine that you have two AP Clerks handling your incoming invoices, each with an average salary of $39,500 (the midpoint salary in Robert Half’s 2014 Salary Guide for Finance).1 Your first shot at costing may ask a simple question: how much time does one of our clerks spend processing an invoice? They do a good job and are able to process 100 invoices per day, so they contribute $1.99 cost ($51,675 / 260 business days per year / 100 invoices per day).


That’s not that much.

Let’s give them a boss. According to Robert Half, that’ll cost us somewhere around $65,750. That’s $1.64 more per invoice ($85,475 / 260 business days per year / 200 invoices per day total) – or $3.63 total. We could add in technology costs, but in the paper environment we’re assuming they don’t have any. I suppose we could assume that all of their payments are made via international wire transfer, pushing the price up by $35-40 per transaction, but that’s cheating.2

So how could their processing cost be $10 or more? They’d need to be processing fewer invoices, of course. If they only received 100 total invoices per day, that could spike the cost to $7.26! But then they’d probably only need one AP clerk, so it’s more likely that the cost would top out at $5.28. On top of that, we’ve assumed that 100% of his/her time is spent processing invoices.

What’s going on here?

First, we haven’t accounted for the time spent reviewing invoices by line-of-business managers. We haven’t accounted for smaller line-items like AP’s share of IT support costs, phone lines, fax lines, internet access, or computer hardware (such as it is in a paper environment). This will add to the measure, but not by much. Remember: in this example, any added cost will be divided by at least 26,000 (100 per day @ 260 days) to get your per-invoice cost.

On the other hand, we haven’t accounted for time not spent processing invoices – say, on employee expense reports. This would just decrease the labor cost allocated to invoices, further reducing the figure. What these difficulties underscore is that when you attempt to build a per-invoice processing cost from the bottom-up, you may wind up with something that is less than the sum of its parts. If you aren’t careful to capture everything, you’re really just reallocating costs away from invoice processing – but not actually helping the bottom line.3

Where does this leave us?

There are two main takeaways that I’d like to end with. First, you should undertake a careful and detailed top-down analysis when calculating your invoice processing costs (say, with the handy Worksheet and explanatory Guide available for download free from Blue Hill). Second, when evaluating a prospective service or solution, you should be specific about which cost category (or categories) it will address and what direct impact it will have. Remember, if your new approach will allow Dave to process 10% more invoices in a given period of time, you’ll still be paying 100% of Dave.4



1. For both this example and the managerial position coming up in the next paragraph, I use a 30% mark-up on top of this base salary to approximate the employee’s fully-burdened cost (health care, vacation, additional employer obligations, etc.) That’s why the $39,500 salary morphed into $51,675 for cost calculation purposes.

2. This would be cheating for two reasons: 1) we’re not incorporating payment into our invoice processing cost calculation, and; 2) international transactions open up an entirely separate discussion of trade finance and associated fees (ex. Your international supplier may have a domestic collections account, thereby avoiding the need for an international wire transfer).

3. And that’s my central point for today. If you can conduct time and motion studies that account for 100% of each employee’s time and properly apportion any downtime (assuming sub-100% capacity) across their various tasks, then you will be safe. If not, you run the risk of expecting hard-dollar savings from changes that lower time used without actually lowering time compensated.

4. In a two-person department, you would need to double processing throughput in order to reduce headcount by one. Even in larger organizations where a 10% efficiency gain could be matched to a whole number of staff for reduction, the smart management decision may be to redeploy them to other tasks rather than terminate them. In that case, the real (and difficult) question is how much value the new activities provide, since the efficiency gain did not result in a corresponding change to actual expenses.

Posted on by Scott Pezza

Leave a Reply

Your email address will not be published. Required fields are marked *

Latest Blog

Blue Cedar Puts Mobile Application Security Far Ahead of MDM Apple iPhone X Highlights Enterprise Corporate-Liable vs. BYOD Conundrum Blue Hill - AOTMP 2018 Q1 Agenda

Topics of Interest

Advanced Analytics




Artifical Intelligence


Augmented Reality



Big Data


Business Intelligence



Cognitive Computing

Corporate Payments

Data Management

Data Preparation

Data Wrangling





design thinking


Emerging Tech

enterprise applications

Enterprise Mobility

Enterprise Performance Management

enterprise video

fog computing

General Industry



Hadoop World

Human Resources


IBM Interconnect




Information Builders


Internet of Things





legacy IT


Legal Tech

Log Data

Machine Learning

Managed Mobility Services


Mixed Reality


Mobile App Security

Mobile devices

Mobile Managed Services







Predictive Analytics

Private Equity



Questioning Authority

Recurring Revenue

Risk Management


Sales Enablement



service desk

Social Media



Supply Chain Finance

Switchboard Software




Telecom Expense Management




Unified Communications


USER Applications

User Experience

User Interface

video platform

Virtual Reality



Wearable Tech