Utility Curves
A utility curve maps the value a user extracts from a product against the effort or investment required to reach that value. Stewart Butterfield used the model at Slack to reason about which user segments to invest in and how much improvement in any dimension was worth pursuing.
The shape
The curve is an S-shape. On the x-axis: investment — effort, features, integrations, polish. On the y-axis: utility extracted by the user.
The typical structure:
- Long flat entry. Most users, spending moderate effort, extract little value. The product is not yet useful to them.
- Steep middle. A core segment, once past a threshold, extracts enormous value from incremental investment. This is where the product’s competitive advantage concentrates.
- Plateau. Further investment produces diminishing returns; the segment is extracting near-maximum value and additional features add little.
Butterfield describes this colloquially as: ‘junk, junk, good, great, then flat.‘
Why segment matters
The utility curve is not a single curve for all users — it varies by segment. A team using Slack for complex asynchronous coordination across time zones sits on the steep part of the curve; additional features create large value. A team using Slack as a glorified IRC sits on the flat entry; the same investment creates little.
These two segments coexist in the same product at the same time. Treating them identically is an investment mistake.
Strategic implications
Investment allocation. Before any improvement decision, identify which users it serves and where they sit on the curve. Serving a steep-part segment with even modest investment produces high returns. Serving a flat-part segment requires either accepting low returns or doing a different kind of work: moving them further along the curve.
Accepting limits. Some segments will never extract much value regardless of investment. Explicitly acknowledging this — rather than trying to serve everyone equally — frees investment to concentrate where returns are highest.
When to stop. The plateau identifies the point at which a segment is at or near maximum value. Additional investment in features that serve them specifically yields little. Either serve a new segment or invest in a new capability that creates a new curve.
The @channel anecdote
Butterfield used utility curves to evaluate the @-channel mention redesign at Slack. The product team proposed restoring a pre-populated @mention in thread reply boxes; instrumentation showed a marginal increase in average thread length (2.17 vs. 2.14 messages).
His rejection: the expected value of the decision was guaranteed to be negative regardless of the statistical result. The improvement served a segment that was already on the flat part of the utility curve — an average thread growing by 0.03 messages reflects near-zero utility increase. The cost of the analysis (feature flags, A/B infrastructure, dashboards, rescheduled meetings) far exceeded any plausible benefit. Running the test was itself the mistake.
Relation to other frameworks
Utility curves share logic with the concept of diminishing returns in economics and with the Jobs to Be Done distinction between under-served and over-served customers. The specific contribution is the segment-aware S-curve: not a single diminishing-returns function but a curve that differs by user type and can be steep for one segment while flat for another.