I will never forget a call I had with a client. It seemed like a pretty straightforward request, the CFO of a mid sized company wanted to “modernize” their financial reporting (mind you, she recently got promoted to this role so she wanted to make her mark). Her idea was creating a Balance Sheet in Tableau, instead of using Excel.
“Think about it”, she said during our initial meetings while her eyes were lighting up, “Interactive dashboards, drill-down capabilities, real time updates. This is exactly the kind of digital transformation we need”.
Being that I was engaged on it in freelance basis, I could’ve just said yes, built what she was asking and collect the fees. I have a Finance background so I was familiar with the BS and PL statements and I tried challenging their expectations but somehow they kept pushing that this is the direction they want to take as part of their transformation efforts. Sure enough, I was able to deliver exactly what they asked for. A visually appealing Tableau dashboard branded based on company assets. Executives were really happy.
Part of the services packages I offered was to also help people get familiar with the navigation and besides the documentation provided, I also wanted to see how difficult is for people to navigate and see if there is room for improvements.
Read also - The $200k Dashboard That Nobody Trusted
As I created a form to collect feedback, the number one request was to add an “Export to Excel” button. And the reason is very simple, that is what they are used to and that is where they do their analysis, built their models and even prepared reports for regulators.
In my mind, we essentially crated the most expensive Excel file viewer in the company. If you see it from the perspective of a freelancer, I delivered exactly what the client asked, however executives missed what was actually needed.
The Automation Obsession
I wish example above was an isolated incident. But unfortunately it isn’t. I have seen organizations spend hundreds of thousands of dollars trying to automate processes that worked perfectly fine in the way they were. But after moving up the ladder in corp world, I quickly realized that the pressure usually comes from two places: Executives who are trying to come up with ways to differentiate themself with the “digital transformation’ and IT teams eager to showcase their technical skills.
Just the way you see it happening with AI now. Companies are rushing to implement AI solutions trying to prove themselves as “innovative”, which success is mainly measured based on how many AI models they have deployed rather than the actual business impact and the value it delivers. I have seen teams build AI powered forecasting models that perform worse than the Excel based approach that a business analyst has been using for years.
I think there is only one question we need to ask ourselves before proceeding with automation: “Does this actually need to be automated?”
If you find this valuable, consider subscribing
When Excel Wins
Here is the uncomfortable truth that most people get surprised: some of the largest corporations run critical processes on Excel spreadsheets. It is not like they cannot afford better tools, but believe it or not, Excel is actually the right tool for the job.
Excel excels (you see what I did there!.. dad joke) when:
The person who builds it, is the person who uses it
The logic is complex and changes frequently
You need the flexibility to model different scenarios
The process involves creating problem solving, not just data processing
The Balance Sheet dashboard? It was just fine in the Excel territory. Financial Statements are standardized, the logic rarely changes and the real work happens in the analysis that follow, which everyone was doing in Excel anyway.
The Real Cost of Over-Automation
Over automation comes with a costs, which for the most part are hidden costs for a lot of companies:
Maintenance overhead: Dashboards require ongoing maintenance, server resources and license fees. Excel does not require most of these.
Change management: When accounting policies change, updating the Excel file takes minutes. Updating the Tableau dashboard requires going through development cycles, testing and deployment process which can take weeks.
User adoption: Despite months of trainings and proper documentation, users never fully adopted the dashboard I created becausr it did not fit their workflow. They kept requesting for features that would allow them esentially do the same all over again in Excel.
Opportunity Cost: While I built something which I was tasked to do as a freelancer, I built something that nobody wanted and the real problems were still unsolved.
A Framework for Automation Decisions
Seeing the same mistakes over and over again, I developed a simple framework to evaluate whether a process should be automated or not. Before any automation project, I ask those questions:
Frequency & Volume
How often this process runs? (FYI, daily processes have different ROI than quarterly ones)
How many hours per month are currently spent on doing this?
How many FTE (full time employees) are involved?
Complexity & Change
How complex is the underlying logic?
How often do the requirements change?
Does the process require creative judgement or is purely mechanical?
Data & Infrastructure
How many data sources are used to currently to perform this exercise?
Is the data already clean and consistent?
Do we have the infrastructure to support the automation solution?
User Needs
Who will be using this?
How do they currently work with the output?
What is their technical skill level?
Business Impact
What specific business problem are we solving here?
How will we measure success?
What happens if this fails?
Score each category from 1-5. If the total score is below 15, the process probably belongs in Excel.
The Balance Sheet project? It scored an 8.
The Hard Truth
Sometimes the best tech solution is no tech solution at all. Sometimes the most innovative thing you can do is sticking to a spreadsheet you have been using for years and does the work just fine.
This does not mean we should never automate. When processes are repetitive, high in volume, involve multiple data sources, serve many users etc., then automation is great for those instances and it can really add value. But automating stuff for the sake of automation, we are basically creating expensive solutions to problems that did not exist.
The next time someone asks you to build a dashboard, ask yourself “Is this actually better than a spreadsheet?”, you might save your organization a lot of money and frustration by doing so. Btw, if you proceed further with a dashboard, then you should read the article I recently sent out on BI Strategy For Leaders.
And if you are feeling pressure to implement AI just because everyone else is doing it, remember that the most sophisticated solution is not always the smartest one.