Stop Guessing How Good Your Design Org Is. Here's the Rubric.

You know the feeling in a review when someone asks how mature your design system is, and the answer is a shrug and a story. "We're getting there." That answer doesn't hold up anymore. There are now real rubrics you can score your team against, out loud, in a room. Let me catch you up on three of them and how to use them Monday.

Maturity isn't a ladder, it's a shape

The old story goes: build components, drive adoption, hit scale, reach steady governance. Nice and tidy. But NN/g's new framework throws out the ladder. Design systems don't climb in a straight line. A merger drops a second system with clashing tokens, and your "mature" system is suddenly fighting foundational battles again.

Instead you score six dimensions: organizational alignment, team effectiveness, infrastructure robustness, governance, support, and adoption. Plot each one and you get a shape, not a rank. That shape is the point. It shows you the imbalances, like a strong component library nobody trusts, or great tooling with no executive sponsor to protect the budget.

One line landed for me: a system "cannot survive an organization that has decided it isn't worth the investment." You can limp along with thin docs. You can't limp along without funding. Score alignment first.

Run the assessment like a team, not a solo grade

Here's the part you can actually do this week. Pull together four to eight people from different seats: design-system folks across design, engineering, product, and content, plus a couple of product teams who use the thing daily, plus a sponsor who sees the org politics.

Everyone scores all six dimensions on a 1 to 5 scale, independently. A 1 is ad hoc, no owner. A 3 is functional but fragile under change. A 5 is mature and holds through reorgs. Then you triangulate. Dimensions where scores cluster mean shared understanding, note it and move on. Dimensions where scores split wide are your real conversation.

The gaps are the gold. When a product designer rates governance a 2 and the system lead rates it a 4, you found a blind spot before it cost you.

Adoption numbers lie if you stop at usage

Most teams report adoption as a single number: how many teams touch the system. That hides the truth. NN/g splits it into three layers worth grading separately. Usage is just access. Conformance is whether teams apply it right or quietly build their own variants. Trust is whether they believe the system will keep up with their needs.

You can have high usage and low trust. Teams use the buttons because they have to, then work around everything else. That's a system rotting under a green dashboard. Ask which layer your "adoption" metric actually measures, then measure the other two.

The same rubric trick works on your chatbot

If you're shipping a site-specific AI chatbot, NN/g gives you five qualities to grade it against: handoff willingness, flexibility, proactivity, emotional responsiveness, and transparency. Same idea, different artifact. Score where you sit before you ship, not after users abandon it.

The sharpest rule is on handoff. Never gatekeep a human. One study participant described the AI runaround as "a hamster wheel kind of spinning around and around." Wilson's Virtual Agent refused a human after three tries. Wyze handed off on the first ask. Guess which one users trusted. When someone asks for a person, give them a person.

Flexibility runs on a spectrum too. Too narrow and the bot only answers FAQs. Too broad and it drifts off-brand and burns compute answering things you never authorized. Aim for the middle: handle questions adjacent to your product, redirect the ones that aren't.

The deep cut

The real value in these frameworks is the disagreement they force, not the final score. A rubric turns "our system is fine" into a room where a product lead says 2 and you said 4, and now you have to explain the gap. That conversation is the deliverable.

So don't chase a perfect shape or a five across the board. A ten-person startup and a ten-thousand-person enterprise can both be mature, because maturity is contextual to your scale. Bring the scored profile to your next review and point at the widest gap. That's the one thing your team disagrees about, which means it's the thing quietly costing you.

Three questions for your team

1. If we score all six dimensions today with a mixed group, which one gets the widest spread of scores, and what does that gap tell us we're not talking about?
2. Our adoption metric measures usage, conformance, or trust. Which one, and do we have any read on the other two?
3. For our chatbot or any AI feature, when a user asks for a human, how many steps does it take them to get one, and can we cut that to zero?