I am setting this down in the second week of Phalgun, in a flat in old Dhanmondi that has belonged to my wife's family since the year after the war. A young friend of the family sat for the 45th BCS preliminary last winter and failed, and his mother phoned me crying as if he had been arrested. The boy had spent fourteen months on Bangla literature, mathematics, Bangladesh affairs, and the front page of Prothom Alo. His coaching teacher in Farmgate had told him on the first day, very calmly, that nineteen out of twenty people who walk into the prelim hall walk back out with nothing. The teacher was telling the truth, and the boy heard it, and the boy walked into the hall anyway and came back out at noon with the rest who had nothing to show for it. His mother believes the country has wronged him. I think the country has run him through the same machine it has been running my friends through since 1972, when the first batch was constituted under a different name. The machine itself has not changed its shape.
The numbers have always been public, and everyone in the room has always known them. About twenty lakh people fill out the form in any given cycle. The Bangladesh Public Service Commission, the BPSC, runs the whole thing as a tournament with multiple stages that takes the better part of two years to complete. My closest friend from the Liberation War, Aminul Haque, sat for the third BCS in 1981 after coming home from a posting in Saudi Arabia. He used to say that the form is a gentleman's contract between the state and twenty lakh of its own citizens, in which the state agrees to give them all an equal chance to be disappointed.
Aminul died of a stroke in 2018.
He never got the cadre, and most of his 1981 batchmates never got it either, and the few who did spent the rest of their working lives moving between Khagrachari and the Secretariat.
A casual observer, asked where most BCS hopefuls die, will guess viva, or written, the romantic stages where one can imagine a candidate sweating in a hot room with a panel of secretaries staring at him. The truth is somewhere else entirely. The biggest absolute drop is between filling the form and showing up to the prelim. About fifteen lakh people, in round numbers, gone before the contest has even started. The next biggest drop is the prelim itself, an MCQ paper lasting two hours where one needs around a hundred and five marks out of two hundred to clear the cut. Almost everyone who shows up still fails. It is, structurally, also the easiest stage to fix, because the syllabus has been fixed for thirty years, the question patterns are public, and previous papers are sold in every footpath stall on Nilkhet for forty taka apiece.
If a person who is building a software product reads what I have just described, that person will recognise the shape. The same drawing has been on every product manager's whiteboard since the late nineties. The signup curve looks like this, the checkout curve looks like this, and so does the retention chart for whatever app for hailing rides a junior product manager is staring at the night before a quarterly review. The absolute numbers are smaller, but the sketch on the back of the napkin is the same one BPSC has been drawing in Agargaon since the late seventies. The lesson from one applies to the other if you have the patience to sit with it for an hour.
Anyone who has actually prepared for the BCS in the last forty years will tell you the same thing. It takes two years of real elapsed time from filling in the form to final posting. Two years in which your university friends start jobs at BRAC or Grameenphone, get married, and move to Toronto or Sydney, while you sit in a flat with a single room in Mohammadpur reading the Liberation War chapters of your guide book for the eleventh time. The falling away is not because people are failing the test, although they are. The falling away is because people are quietly stopping. They get a job at a cousin's import business. They get pregnant. The mother gets sick and somebody has to manage the prescriptions. The father retires and the household account quietly shifts from a salary to a pension, and certain things stop being affordable. None of these is a failing of the candidate, but each is a failing of the funnel, which was set up to test whether a person could sit through fifty years of bureaucratic furniture. It was not set up to ask whether the person could keep paying the bills while they did so. Avinash Kaushik wrote about this in the marketing context under the name See-Think-Do-Care, and his argument was that the people who run the funnel obsess over the bottom of it, where the money is, and let the top rot from neglect. The coaching teachers in Farmgate worked this out without ever reading a Western analytics blog. They sell batches specific to the prelim in sprints of two months, priced at a level that a working bus driver in Mirpur can afford for one of his children, and the teachers are right.
I think about this every time I sit at a wedding in Gulshan and a young woman from a multinational tells me, with the sincerity of someone not yet disappointed, that her team will soon ship a redesign worth four percent on conversion at step three. Perhaps the redesign will deliver. I will believe the chart when I see it. But how does the redesign hold a person who has just received a WhatsApp from his mother asking what time he is coming home, and at the same instant a message from a colleague about a meeting he had forgotten he was hosting? Show me the variant of the redesign that survives both interruptions and I will agree to cancel the next two sprints to ship it. The young woman did not have an answer at the wedding, although she was kind enough to stop talking and let me eat my biriyani.
A few years ago I helped a couple of younger friends start a small side project called PorashonaOnline, or PO. They came to me because they remembered that I had spent a decade in the seventies and eighties working in the Education Ministry's planning cell, and they wanted an old man's view of whether the idea was foolish. The pitch was modest. The intended user was a young man on a Pathao bus, stuck in traffic on Mirpur Road, somewhere between his office in Karwan Bazar and his coaching centre in Farmgate. He had 45 minutes of stationary bus time on his hands. He did not want a video lecture. He wanted twenty MCQs and a small dopamine reward for getting one of them right.
It worked, in the quiet way that most software does not. The lesson was not about the technology, and it was not about the syllabus either. What the funnel actually cares about is not your product at all, but whether you have respected the real, daily constraints of the person inside it. The constraints we had to name out loud included Mirpur Road traffic, the duty to sit with one's mother in the evening, and the cha break at five with whatever instant coffee Pran is selling that month. There were also the thirty days of Ramadan during which nobody studies after Iftar because nobody is awake. If you cannot name those constraints for your user with that much specificity, you are not building a funnel. You are building a wishlist with a chart laid on top of it, and the chart will look very fine until the day someone asks why the numbers have not moved.
Next time a young product manager in your office stands up at a planning meeting and begins to speak about the funnel, ask him which stage in his chart maps to the prelim, and where the equivalent of fifteen lakh missing people is leaking out of his product before the contest has even started. Then ask what his user is actually doing on the bus when the retention email arrives at quarter past five on a Wednesday afternoon in late Boishakh. If none of that comes back without a fresh dashboard tab being opened first, the chart behind him is decorative and the cohort it claims to describe half emptied weeks ago.
I say this with some sympathy, because the chart is not the problem and the young man is usually not lazy. The problem is that a funnel diagram is built to show you the stages that exist, and the expensive leaks are almost always happening at a stage nobody drew. Nothing in the BPSC's own reporting has a box for the fifteen lakh who paid the fee, took the receipt, and then quietly declined to sit the exam. That absence is the real tell. They were never a stage in anyone's chart, only the space between two stages, exactly where they went. The only real way anyone ever finds them is by standing quietly in the room at Agargaon on the morning of the prelim, counting the empty chairs against the receipts the clerks filed.