How Four Hours in a Post Office Led to My Big App Idea
I went to the post office just to withdraw some money, but I ended up analyzing the queue system and people’s behavior.
I’m going to milk this story dry until I get back the worth of those four hours. This is the first article in a series of milking.
I needed some cash the other day. It was a beautiful Sunday, so I decided to go with my brother early in the morning before the post office opened at 8:00. We got our queue numbers at 8:09. To our surprise, our numbers were 111 and 112. How could that be? The office supposedly opens at 8:00, but let’s be real, you can always add five or ten minutes to that.
We figured this was just because it was the main office. Since we’d already reserved our spots here, we decided to check the nearby branch. It was packed with elderly people, and since we weren’t sure how much we could withdraw there, we didn’t want to risk waiting. Even at the main branch, we had to split the amount between us to avoid any issues.
The Setup
Back at the main office, my brother preferred to stay in his car and work while I went inside and sat on the benches. The office had eight rows of benches, with about eight seats in each row — 64 seats total, and half of them were occupied.
We were facing a monitor at the end of the office that played promotional videos about post services. It also displayed the queue number, and whenever a clerk pressed a button, the number changed, showing the next person in line and which clerk they should go to.
On the left side, there were eight clerk desks, though only four were occupied. Two clerks were handling money-related services, while the other two dealt with packages and other tasks. The queue was for the money-related services only, as the other services weren’t as crowded. But it didn’t seem that way to me. I think the clerks assigned to the other tasks were just much faster, which is why no one noticed a backlog.
On the right, there were two more employees. One was responsible for manually dispensing queue tickets from the broken ticket machine, which had two paper flags on the screen. The employee would click under each flag to print a ticket. The other employee handed out forms for withdrawing or transferring money.
The queue system screen was attached to the cashier’s wall. The cashier handled large sums of money and bigger transactions, while the other clerks dealt with smaller amounts and verifications. If you needed to withdraw a large amount, they’d redirect you to the cashier.
The Queue
At first, I was just observing the queue, but then I noticed some people were taking longer than others. That’s when I decided to track it. I focused on Clerk #6 and logged their activity for the next four hours. With the help of ChatGPT, I prepared the logs with special flags to make everything clearer.
A Miss or Balance Check
Each time Clerk #6 called someone, I logged it. Short interactions, lasting less than a minute, were usually balance checks. These were often done by elderly people or those less familiar with technology. While you can check your balance via the website, mobile app, or even at the terminal with your card, most elderly folks don’t trust technology, and it’s tough to convince them that these tools are meant to make their lives easier.
Sometimes, people missed their turn altogether. Of course, many of them came back later, causing a bottleneck in the queue.
Prepared Customers
These people were a relief. They kept the queue moving, finishing their transactions in under five minutes. Some caused minor issues, but most were quick and efficient.
The Cycle Bottleneck
This is where things got frustrating. These bottlenecks destroyed any hope of getting out quickly. After paying attention, I realized that the people who had missed their turns or checked their balances earlier were returning. They’d fill out forms and come back all at once, overwhelming the clerk and causing delays.
Special Cases
I was surprised to see that there weren’t many people skipping the queue, despite the fact that in our country, people with connections often bypass public service lines. In four hours, I only saw three instances of this — two police officers and one employee from the electricity and gas company. The officers asked the ticket dispenser guy for help, and he went to the clerks to process their requests. The employee just said, “I’m from that company,” and his request was handled right away.
The Red Marks
The red marks are my own. The first indicates when I was called, the second when I was redirected to the cashier, and the third when the cashier handled my request.
The Problem
Our city is a commercial hub, attracting people from all over, including neighboring countries like Tunisia and Libya. While we have many post offices, only three or four are main branches. Most people with large transactions come to these branches, which contributes to the long queues. If you don’t get a ticket within the first hour, you might as well forget about getting served that day.
Even though this main office has eight clerk desks, only four were operational, with just two clerks handling money-related services. Based on my logs, each clerk spends about four minutes per customer. If we assume they work six solid hours a day, they can only serve around 90 people each. With two clerks, that’s 180 people in total, but that’s an optimistic estimate given the endless interruptions — network issues, money shortages, and so on.
I mentioned my queue number was 112. After four hours, Clerk #6 had handled 61 people, with 30 more to go. I wish I could’ve stayed longer to gather more data, but based on how tired the clerk seemed when I left, I doubt she made it to 90.
Solutions
The problem lies with both the system and us as users. Here are a few potential solutions, starting with ones I’ve seen work in other offices in the capital.
A Clerk Per Task
At a small post office near our university in Oued Smar, they manage the queue brilliantly. This office serves a train station, an industrial zone, a tech institute, and residential buildings. Despite the high traffic, I’ve never waited more than ten minutes.
They have six clerks, with one dedicated solely to balance checks, which significantly reduces wait times. Most people come to check their salary or pension, and this one clerk takes a huge load off the other clerks.
The three money-withdrawal clerks operate in a streamlined manner. One confirms the data and passes the request to the next clerk, who processes the withdrawal, while a third fetches the money from the back, so the clerks never have to leave their desks. This system speeds up the process considerably.
Separate Queues for Men and Women
In another post office in the capital, they divide the queue by gender. One clerk handles the women’s queue, while two clerks handle the men’s queue. Despite the limited resources, it works efficiently. However, this office doesn’t process large withdrawals, which may contribute to its speed. This approach is also used in utility companies when people pay their bills.
Stochastic Chains and Queues
I remember a course from college called FAS, Les Files d’attente Stochastiques. The professor introduced us to Markov chain models, which are used to solve queuing problems by predicting how long people will wait or how quickly service will be provided. I refreshed my memory on these models using ChatGPT, and here’s how they could apply to our situation.
Markov chains model the system as a series of states: how many clerks are free, how many are busy, and how many people are waiting. These models use probabilities to transition between states. For example, if a clerk finishes with a customer, they move to a “free” state, and the next person in line is served. This can help calculate average wait times and how many clerks are needed.
In our post office scenario, Markov chains could model the flow of people, from taking a ticket to completing their transaction. By using such models, the post office could, in theory, optimize clerk distribution, minimize wait times, and predict peak hours.
But that’s all theoretical. Applying these techniques in real life, especially with the bureaucratic inefficiencies in our country, is a different story. This is something they should have considered when building these offices, but we’re far from that becoming a reality.
My Solution is an App
While I waited in that long, slow queue, I kept thinking about how to fix this problem with an app. Here’s what the app should do:
- Selective Audience: It would target people with smartphones who are in a rush and don’t want to waste time in queues. This would exclude most elderly people, who often don’t use technology, and those who don’t mind waiting.
- Easy to Use, No Nonsense: The app should be simple, with no unnecessary loading screens or complications. When people are in a hurry, they don’t have time for that.
- Paid Agents: You could become an agent at any establishment and update the queue in real-time. Users would pay for this service.
- Cooperation and Solidarity: Users can also update the queue themselves, ensuring that others have up-to-date information. This helps filter out any bad agents.
- Fair Pricing: The cost of the service will depend on what agents charge and how much users are willing to pay. Initially, we’d set a fair price and adjust based on feedback.
Let’s end it here. This post was supposed to share my bitter experience at the post office, but instead, it turned into something much bigger.
Next up, I’ll be writing an article explaining the app idea in detail and whether I decide to move forward with its development.
