A rock formation · Inspired by Zion National Park · And a way of working
It is a band of Early Jurassic river-rock on the Colorado Plateau, sitting between two of the Southwest's great cliffs — and the clearest model we've found for how good product work gets made. Start with the rock. The rest follows.
Where it sits, how old it is, and what it managed to keep.
The Kayenta is the odd member of the Glen Canyon Group: softer, so it erodes into ledges while the Wingate below and the Navajo above stand as sheer walls — the notch that lets you read the whole group at a glance.
Navajo Sandstone
Cross-bedded dune sand, hundreds of metres deep, burying the Kayenta river system as the desert took over. These are the walls of Zion Canyon.
Kayenta Formation
Red-brown sandstone with siltstone and limestone, laid down by shifting rivers. Cross-bedding inside it shows the desert reaching this far south — briefly, more than once — before it took hold for good.
Wingate Sandstone
The massive lower wall of the group. Where Kayenta rock caps it, the cliff holds; where the cap is gone, it retreats.
Chinle Formation
Not part of the group — the surface everything above was deposited onto. Nobody chose it; it was already there.
The formation spans the Early Jurassic, roughly 198.0–178.8 million years ago. Then a second, independent method — detrital zircon dating — put a number on it.
Pliensbachian–Toarcian, marked by the white tick above. One number, one error bar, and an estimate meaningfully younger than the old range. Remember it; we come back to it.
Braided channels, floodplains, standing water, occasional dry spells — deposited at the wet edge of the desert that became the Navajo. Conditions favoured preservation, and it shows.
Shifting streams
Fluvial sandstone, cut and stacked by rivers rising and falling with the seasons.
Mudstone flats
Siltstone recording the quiet spells between floods — and the footprints.
Lakes and ponds
Freshwater limestone with mussels, snails and ostracods — lake patches on the plain.
Dune incursions
Cross-bedded sandstone: the desert reached this far south, briefly, and left the proof.
The closest modern analogue for Kayenta time: rainy summers, dry winters, at a desert's southern edge — the climate North America was steadily leaving behind.
Body fossils
The animals themselves
The formation's best-known theropod and its main identified dinosaur.
A small armoured thyreophoran — the commonest dinosaur here.
A massospondylid sauropodomorph from Gold Spring Quarry 1.
The earliest known true frog — the first hips built for jumping.
An early caecilian, still with limbs.
A tritylodont — a mammal relative, named for the formation.
One of the oldest turtles known from North America.
A dimorphodontid pterosaur.
Trace fossils
The marks they left
Large three-toed theropod tracks — the biggest predator prints in the formation.
Slender theropod tracks, a smaller counterpart to Eubrontes.
Three-toed reptile tracks with long, clawed impressions.
Ornithischian footprints, often preserving a hand impression beside the foot.
Stand in Zion Canyon and the wall stops you: two thousand feet of Navajo Sandstone, made entirely of sand the rivers below delivered.
Look down instead — the bench you're standing on is the Kayenta. Softer, less photogenic, and it's what carries the fossils and keeps the wall from retreating.
That is the company in one cross-section. Everyone wants the wall. The work is the ledge.
Everything above, applied to building software people actually want.
Stratigraphy reconstructs a process nobody watched from what it left behind — the job description for user research, too. Every term on the left has a mirror on the right.
Superposition
In an undisturbed sequence, lower means older.
Evidence precedes decision
Every design choice traces to the finding that justified it.
Differential erosion
Kayenta caprock stops the Wingate cliff below from crumbling.
Design protects research
Research that never ships erodes into an unfindable deck.
Body vs. trace fossils
The animal, or the footprint — two different kinds of evidence.
Attitudinal vs. behavioural
What people are, or what people did — never let one answer the other's question.
The zircon date
A better method produced a number with an error bar and moved the estimate.
Nothing ships undated
Method, sample and confidence travel with every claim — so it can be revised.
Unconformity
A surface where time is missing. Geologists draw it and estimate what is gone.
Gaps stay visible
Unresearched decisions are logged as assumptions with an owner.
Facies change
Channel sandstone here, lake limestone twenty miles away. Same system, different rock.
Segments aren't averages
We map where the experience differs instead of averaging it into a persona.
Same column as Part One, relabelled. Start at the basement — nothing on top exists without what's below.
Phase 01 — Inherited ground
The legacy product, the tech debt, the old research decks — surveyed before we add to them.
- Heuristic audit
- Prior-research archaeology
- Stakeholder map
- Constraint register
- Analytics baseline
Phase 02 — Foundational research
Generative UXR before a single pixel — one coherent body of understanding, not a scatter of quick reads.
- Research plan
- Recruiting & screeners
- Field & remote sessions
- Synthesis
- Opportunity map
- Prioritised problems
Phase 03 — Design in the open
Many thin beds — concepts, prototypes, usability rounds — instead of one heroic reveal, so stakeholders can walk the reasoning at eye level.
- Concept sets
- Information architecture
- Prototypes
- Evaluative rounds
- Content design
- Decision log
- Design system work
Phase 04 — Ship & scale
Specs that survive handoff, instrumentation that answers next year's question, and a system that scales the pattern.
- Specs & redlines
- Build partnership
- Launch instrumentation
- Adoption tracking
- Post-launch read
Skeletons in the quarries, footprints in the mudstone — confusing the two is the costliest mistake a product team makes.
What people are
Rare, expensive, richer than any track — and never a large enough sample to count.
- Methods Interviews, contextual inquiry, diary studies, moderated usability
- Answers Why. Motivation, mental model, vocabulary, workaround, felt cost
- Fails at Frequency. Six interviews are never a percentage
What people did
Common, cheap, continuous — honest about behaviour, silent on why.
- Methods Telemetry, funnels, session paths, search logs, tickets, churn cohorts
- Answers What, where, how often, at what scale
- Fails at Intent. A path is not a reason
Palaeontologists name the footprint, not the species — a track is a guess dressed as a fact. We hold the same line: the trace wins on what happened, the body wins on why. A finding needing both gets both, or ships half-evidenced.
Geologists date twice: relative for sequence, absolute for a number with an error bar. Findings earn their place the same way.
Stated, unsupported. Carries an owner and an expiry date.
One method, small n — safe to design against, not to bet a roadmap on.
Two methods agree. Design proceeds.
Quantified, with margin. Survives handoff and turnover.
The number from Part One, again: a better method moved the age, and the field moved with it — nobody defended the old range on seniority. Findings should be revisable without anyone losing face.
An unconformity is missing time. Geologists don't paint over it — they draw it, name it, and estimate the loss.
Above the surface
Decisions that were made, shipped, and are now load-bearing for everything after them.
Below the surface
The last research that actually touched this problem — possibly two reorgs ago, possibly for a different product.
Dry intervals happen
Deadline-driven, assumption-led stretches happen — real, survivable, and recorded as what they were.
Every gap has an owner
Each unresearched decision is logged with who made it and what would settle it.
One reference record
Every engagement gets one readable record of what we learned, decided, and deferred.
You rarely need the whole column. Pick the exposure that fits.
Survey the section
A short, sharp read of what you already have.
- Heuristic and accessibility audit
- Prior-research archaeology
- Analytics and support-ticket read
- Gap map with confidence labels
Cut a new bed
All four phases, on one problem worth doing properly.
- Foundational research programme
- Concept sets and prototypes
- Evaluative rounds with real users
- Spec, build partnership, launch read
Field school
We run the method with your team until your team runs it without us. Capability, not dependency.
- Research operations setup
- Interview and synthesis coaching
- Decision-log and evidence practice
- Design system contribution model
Rivers laid the Kayenta down for twenty million years, legible today because the sequence was never broken. Speed feels like progress; sequence is what survives the reorg.
Evidence below · Design above · Nothing undated