Product Innovation vs Process Innovation: Which One Pays

Both changes cost money, and both are earned back out of contribution. The two part on where that contribution has to come from. A change to how something is made is earned back against volume that already sells. A change to what is sold has to find volume nobody is buying yet. Same rupees out, two entirely different jobs to bring them home, and only one of those jobs needs a stranger to say yes.

Both cost money, so where do these two actually part company?

Two definitions and a table with a column each would be the obvious way to tell these two apart. Each kind of change is defined, told apart from the other, and split on what a change touches rather than on how clever it looks, under Innovation: The Types and Which Ones Threaten Incumbents, where both are taught at length with the case worked. Neither definition settles what follows.

A single question survives both definitions without either one asking it. Both kinds of change cost money. So what is each one earned back out of, and out of which volume? Recovery, not meaning, is what separates the two.

The same split shows up on an ordinary road. There are two ways to make a tea stall better. The first is a kettle that boils in half the time, and it gets earned back on the cups already going out every morning, to the same regulars, on the same road, with nobody outside the stall having to change their mind about anything. The second is a snack the stall has never sold. The same notes leave the till in both cases. Only one of the two needs a stranger to decide, for the first time, to buy something they have never bought.

With those two beside each other, the shape of everything below is already visible. The kettle's return sits inside a queue that exists. The snack's return sits in a queue nobody has joined. That one difference is worked first on published figures and then again on a case where the word process turns out to settle nothing whatever.

What is any change earned back out of, and at what rate?

No hedging is needed for what follows, so take it in one clean passage. A change costs a sum. The sum arrives in the standing baseSpending that arrives at the same size whether a year turns out busy or quiet. Rent, a salary bill and a building lease all belong to it, and which costs qualify is worked out elsewhere in these notes., the spending that turns up at the same size whether the year is busy or quiet. And it comes back one unit at a time, out of what each unit leaves behind after everything that moves with output has been paid.

Anjani Stationers Private Limited, invented for these notes and trading nowhere, makes hard-bound registers. One register carries Rs 46.20/- of contributionWhat a single sale leaves behind once every cost that rises and falls with output has been taken off it. Price less the costs that move, counted one unit at a time., a published figure. The year's published output is 2,50,000 registers. The two multiplied together give Rs 1,15,50,000/-, exactly the contribution that year published. The multiplication is not decoration. It checks that the two figures belong to the same year and the same works, and a count built on figures that failed that check would be a count built on nothing.

So here is the recovery division, in words before it is in figures: rupees added to what stands still, divided by contribution a unit, gives units of recovery. The honest unit of any recovery is a count of units, not a period. A period would need a rate at which those units arrive, and these notes publish no rate of arrival for anything at all. So no length of time is computed here. The arrival of the units is precisely the unknown thing, and putting a number of months on it would bury the one gap a reader most needs to see.

What does the one published rise in standing cost take to earn back?

Exactly one movement in what stands still is published for this works. The standing base went from Rs 49,60,000/- in the earlier year to Rs 74,00,000/- in the published year, a rise of Rs 24,40,000/-. Earn that back at Rs 46.20/- a register and it takes 52,813.85 registers of contribution in a year, or 21.13 per cent of the 2,50,000 registers the works turned out that year.

Now say what that figure is and, in the same breath, what it is not. The count is the size of the recovery job, measured in the only unit that carries any contribution at all, and it is not a statement about what the money went on. The account publishing that rise does put the three lines it sits in on the record, each with a figure against it: Rs 6,00,000/- more of employee benefits, Rs 11,40,000/- more in the standing half of other operating costs, mostly the year's second warehouse, and Rs 7,00,000/- more of depreciation and amortisation against assets bought. The three added together come back to Rs 24,40,000/- to the rupee, and the source carrying them marks that split an estimate rather than a disclosure. Which line a rupee landed in, though, is not which kind of change bought it, so which kind of change any of it was stays unstated, and from outside a set of filed accounts that distinction is invisible. Why it is invisible is settled elsewhere in these notes under Research and Development: Spending Today for Revenue Later.

Both the rise and the count are priced at the published year's contribution a register. The earlier year's contribution a register is not published, and its marginThe share of a sale price left standing after a stated group of costs has been taken off. Which costs are taken off is the whole of what a margin's name conveys, and the naming is settled elsewhere in these notes. differs. The pricing note travels with the figure or the figure does not travel. A reader meeting 52,813.85 registers without it cannot tell which year's rate priced a movement spanning two years, and a count whose rate is unknown is not a count at all. It is a shape.

Bring it home to a kitchen table. A household takes on a fixed monthly instalment and works out how many extra shifts would cover it. The shift count is honest arithmetic and takes a minute with a phone. Whether those shifts are on offer is a completely different question, nobody at the table can answer it, and it is the one that decides the year.

Which volume earns it back, and does that volume already exist?

Take that count, 52,813.85 registers, and ask the one question nobody has asked yet. Where do those registers come from?

For a change to how something is made, landing where the rate of the works is actually settled, the registers are already selling. The recovery is a slice of a block that exists. The slice can be set beside the order bookOrders taken and not yet delivered against. What sits there is a set of commitments other people made, which is exactly what separates it from a forecast., argued about with delivery notes on the table, and nobody outside the building has to do anything at all for the slice to be there. It was there before the proposal was typed.

For a change to what is sold, the registers are not there. The recovery is a block hanging off the end of the year with nothing underneath it, and every register inside it is a register somebody who is not yet a customer has to agree to buy. The same count means two completely different things depending on which side of the business's own volume it sits.

One sentence of consequence, and then it can stand on its own. One of these two recoveries can be argued about with the business's own records on the table. The other cannot be argued about at all without evidence nobody in these notes holds, and no amount of care with the arithmetic will supply it.

Two changes of the same kind in one works: why is one worth a great deal and the other nothing?

Throughput, covered separately in these notes, takes this same works and proves why a run of stages turns out what its slowest stage turns out, working three moves on paper to get there. Two of those finished moves are taken up here, with a question that treatment never had to ask: what would each one have had to be earned back against?

In that treatment, cutting runs at 150 registers an hour, printing at 125 and binding at 100, so the works makes 100 an hour and a rated abilityThe ceiling a works sets for itself: its own rate multiplied by the hours it counts as open. Nobody has promised to sell that much, and nobody has promised to make it either. of 4,00,000 registers a year on 4,000 line-hoursOne hour of availability on one production line. Count the lines, count the hours each was open, multiply the two, and that is the year's supply of them., against 2,50,000 registers actually turned out. Two changes are then worked there, and both are worked as supposings rather than as things that took place.

Suppose a faster press took printing from 125 registers an hour to 150. Binding would still take 100 an hour, so the works would still make 100 an hour and would still turn out 2,50,000 registers a year. The return is zero rather than merely small. The cost, meanwhile, would still arrive in the Rs 74,00,000/- that stands still, so the denominator of that recovery would be nothing at all while the number on top of it was entirely unchanged.

Take binding from 100 registers an hour to 125 instead, and rated ability moves from 4,00,000 registers to 5,00,000 on the same 4,000 line-hours, a gain of 1,00,000 registers of yearly ability. The constraintThe one stage in a run of stages that settles how fast the whole run goes. Every other stage waits on it for some part of its time, and which stage it is can move. does not disappear at that point. It relocates. Printing also runs at 125, the two then tie, and passing 125 an hour would need both raised together.

Two changes of the same kind, in the same works, in the same demonstration, are worth 1,00,000 registers of yearly ability and exactly nothing, so the word did no work in either case. The ruling deserves stating without softening. A change to how something is made is earned back against volume that already exists only where it reaches that volume. Where it does not reach it, the denominator is nothing while the cost stands exactly as it was.

Figures this vivid crowd out the plain fact, so state it plainly. Neither move happened. Both are worked elsewhere as demonstrations. The business's rated ability is 4,00,000 registers, its works runs at 100 an hour, and no source says it bought a press or lifted a stage.

Which of the two returns can be worked out before anybody buys anything?

The plain answer flatters the wrong side, so answer it plainly first and then turn it over. The return on a change to how something is made, landing where the rate is settled, closes on figures that never leave the building: three stage rates, a count of hours, and what one unit leaves behind. Every one of those is a fact the works already holds about itself, written down somewhere in its own records.

The return on a change to what is sold needs one figure that sits in no ledger anywhere. The figure is a count of the people who are not buying today and would start. The count is not hard to find. The count has never been taken. No survey, no filing and no trade paper carries it for this works or for any other, and an appraisalThe piece of work done before a decision, putting a figure on what something would be worth if it went ahead. It is a claim about the future written in units, so the units decide what evidence it needs. that needs it is an appraisal with a hole in the middle. One appraisal closes on figures that never leave the building and the other does not close at all without evidence about people who are not customers yet.

Stopping here would leave the first kind looking simply the better one, and it is not. Being computable is not the same as being valuable. A business that only makes the changes it can appraise will make the same kind of change every year, and it will do so through a review that behaves impeccably at every single step. The easier appraisal is the one that gets done, and what gets appraised is what gets proposed. The pull is real, and naming it is not the same as recommending anything.

When does the earning back actually start?

Two starting guns, and they are not on the same clock. For a change landing on volume that already sells, the first register counting towards the recovery is a register that was going to be sold anyway. The recovery starts with the next order out of the existing book, more or less at once, and without a single person outside the building being asked for anything.

For a change to what is sold, the first register that counts is the first register somebody agrees to buy. Nothing in these notes says when that is, or whether it happens at all. One recovery starts with an order that already exists and the other starts with a decision somebody else has not made yet.

The obvious next move is then refused, as a matter of evidence rather than of taste. No length of time is worked out here in either case. A period needs a rate at which units arrive, and these notes publish no path, no rate and no future period for anything whatever. A figure supplied at that point would be a made-up number, and it would be the one carrying every ounce of the weight.

Where does this go wrong when nobody has made a mistake?

What should be asked of any proposed change before its number is looked at?

Run the two supposings through that card and watch where they come apart. Question one prices both, and neither carries a price anywhere in these notes, so both stop at the same place. Question two puts both on volume that already sells, so both pass. Question three separates them completely: one lands on the stage that settles the rate of the works and one does not. Questions two and three between them would have told the two apart before anybody had costed either machine, and that is the entire reason for asking them in that order.

Who actually uses this, and how. A lender looking at a works that has just added to what stands still will be told at most which lines rose, never which kind of change they bought, and does not need the second. The lender can divide the rise by contribution a unit and read off the count of units that has to keep arriving before the year merely stands where it stood. An analyst reading an announced change does the same division and then asks the only follow-up that matters. Are those units in the order book already, or is the announcement quietly a claim about people who are not customers? A household weighing up a second-hand delivery scooter against the deliveries it already makes every evening is doing precisely the same arithmetic on a smaller sheet of paper.

What is handed on to other subjects?