The Balanced Feedback Portfolio
Treat every feedback channel as a biased sensor, then hold a portfolio so you have no blind spots.
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 90%
Every feedback source Figma has is systematically biased: support tickets skew dissatisfied, sales calls surface perception rather than usage, and Twitter over-represents designers who happen to be online. Yamashita's answer is not to pick the 'real' signal but to deliberately hold a portfolio of channels, each understood for its bias, and to treat low-engagement complaints as canaries rather than as demands. Figma operationalized this with a private 'concerning tweets' channel that contains the CEO's firehose without letting it derail teams.
Origin
Emerged from Figma's founding practice — Dylan Field putting the product in front of designers, collecting feedback, and returning with it implemented — and from a specific failure mode: Field dropping tweets he found concerning into open Slack channels until teams felt obliged to drop everything and react. Yamashita and colleagues intervened to build the containment channel. He layered on research and data functions after arriving.
Core principles
- 01A viral tweet is not the signal; the zero-like tweet with an essence of truth might be.
- 02Every channel has a bias — support skews angry, sales skews perception, Twitter skews who is on Twitter.
- 03The point of the portfolio is blind-spot elimination, not consensus.
- 04Loud feedback deserves investigation, not automatic implementation.
- 05Founder-level customer obsession has to be routed, or it becomes an interrupt storm.
How to run it
- 1
Inventory your channels and name each one's bias
List every place customer signal arrives — support tickets, sales conversations with prospects, social, research, product data — and write down what each one systematically over- and under-represents.
Pro tip Ask specifically: who is NOT in this channel? For Figma, most of the audience was not on Twitter even though the loudest designers were.
- 2
Build the missing capabilities
Where a channel is absent, invest to create it — research and data functions exist precisely to balance the vocal minority against what is actually happening across the base.
Watch out If you only have one strong channel, you will mistake its bias for reality.
- 3
Treat loud signals as canaries in the coal mine
A tweet or ticket is an input, not an instruction. Read it as an early-warning indicator that something bigger may be wrong, then go check the other channels for corroboration.
Pro tip The tweets worth investigating are often the ones with one like or zero likes but an essence of truth to them.
Watch out Not everything gets implemented verbatim — Figma explicitly does not.
- 4
Create a containment channel for exec-sourced feedback
Route founder/CEO customer observations into a small, private, low-stakes channel with a designated group who triage them — instead of into public channels where teams feel compelled to drop everything.
Pro tip Figma calls theirs 'concerning tweets.' The name sets the expectation: worth a look, not an emergency.
Watch out Without containment, a CEO's genuine customer obsession becomes an organization-wide interrupt.
- 5
Ask the community directly, in the open
When picking a focus area, tweet the question out — 'we're thinking about focusing more on prototyping, what are the top problems you're seeing' — and read all of it for pulse, not for a mandate.
In the wild
Dylan Field, who has read customer feedback daily for a decade, used to drop tweets he found concerning into various Slack channels. Once Figma got big enough, teams felt they had to drop everything and deal with each one. Yamashita and colleagues created a small private channel called 'concerning tweets' where Field can drop them for a designated group to triage.
→ The CEO keeps his direct line to user sentiment, low-signal complaints still get evaluated for whether they point at something bigger, and teams stop being derailed by a single tweet.
Figma's team was exceptionally good at Twitter and staying on top of designer sentiment there — and luckily many designers are on Twitter. But Yamashita recognized on arrival that most of Figma's actual audience is not.
→ He prioritized building capabilities to extract insight from research, data, and other sources so the loud channel would not define the roadmap.
Common mistakes
Treating the vocal minority as the market
The people who tweet, file tickets, or shout in sales calls are not a representative sample. Without balancing channels you will build for the loudest cohort and miss the silent majority entirely.
Letting exec feedback arrive as an interrupt
When a founder posts a customer complaint into a public channel, teams read it as a directive and abandon their priorities. The feedback is valuable; the delivery mechanism needs a container.
Is it for you?
Best for
Product leaders at companies with a passionate, vocal community whose loudest users are threatening to become the roadmap.
Not ideal for
Very early startups with almost no users, where the loud minority IS the market and over-indexing on them is correct.
From the transcript
“we created this new channel private Channel called concerning tweets”
“it's really important to kind of balance out the vocal minority with you know what's actually happening right so I really view some of those…”
“I think it's just important especially as a product manager to feel like you have this balanced portfolio of different kinds of feedback to know…”
From the episode
An inside look at how Figma builds product
Yuhki Yamashita (CPO of Figma)