Steve Jobs

The pattern they actually represent

Apple lost about a billion dollars in fiscal 1997. Jobs returned that September as interim chief executive, seven months after Apple bought NeXT for roughly $427 million. In fiscal 1998 the company reported a profit of $309 million.

The turnaround is usually told as taste. The copyable part is a shape, and you can read it off the shipping schedule rather than off any anecdote. Two questions crossed — consumer or professional, desktop or portable — make four boxes. The Power Macintosh G3 arrived in November 1997, the iMac in August 1998, and the iBook, announced in July 1999, filled the last empty box beside the PowerBook G3.

A grid does two jobs, and the second one gets dropped in the retelling. It gives every cancellation a stated reason: this product has no box. And it puts every survivor in debt: each box must have something shipping into it, or the cut was not focus, it was shrinkage.

Watch how the cuts read with that frame in place. In September 1997 Apple bought Power Computing's customer list and its Mac OS licence back for $100 million in stock and closed the clone programme. In February 1998 Apple ended the Newton — a programme begun in 1987 whose first MessagePad shipped in 1993 — and the press release carried the reason in one sentence, on the day of the cut.

That is the whole move, and it is unsentimental rather than brave. Kill a decade-old flagship, publish the criterion in the same breath, and let the criterion take the argument for you. Asked from the floor at WWDC in May 1997 why Apple had killed OpenDoc, he defended the cut rather than distancing himself from it, and stated the same test: start with the customer experience and work backwards to the technology.

The blockage it speaks to

Read this one if your product has grown more surface than you can hold, and every piece of it has at least one real customer attached.

You know which parts are dead weight. You can name them. What you cannot do is remove them, because removal has a face — the client who uses that one report, the integration that took a quarter to build. Every item defends itself convincingly, and no single item is ever worth the fight. So the scope stays, and the thing you want to ship moves out another month.

Restraint is not the study you need. Restraint is for scope that has not been built yet, and yours is built, staffed and in production. Jobs is the right example precisely because his cut was retrospective and expensive: a programme with ten years and a shipping product inside it, ended in one sentence. He was not protecting a clean line. He was taking a sprawling one back.

And the reason he could is that the grid made the decision, not the moment. Once the line is four boxes, the Newton stops being a beloved product ended by a man with strong opinions. It is an item with no box. That is what makes the move repeatable by someone who is not him.

Three moves you can steal

Draw the boxes before you look at the list. Pick the two questions that actually divide your market — who it is for, and which situation they are in — and cross them. Do it with no product names anywhere near the page. If you start from the list you will defend the list, one sympathetic item at a time.

Give every surviving box a debt. A grid is not a filter, it is a set of promises. Each box needs exactly one thing shipping into it, and a box you cannot staff is a box you should not have drawn. This is what stops subtraction becoming retreat: you are not cutting to have less, you are cutting so four things get finished.

Publish the cut and its reason in the same sentence. Not a quiet deprecation notice six months later. One line to the customers who used it, naming what stops and where the effort goes instead. Saying the criterion out loud is what makes the next cut cheaper: it has now taken an argument and survived one.

Where the pattern breaks

Subtraction is only as good as the arbiter, and a single arbiter has no counterweight. The Macintosh Apple introduced on 24 January 1984 carried 128 kilobytes of memory and no expansion slots — removals made on one judgement, on the same instinct that later produced four boxes. By June 1985 Apple had reorganised, merged the Macintosh and Apple II divisions, taken Jobs out of operating responsibility, and announced 1,200 layoffs, about a fifth of its workforce, alongside the first quarterly loss in the company's history. A cut that removes what customers need is not focus. Which is why the grid outranks the taste: a grid can be argued with by someone who is not you.

What to do this week

Draw two axes over your own market on a blank page. Label them and leave every product name out. Four boxes, and no more than four — a five-box grid is the same avoidance wearing geometry.

Now put each thing you currently sell, support or maintain into exactly one box. Anything that lands in no box goes on a list. Anything that lands in a box already occupied by something better goes on the same list.

Take the cheapest item on that list — the one with the smallest number of customers attached — and write the sentence that ends it: what stops, when it stops, and where the effort goes instead. Send it to the people it affects.

Then show someone the grid before you send anything. Not the list, the grid. If the axes survive somebody else's questions, every cut underneath them is already decided.