Reed Hastings

The pattern they actually represent

Netflix wrote the retirement of its profitable core business into a legal filing, and did it while calling that business healthy.

The annual report for 2007 states both halves in one paragraph. The disc, it says, "will continue to be the main vehicle for watching content in the home for the foreseeable future" — and, in the same breath, that "by growing a large DVD subscription business, we will be well positioned to transition our subscribers and our business to Internet-based delivery of content."

Read that twice. A public company is telling its shareholders that the purpose of growing the disc business is to pay for what will replace it. Instant watching had shipped that January, on PCs only. The crossover was four years away.

That is the pattern, and it is rarer than it sounds. Most founders can retire a losing line, because a losing line argues its own case. This record is a run of decisions to retire winning ones, early, on a schedule nobody else set.

It had happened once already. In March 1999 Netflix stopped selling new discs, and that September it replaced per-disc rental revenue — the revenue it actually had — with a flat monthly fee and no late fees. It happened again at the end: in the third quarter of 2011 there were still 13.9 million disc subscriptions in the United States next to 21.4 million streaming subscriptions, and the disc business was already walking toward the close it reached in 2023.

None of those moves was a rescue. Every time, the business being replaced was the one paying the bills.

The blockage it speaks to

The catalogue defines this obstacle as pouring money into failing bets because of what is already spent. That is the easy version. You can point at the loser.

The hard version is the one on this page. The asset is still paying.

In the quarter Netflix reported its worst results yet, the disc business it was retiring had 13.9 million subscriptions and was making money. Nothing was failing. Nobody was going to force the decision. The mechanism is identical — past investment does the present allocating — but the disguise changes. Instead of stubbornness it wears prudence, which has no rebuttal in it and no deadline.

You are the reader for this if you have one line that works and have privately known for a year that it is not the future. You are not neglecting the replacement; you are resourcing it from the leftovers, and the old line keeps your strongest people because it pays. Every allocation is defensible alone, and together they set the date the replacement is ready — after the market turns.

Two structures hold it there. One is the blended number: while old and new sit in one revenue line, a declining line rides on a growing one and nobody has to decide anything. The other is waiting for permission — a bad quarter, a lost client, a rival's launch — because a decline decides for you at no cost to your story. Permission arrives exactly when the transition is no longer affordable.

Three moves you can steal

Write the retirement date for your best line in a document somebody else reads. A named quarter, in the board update or the investor note, beside an honest statement that the line is healthy today. That is what the 2007 filing does. It separates direction from timing, so the old line keeps being run properly rather than quietly resented, and it turns a private intention into a commitment somebody else holds. Deniability is what lets a date slip.

Split the P&L before you split the product. Report the old line and the new line as two numbers — revenue, contribution, customers — even if they share a team and a bank account. Netflix did this in October 2011, replacing a blended domestic margin with separate contribution profit per segment. Split the reporting, not the customer's experience.

Grade the new-customer mix, never the installed base. Once a month, work out one figure: of the customers you won in the last thirty days, what share bought the old thing? Put it beside the same figure from a year earlier. About 7% of new Netflix members took the combined offer in 2011, while the disc base was still 13.9 million. The installed base measures decisions made years ago. The new-customer mix turns first.

Where the pattern breaks

The pattern has a bill and Netflix paid it in public. In July 2011 the company separated disc and streaming pricing, then announced a separate disc brand called Qwikster and retracted it weeks later. US subscribers fell from 24.59 million to 23.79 million that quarter, the only decline in the nine-quarter run Netflix published. Shares that had briefly touched $300 in July closed at $77.37 on 25 October. The company's own letter named the error: what it misjudged was how quickly to move. Hastings later called it roughly the right idea, five years too early. Direction and pace are separate decisions, and getting the first right does not buy the second.

What to do this week

Pick the line you would defend hardest in a meeting. Not the weakest one. The one you are proud of.

Work out a single number for it: of the customers you won in the last ninety days, what share bought that line? Then the same number for the same ninety days a year earlier. It is usually already in your invoices.

Then write one sentence naming the quarter in which you would stop selling it. Do not act on it and do not tell the team. Send it to one person who will remember you wrote it.

The two figures will either move you or they will not. Either answer is useful, and neither is available while that line sits inside a total.