Chapter 1 The Reorder Book

Part No. 1147 · Bush, bronze, 32 mm · Reorder level 40 · Checked: 14 March 1988 · Initials: R.K.

A note on the people. The company in this chapter was real and so was the work. The men who ran it are composites, and their names are invented, because the point of the story is the chain they formed and not the men who happened to stand in it. Naval is real, and so is Orange.

Once a month, Mr Malhotra shut his door.

He was the chairman of a company that built printing machines, and for one week in every four he was not available to anyone. Not to customers, not to the bank, not to his wife, who rang on the third day of every projection week to ask whether he was still alive and was told by the peon, with great seriousness, that sir was inside with the ledger. The sales ledger went in with him, and the file of pending orders, and a pad of lined paper, and a flask of tea that was refilled through a door opened six inches and no more. He came out five or six days later, unshaven, holding a single sheet: how many of each machine the company would sell over the next six months.

Nobody else was allowed to make that projection. He had built the company from one machine and a rented shed, he knew every customer by name and most of them by temperament, and he believed, probably rightly, that nobody else could guess the future of the order book as well as he could. So every month the company’s most capable mind spent a week doing arithmetic, and the company’s most expensive door stayed shut.

The sheet went to Dhawan, the plant manager. Dhawan was a large, unhurried man who had been with the company since the first machine, and he carried the whole plant in his head: which part went into which subassembly, which subassembly into which machine, how long a casting took to come from the foundry, how long a bush took to turn, which of the thousands of parts on the shelves had to exist before anything above them could be built. He never consulted a drawing. If you asked him how many gears went into the folding unit he would close his eyes for a moment, as though reading something off the inside of his eyelids, and tell you. From the chairman’s single sheet he worked downward, machine by machine, assembly by assembly, part by part, and wrote it all out by hand. It took him two weeks and filled a stack of foolscap three hundred sheets thick. He did it from memory, because there was nowhere else the information lived, and if Dhawan had been knocked down by a scooter on the Mathura Road the company would have discovered that it did not know how to build its own machines.

The stack went to his clerk, a thin, careful man named Ramesh, who copied every line into the part registers. There was one register page for every part, thousands of pages in a row of bound ledgers along the stores wall, and each page carried the part’s history: what had been issued, what had been received, and now, in Ramesh’s small, upright hand, what the plant manager said the next six months would need. Ramesh worked for most of a week. He wrote with a fountain pen, and he kept a second, cheaper pen for corrections, and the registers were the one place in the building where you could read, in two inks, the history of everyone’s mistakes.

The registers went to Verma, the production planning manager, whose job was to find the mistakes, and there were always mistakes, perhaps fifty in a month. Not because anyone was careless. Because some parts that the plan said must be made were already on the shop floor, half-finished, as work in progress; because some parts had been replaced by other parts since the last plan and the old number was still in Dhawan’s head; because a line copied from one sheet to another by a tired man at four in the afternoon is sometimes copied wrong. Verma took a grim pleasure in the fifty. He kept a count, month by month, on the inside cover of his diary, and once told me, as a point of professional pride, that he had never found fewer than thirty.

And last, the store clerk, Rao, walked the shelves with the registers open and compared what the plan said with what he could see and count, and with what he expected to use in the weeks it took a reorder to arrive, and raised the orders. Rao had been doing this for eleven years and did not entirely trust anything written down, including his own register, so he counted. Every month, every shelf.

Five men. Five desks. A month. And at the end of the month, a set of reorder levels that had been true at the beginning of it.

What a month costs

I want to be exact about why this was expensive, because the expense was not where a visitor would have looked for it.

The visible cost was the time: a week of the chairman, two weeks of the plant manager, most of a week each from the clerk, the planner and the storekeeper. Add it up and the reorder book consumed something like six man-weeks every month, most of it from the four or five most experienced people in the building. That was bad enough. But it was not the real cost.

The real cost was fear. Every man in the chain knew that the number he wrote down would not be looked at again for a month, and that if it was too low, somewhere in that month a machine would stand on the floor unable to ship because one bush or one gear was not on the shelf. A finished printing machine is a large, expensive object, and a customer who has been promised it does not want to hear about a bronze bush. So every man, at every desk, rounded up. The chairman projected a little high, to be safe. Dhawan added a margin on top of the chairman’s margin. Rao, walking the shelves, added his own. By the time the reorder levels reached the supplier, they carried four layers of caution, each reasonable on its own, and the shelves filled with parts that would not be needed for months. The money tied up in that stock was the largest sum of money in the company, more than a year’s wage bill, and nobody had decided to spend it. It had simply accumulated, one careful margin at a time, as the price of an answer that took a month to compute.

And even then the answer was wrong, because it was stale. The projection was made on the first of the month from orders that had arrived by the last day of the month before. By the time Rao raised the purchase orders, four weeks had passed, two new orders had come in, one had been cancelled, and a customer in Ludhiana had rung to ask whether his machine could be brought forward. None of this was in the registers. It could not be. The registers were a photograph of the company taken a month ago and handed to people who had to act today.

The machine that could not go

I remember the first stockout I saw, because it was my second week in the building and I did not yet understand what I was looking at.

A machine stood on the assembly floor, complete in every respect but one. It had been promised for the end of the month; the customer had already paid part of the price, and had, I learned later, already told his own customers. The missing part was a small thing, a hardened steel component that went somewhere inside the drive. The register said there should have been six on the shelf. There were none. Three had gone into machines built the previous month, which the register knew about. Three had been issued to the shop as work in progress for machines not yet built, which the register knew about too, on a different page, in a different hand, and nobody had put the two pages together.

The part took three weeks to source. The machine stood there for three weeks, with men walking round it, and after the first few days they stopped looking at it, the way people stop looking at a relative who has overstayed. The customer rang every second day, and then every day, and then stopped ringing, which was worse. On the Tuesday of the third week Mr Malhotra came down to the floor, which he did not often do, and stood in front of the machine with his hands behind his back for what felt like a very long time. He did not shout at anyone, because there was nobody to shout at. Every man in the chain had done exactly what the system asked of him, and the system had produced a machine that could not go. He asked Rao how many of the part were on order. Rao told him. He nodded, and went back upstairs.

That was the moment I understood what we had been let in for. Not a stores problem, and not a people problem. A problem of a result that mattered, which no one person could hold in his head, and which had therefore been cut into five pieces and handed down a line, and which leaked a little at every handoff until, at the end, something was simply missing.

Two outsiders

I should say what I was doing there, because I was not an employee and never became one.

I had a company, or half of one. It was called Orange Technology, and I had started it after engineering college and eighteen months at Eicher, with Naval Bansal, who had sat two rows behind me for four years and was the only person I knew who was more certain than I was that we should not work for anyone. Orange wrote custom software for anyone who would pay for it. It consisted of the two of us, a second-hand PC, a dot-matrix printer that could be heard from the street, and a scooter on which we visited every factory and office within a day’s ride of the city, one after another, to tell people who had not asked that their paper could be made to disappear.

That is how we found Mr Malhotra’s company, and how we had found the ones before it. There was no introduction and no uncle. We had a list of factories from the industrial directory and we went down it, gate by gate, and at most gates we were told that sir was busy, and at some we were given tea and shown out, and at a few, after the third or fourth visit, we were given a problem. By the time we came to the printing-machine works we had several of these behind us, each one sold the hard way, and we had learned the first rule of the trade, which is that nobody buys software. They buy the end of a particular pain, and you have to find the pain before you can sell them anything.

At Mr Malhotra’s, the pain was on the assembly floor, complete in every respect but one.

The second rule we had learned was not to promise anything until we understood it. So we did not arrive with a program, though we were sure enough of ourselves to have written one on the spot if he had asked. We arrived with a proposal for a systems study: a few weeks in the building, at a fee that was small enough to be agreed on the spot, at the end of which we would tell them what could be done and what it would cost, and they could say no. Mr Malhotra said yes to that in about four minutes, which we took to be a sign of how much pain there was, and gave us a chair in the corner of Verma’s office, a cup of tea every two hours, and the run of the building.

I knew what a reorder level was in the way that one knows what a carburettor is from a diagram. Inventory control had been one chapter in one course at college, examined once and forgotten. Neither of us had ever seen a register, never walked a stores shelf, never met a man like Dhawan who carried a plant in his head.

So we spent the study doing nothing but understanding. I sat with each of the five men in turn and asked them to show me what they did, and then asked them to show me again, slowly, and I drew it. I drew the chairman’s projection sheet and the columns on it. I drew the way Dhawan exploded a machine into its assemblies and the assemblies into their parts, and the lead time he attached to each, and where that lead time came from. I drew a register page and every entry on it. I drew what Verma looked for when he hunted for mistakes, and what Rao did on the shelves. Naval, who was better at arithmetic than I was and worse at listening, took my drawings in the evenings and turned them into tables and counts: how many parts, how many assemblies, how many entries a month, how many hours. By the end of the month the whole chain was on paper, in our hands, as a set of boxes with arrows between them, and we could see something that nobody standing in the chain could see, because each of them saw only his own desk.

We could see that it was arithmetic. All of it. The chairman’s week was judgement, and judgement we could not replace; but once his single sheet existed, everything that happened to it afterwards, on four desks over three weeks, was multiplication, addition, subtraction and lookup. Dhawan’s three hundred sheets were a bill of materials multiplied by a sales forecast. Ramesh’s week was copying. Verma’s fifty corrections were the places where two copies of the same fact disagreed. Rao’s walk was a comparison of two columns. There was not one decision in the three weeks after the chairman’s sheet that a machine could not make, faster, every day, from data that was as new as that morning.

We were certain of this from the first week, long before we had earned the right to be. We did not say it loudly, because Dhawan had been doing the job since before we were born and we had the sense to let the study finish before we told him his fortnight was arithmetic. But we did not doubt it for a moment, and we never did, through everything that followed. We were two very confident young men. We had been short of money since the day we started, and we would be short of money for years, and it had never once occurred to either of us that this was a reason not to take on whatever was put in front of us. The only question was whether they would let us build it.

The presentation

At the end of the study we asked for an hour with Mr Malhotra, and were given it, in his office, with Dhawan and Verma called in and the door shut, which we understood to mean that it was serious.

We had our drawings. We laid them out on his desk, all five desks and all the arrows, and we told him what I have just told you: that after his sheet, it was arithmetic, and arithmetic could be done by a machine in an hour instead of by four men in three weeks, and done again the next morning, and the morning after that, from numbers that were never more than a day old. We told him the stockouts would stop, because the register and the shop floor would be the same page. We told him the stock would fall, because four men would no longer each be adding a margin against a month of blindness. Naval had a sheet of numbers, and he went through them one by one, looking Mr Malhotra in the eye after each, as though daring him to find one wrong.

Mr Malhotra listened without saying anything, which was his way, and then he asked three questions. The first was how long. We said six months, and he looked at us in a way that told us he had heard that number from other people. The second was what it would cost. We told him, and he did not react at all, which was worse than if he had. The third was the one we had known was coming. He turned to Dhawan, who was sitting with his hands on his knees, and asked: and what happens to you?

Dhawan said, after a moment, that he supposed he got his fortnight back. Mr Malhotra did not say anything to that either. But he picked up one of my drawings, the one of his own projection sheet, and looked at it for a while, and I saw something in his face that I did not understand until years later, when I had my own company and my own shut door: the look of a man who has been told that the thing he does with his best week every month is not, after all, the thing that only he can do.

Then he told us his terms, and they were the terms of a man who had bought a great many machines. There would be a small advance, small enough that I remember thinking it would not cover the petrol. After that we would be paid in stages, and only when each stage was done to his satisfaction and not before: first a full specification of the system, every screen and every report, which Verma would check; then a pilot, on one family of parts, run beside the registers until it agreed with them; then the implementation, piece by piece, each piece signed off before the next was paid for. If a stage was late, we would wait. If it was wrong, we would do it again, at our cost. He said all this without any unkindness, as a man states the price of bread, and then he asked whether we agreed.

We agreed before he had finished asking. We were young, we had never once doubted that we could do anything that was put in front of us, and a man offering to pay us only for what worked did not frighten us; it flattered us. We walked out into the corridor and Naval said: he is going to make us work for every rupee. He was grinning. He was right. It was the best thing that ever happened to us.

Six months

There was no software to buy. In 1989 in India there was no package we could find that did this, nothing to install, nobody to call. If the reorder book was going to become a program, somebody was going to have to write it, and the somebody was the two of us, in Verma’s corner, on our own machine, which we carried in from the scooter every morning in two trips.

We wrote the specification first, because we had to, because it was the first stage and nobody was paying for anything until Verma had read it. I had never written one before. It took six weeks and ran to a thickness that embarrassed us, and Verma read every page of it with the same pleasure he brought to the monthly fifty, and sent it back twice. By the third time it was, I think, the most complete description of that company’s stores that had ever existed, and it was also the first money Orange earned from Mr Malhotra. I remember the cheque. It was not large, and it was already spent, because we were always short of money and had been living for two months on the advance and on what the earlier clients still owed us. We looked at it for longer than it deserved, and then Naval took it to the bank on the scooter before the ink was dry.

Then the program. I wrote it first in FoxBASE, with pieces in C where FoxBASE was too slow, and moved it to FoxPro when FoxPro appeared, which happened in the middle of the work and felt like a gift from somewhere. I will not describe the program, because the program is not the point and because it was, by any later standard, crude. What matters is what it held. It held the bill of materials, every machine exploded into its assemblies and parts, which had lived only in Dhawan’s head. It held a lead time against every part, which had lived there too. It held the register, every issue and receipt, keyed in as it happened rather than copied once a month. And it took the chairman’s sheet, which still came out of his room once a month in his own hand, and did in a few minutes of machine time, and an hour from keying to checked report, what four desks had done in three weeks.

Dhawan gave us his afternoons, which was the real gift. He would come to Verma’s corner after lunch, lower himself onto a stool that was not built for him, and take the plant out of his head one assembly at a time while I put it into tables. It went slowly, because he would stop to tell us why a part was the way it was, and who had designed it, and what had gone wrong the year they had tried a cheaper one, and I learned more about printing machines in those afternoons than I have learned about anything since. He was not threatened by what we were doing. He said more than once, with something like relief, that he had been waiting twenty years for somebody to write it down, and that he had sometimes lain awake wondering what would happen to the company if he forgot.

Verma was another matter. Verma had understood, before anyone else, that a program which found mistakes automatically would put an end to the monthly fifty, and with it to the only part of his job that anyone noticed. For the first two months he was correct, distant and exacting, and since he was also the man who signed off each stage before we were paid, he was in a position to be exacting at some length. We did not mind. We had met Vermas at other gates, and we had learned that the man who is hardest to convince is the one worth convincing, because once he is on your side he brings the building with him. What changed him was not anything we said. It was the pilot. We ran it on one family of parts, the drive components, beside the registers, as the terms required, and the first time it was run against a month-old register it found sixty-one mistakes to his fifty. He spent an afternoon checking them and found that it was right about fifty-eight. He came to our corner at six in the evening with the printout, put it on the desk, and said, with the air of a man conceding a chess game: it is better than me. Then he said: now make it find the other three. Then he signed the pilot off, and we were paid for it, and after that he was the program’s fiercest tester, and I think its proudest user, and his diary got a new count.

Ramesh learned to type in a fortnight and never looked back, though he kept the fountain pen in his pocket for a year, the way a man keeps a watch that has stopped. Rao did not trust the screen and said so, and went on counting the shelves every month, and the program was better for it, because every time his count and the screen disagreed we found something: a receipt note never entered, a part issued on a chit and never recorded. He was, without knowing it, our audit.

The implementation went in piece by piece, as the terms required, and each piece was paid for only after the men who used it said it worked. By 1989, after about six months, the reorder book was a program. Then we did the thing that I now believe was the most important decision in the whole project, though at the time it felt like cowardice, and though in truth it was Mr Malhotra’s condition and not our idea. We ran both. For six more months, the registers were kept exactly as before, and the chairman’s sheet went to Dhawan exactly as before, and the three hundred pages were written and copied and corrected and walked, and alongside all of it the program produced the same answer from the same sheet in an hour. Every month, Verma compared the two. Every month, the program found mistakes in the registers, and occasionally the registers found a mistake in the program, and I fixed it that day, sometimes that night, in Verma’s corner with the lights off in the rest of the building and the watchman looking in every hour to see whether the boys had finally gone home. After six months of that, nobody in the building trusted the registers more than the screen, and the last stage was signed, and the last cheque came.

We were paid, in the end, every rupee we had quoted, and not one more, and not one of them early. We were short of money the whole way through, as we were short of money through every year of Orange, and I have come to think that this was not a hardship but a kind of education: a company that is always a cheque away from trouble learns very quickly what a customer will actually pay for. I have had easier clients since. I have never had a better one. A man who pays you only for what works teaches you, faster than any college, what working means.

What came back

Nobody lost his job. I say that plainly because it is the first question anyone asks and because it was true. The word at the time was computerisation, and what computerisation did in that company was give time back. Mr Malhotra kept his week of projection, because that was judgement and the program could not do it, but the three weeks that used to follow his sheet simply stopped happening. Dhawan got his fortnight back every month and spent it on the shop floor, where he had always wanted to be, and where, within a year, he had redesigned the folding unit that had been giving trouble since before I was born. Verma, with no copies to reconcile, turned to the scheduling of the floor itself, which had never had anyone’s full attention, and found that it was full of the same month-old answers. Ramesh became the man who kept the data clean, which turned out to be a more important job than the one he had had before, and the first person in the company to be sent on a training course. Rao still walked the shelves, but now with a printout that was true as of that morning, and he stopped adding his margin, because he no longer needed to. He never said the screen was right. He said it was less wrong than it used to be, which from Rao was a speech.

The stock fell. Not at once, and not because anyone ordered it to, but because four layers of caution had been taken out of the chain and replaced by one answer that was never stale. Within a year the money tied up on the shelves had fallen substantially, and the stockouts, the machines standing complete but for one bush, became rare. The company did markedly better, and the only thing that had changed was that five people handing a month-old answer down a line had been replaced by one calculation that was never out of date.

There was no morning when it printed itself and everyone cheered. I had half expected one; I had, I am embarrassed to say, imagined it. Instead there were bugs, every day, small ones, found by the men who used it and fixed by me before the next day, and the program grew the way such things grow: one more register, one more kind of receipt note, one more report that Verma wanted, one more exception that Rao had always handled in his head and now had to be written down. It was never finished. I understand now that it could not have been, and that the feeling of its being unfinished was not a failure but the ordinary condition of a system that is actually in use.

Mr Malhotra’s door still shut once a month. But his wife told me, at a wedding years later, that he came out on the fourth day now, not the sixth, and that he had started to whistle.

The registers

I do not know the exact day the registers stopped. There was no decision. Ramesh kept them through the six months of running both, and then, I think, for a while afterwards out of habit, and then less often, and then there was a month when nobody opened them at all, and nobody noticed that nobody had.

I found them a year or so later, on the shelf along the stores wall where they had always been, with a film of dust on the top ledger. Orange had more clients by then, most of them found the same way, gate by gate, and I had come back on the scooter to fix something small. I took one down and opened it at random. Part number, description, reorder level, a date in 1988, initials. Four men’s handwriting on the page, two of them in two inks, each correcting the one before. I stood there for a minute with it open, and I thought about Dhawan on his stool, and Verma with his printout at six in the evening, and Naval daring a chairman to find one number wrong, and the machine that could not go. Then I put it back, and I did not open another one for thirty-five years.