n o ren
Economics & Markets

Stop Using “One‑Size‑Fits‑All” Pricing

Netflix's 2011 price change cost nothing for single-service subscribers and about sixty percent more for everyone who wanted both.

2026-09-082 min read

A single elasticity number for a product is an average across customers who respond to price in opposite directions. That average describes no actual buyer. Some segments barely notice a change because the product is load-bearing for them; others sit a few dollars from the exit and leave the week the invoice changes. When a firm reads the aggregate and concludes demand is inelastic, it is usually reading the loyal segment's insensitivity plus the marginal segment's sensitivity, netted into a figure that hides both.

Netflix supplied the textbook case in 2011. The company separated streaming and DVD-by-mail into two plans priced at $7.99 each. A subscriber who used only one service paid what they had always paid, while a subscriber who used both went from $9.99 to $15.98, roughly a sixty percent increase, so that group absorbed the entire change. Netflix lost about 800,000 US subscribers in the third quarter of that year, announced a spin-off of the DVD business called Qwikster, and abandoned Qwikster about three weeks later as the reaction got worse. The headline move was modest by any aggregate measure. Concentrated on one segment, it was not modest at all.

The trap is not that the price rose. It is that the decision was evaluated on a number that averaged away the only thing that mattered, which is who was being asked to pay it. Firms repeat this whenever they test a change against blended churn, blended revenue per user, or a blended conversion rate. The blend is where the damage hides, because a segment can be gutted while the average moves a point or two. Before a price moves, the question worth answering is not how much demand will fall, but whose demand, and whether that group is the one carrying the rest of the business.

Key insights
An aggregate elasticity is a weighted average of opposing responses and describes no real customer.
Price changes rarely fall evenly, and the segment absorbing the change is usually invisible in the headline number.
A modest blended increase can be a severe increase for the subset that buys the most from you.
Why it matters

A blended elasticity figure can look safe while the segment it averages over is the one funding the business.

Pricing decisions justified on aggregate churn survive review precisely because the aggregate is the wrong unit, so the same error repeats every cycle.

Use this tomorrow
1
Pull your last price change and split the churn it produced by plan or segment; count how many segments moved at more than double the blended rate.
2
List every customer segment that pays for two or more of your products together, then count how many of your last three pricing decisions changed what that group pays.
Go deeper

Aggregation problems of this shape are old in economics: a demand curve estimated across a heterogeneous population is a composite, and its slope belongs to no individual in it. The practical consequence is that elasticity estimated from past aggregate moves predicts the next one poorly whenever the mix of customers has shifted underneath it. Firms with a bundled or multi-product base are the most exposed, because the bundle is exactly where a small headline change concentrates into a large one. The fix is not a better average; it is refusing to make the decision at the level of the average.

There is a second-order effect Netflix demonstrated and few firms plan for: the segment hit hardest is also the loudest, so the reputational cost arrives well before the revenue cost surfaces in a quarterly number. Netflix reversed the Qwikster structure within weeks but kept the price separation itself, which suggests leadership read the backlash as being about the split rather than about who was paying for it. That distinction matters, because only one of the two is fixable by reorganizing. By the time the aggregate finally moved, the lag was long enough that the causal link was still arguable internally.