The Design-Led Architecture Stream
Give design a protected block to define the product's building blocks — or it will smuggle scope in at the end.
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 87%
After a year of pure feature velocity, Mixpanel found each new feature had low reach because it had to be rebuilt for every part of the product. The fix was to spin up a second, parallel stream that was design-led and purely about system architecture: which core building blocks exist, how few can there be, how do users discover them, how do they relate. Crucially, this required freezing architecture-affecting work for three months so designers could come up for air.
Origin
Instituted at Mixpanel around 2019-2020 (coinciding with their adoption of Figma) by Vijay Iyengar, then an engineering manager, with the PM and head of design. Explicitly modelled on how Notion's pages-and-blocks architecture lets features hang off a few strong primitives.
Core principles
- 01Many great products win or lose on their architecture, not their feature count.
- 02A small set of strong building blocks multiplies the reach of every future feature — new capabilities inherit the benefits automatically.
- 03Design used as a final coat of paint is a waste of a design team.
- 04Designers with no dedicated space for design-led work will smuggle simplification into tactical projects at the last minute — which reads as end-of-project scope blowup.
- 05The protected block is real only if you also commit to not shipping architecture-impacting work during it.
How to run it
- 1
Detect the low-reach symptom
Look for diminishing returns on features and the tell-tale sign that every improvement must be rebuilt separately for each part of the product. That is a holistic-design deficit, not a velocity problem.
- 2
Freeze architecture-affecting work and buy design a runway
Explicitly commit that the next three months of projects will not require design, and that no architecture-impacting work will ship in that window. Iyengar calls this a controversial thing to say; it is what created breathing room.
Pro tip Get the PM and head of design in the room and make the trade explicit — you are trading three months of design polish for a durable system.
Watch out Without the freeze on architecture-impacting work, the designers' output will be obsolete before it lands.
- 3
Answer the four architecture questions
Have the design group define: what are the key building blocks of the product; where do they fit; how few of them can we have; how will users discover them and how do they relate to each other.
Pro tip 'How few can we have' is the load-bearing question — Iyengar calls it a really important step.
- 4
Run it in parallel with the core grind
Keep the tactical gap-closing stream running alongside. The two streams are complementary: one stops the bleeding, the other makes the product coherent.
- 5
Take on projects purely for consistency, reach, and depth
Regroup, set strategy, and greenlight work whose only justification is improving the consistency, reach, and depth of the UX — not shipping a new customer-facing feature.
Pro tip Make design one of your stated key differentiators so this work has a durable mandate rather than being a one-off.
In the wild
The biggest output of the design-led phase was visualisation consistency: making charts interactive in the same way across the entire product. Any new visualisation, enhancement, or sort behaviour added in one report instantly applied everywhere, and every new visualisation inherited the accumulated benefits.
→ Reach of every added feature multiplied; the product became more accessible and dark mode became feasible, making visualisations easier to read at a glance.
Before the dedicated stream existed, design was pulled in at the end of tactical projects and asked to 'make this look nice'. Designers used each project as the only available opportunity to squeeze in simplification, which repeatedly blew up scope at the worst moment.
→ Decoupling design-led work into its own protected space removed the friction point and stopped the recurring late-stage scope explosions.
Common mistakes
Bringing design in at the end to add pixels
Using a talented design team as a final polish layer on engineering-shaped projects wastes them and guarantees scope creep, because the last mile becomes their only chance to influence the product.
Promising design time without freezing architecture work
If the team keeps shipping architecture-impacting changes during the design block, the resulting system design is stale on arrival. The freeze is the mechanism, not a courtesy.
Is it for you?
Best for
Product teams emerging from a feature-velocity sprint who now find each new feature has low reach and must be rebuilt per surface
Not ideal for
Companies still actively bleeding customers on missing table-stakes features — put the fire out first
From the transcript
“we were hitting diminishing returns with some of these features and like not considering the holistic design and consistency meant the reach of every feature…”
“we can actually do the next three months of projects about any design which was a kind of controversial thing to say but we're doing…”
“that's just a classic way to blow up scope at the end of the project because there wasn't a dedicated space for design-led projects”
From the episode
An inside look at Mixpanel’s product journey
Vijay Iyengar (Head of Product)