Where onboarding stalls for the people who carry the blame
Six modeled operator archetypes were taken through Softr’s
marketing site and AI onboarding flow, screen by screen. They differ in tooling
proficiency, operational context, and emotional disposition. Every one cleared the
comprehension gates and stopped at the same one. Not because the product was hard.
Because they could not verify it.
6 archetypes
3 journey stages
9 question instrument
Method: synthetic panel
Panel generated withsyntheticusers.com
01 · Executive summary
The product is not too hard. It is unverifiable.
Six modeled training and enablement operator archetypes went through a first-run
onboarding journey in three stages. Every archetype understood the product, believed it
could use it, and wanted the outcome. None would put real work into it.
The sequence stalled at the same gate for every archetype, and it was not a
difficulty gate. They could not verify what the system had decided on their
behalf, and in a role where a wrong completion record is personally answerable, an
unverifiable system is not adopted no matter how easy it is. The two things the
product is proudest of, speed and finish, were both read as evidence against it.
The study does not observe a commercial outcome, but it produces a testable
prediction. The modeled response points to a specific false-positive activation pattern:
operators complete a trial or a bounded pilot while retaining the spreadsheet as the
trusted system of record. If that prediction holds in product data, trial activation
will overstate genuine adoption.
Where the sequence stalls
Gate 05: Evidence
Can I verify this rather than trust it? Blocked across every archetype
in the panel.
What cleared without difficulty
Legitimacy, relevance, mental model, capability
Nobody found the product hard. Ability was never the barrier.
The adoption criterion
Show me the logic and let me correct it
Emerged across the full modeled panel, in both emotional registers,
without verification being named in the instrument.
Governing law
3rd Law: consequence, not friction
Anticipated blame suppresses the behaviour before it is attempted.
Action 01
Promote the build log from a loading state to the product
It is the only element in the study that drew an unqualified positive, and the only
one that speaks to the gate that actually holds. It currently scrolls past while
the reveal is awaited.
PrimarySecondary
Salience is the lever. The reasoning already
exists and is simply not made prominent at the moment it would do the most work.
Action 02
Add correction before you add publishing
Showing what the system decided clears verification. Changing it clears reversibility.
The journey currently ends on a publish dialog that every archetype read as a trap, with
no correction path demonstrated first.
PrimarySecondary
Autonomy removes the loss-aversion block, and
correcting the system’s assumptions is also how ownership gets built. The fix is not
only defensive.
Action 03
Lead with the domain object model, not the app category
Conversion happened when archetypes recognised their own object names in the
generated data model. Nothing on the marketing sequence names a single one of them.
PrimarySecondary
Familiarity and relevance are the framing levers.
The category language forces translation work that the operator resents doing.
Strategy labels are drawn from the Make It Toolkit behavioral design
framework (Massimo Ingegno). Customer Centric Solutions LLC is a certified instructor for
the framework. Definitions are not reproduced here.
02 · What was studied
Softr’s onboarding, in three stages, as a real buyer meets it.
Everything that follows is anchored to these three sequences, so they come
first. This is also the part that decides whether findings are usable, and it is the part
most synthetic research skips.
SoftrSubject of study
The flow studied
Softr’s public marketing site and AI onboarding flow, as observed in July 2026.
Chosen because it is a well built, actively shipping B2B activation flow, which makes it
a harder and more useful test than a weak one. Softr did not commission, review, or
participate in this study.
Every one of the nine questions was posed against a
specific screen sequence, in the order a real buyer meets it: the marketing site, then the
first session after sign-up, then the path from generation to a working application.
The archetypes were not asked to imagine a product from a description. They were looking at
something, and answering about what was in front of them.
The difference is not cosmetic. An archetype asked would you trust an AI app
builder returns an opinion about a category. The same archetype shown the screen where the
system announces the app is ready, and asked what they believe at that moment, returns a
reaction anchored to a design decision you can change on Monday.
Why it matters for what follows
Nearly every finding in this study is located at a screen. The auth wall, the generated
table, the completion message, the logo row. That is only possible because the questions
were asked in usage context rather than in the abstract.
The three sequences are consecutive stages of one
journey, not competing designs shown in rotation. Each carries its own question block, and
each question in the instrument was asked against the sequence it belongs to.
Concept A
Website
Should I try this?
Questions asked against this sequence
What do you believe this product does, and is it relevant to your needs?
What increased or decreased your confidence in trying this product?
Would you start a free trial? Why or why not?
Every element the archetypes reacted to is visible here: the
headline claim, the enterprise logo row, the internal-only or external-facing
line, and the non-coder testimonial. The logo row split the panel in a way no card
variable predicted. The testimonial produced opposite reactions from two archetypes
with adjacent profiles. Nothing on this sequence names a training object.
Concept B
First Session
Can I do this?
Questions asked against this sequence
How confident are you that the AI understands what you are trying to build?
What moments increased or decreased your confidence during setup?
At the end of this experience, do you believe you could successfully build your application? Why or why not?
Screen one is the single open input, and it produces the
self-efficacy problem: the archetype worries about whether it is wording the prompt correctly before they have learned anything about the product. Screens two through four
are the clarifying questions, the strongest asset in the flow and the only place the
product speaks in curriculum structure. The theme picker sits at five, immediately
before the payoff, and it is the one screen that reliably lowered confidence.
Concept C
Path to First Meaningful Progress
Is this helping me do my job?
Questions asked against this sequence
At what point did you first believe the product was helping you make meaningful progress toward your goal?
How closely did the generated application match what you expected to create?
Would you continue using this product to build a real application? Why or why not?
The grid arrives first, at screens one and two, and that is where
belief lands for every archetype that reported it. The bar chart, where four of six
located first belief, is screen five. The completion dialog is screen six, which is
also where the search for the escape hatch begins. The sequence ends on the one
screen the panel treated as a liability.
Research basis
Where do synthetic eyes look?
There is no average viewer in your target market. A population heatmap
describes nobody in it.
Same pixels, different jobs
One product image, three participants, three different accounts of it. The three
sequences above are that result applied to an onboarding flow.
Portrait or film
A concept test is one image and one reaction. A user test follows a live session
screen by screen. The sequences above sit between the two.
78% of human ceiling
Predicted attention, graded against roughly three thousand images of recorded human
gaze, per participant rather than pooled.
This validates where a synthetic participant looks. It does not validate
the reasoning inside an interview, and we do not treat it as if it does.
03 · Findings
Six findings, replicated across every archetype
Each of these survived variation across the full modeled panel: the whole
tooling-capability spread and both emotional registers. None of them was a question in the
instrument, and none was prompted by it.
Section 04 places them in sequence and shows where the journey stopped. The
archetypes and the method behind them are set out in section 07.
01
They split the product into a shell and a set of rules, and trust only the second
Unprompted, every archetype separated what the product looked like from what it would do.
Three archetypes volunteered a numeric split without being asked for one, and every
split ran the same direction: high on structure, low on logic.
“It got the shell mostly right. I’m still waiting to see if the guts are right.”
Partner Education Manager, 39
02
Speed and polish are both read as warnings
Every archetype treated the velocity promise as evidence against the product, and
treated design investment as evidence of where the team spent its effort. The theme
picker lowered confidence. The publish dialog produced the sharpest negative in the
study.
“When the product emphasizes publishing before it proves governance, it feels backwards. It tells me the product may be proud of the wrong thing.”
Training & Compliance Manager, 49
03
Conversion happens when the data model uses their own object names
The recognition moment was not the interface, the dashboard, or the finished app. It was
a populated grid carrying the objects they think in. Recognising your own vocabulary
is a pre-rational event and it was the strongest positive signal recorded.
“Seeing Lesson Progress as its own entity is the moment I stopped thinking this was a toy.”
Enablement Lead, 45
04
The build log was the only element to earn an unqualified positive
Every other element drew a hedge. The visible record of what the system decided drew
none. Caution for the design team: showing what the system decided clears
verification, not correction. Those are separate gates.
“It feels less like magic and more like a build log. I don’t want a black box handing me a finished app and expecting trust.”
Training Operations Lead, 37
05
Nobody evaluates alone. Everyone rehearses against a named colleague
Every archetype simulated the decision through a specific person, and supplied that
person’s exact objection unprompted. The buying unit is not the operator in the
chair. The deciding criterion arrives secondhand.
“I can hear Derek saying, looks slick, but can it tell me which distributor reps are certified today, not last month?”
Partner Education Manager, 39
06
Two emotional registers, one shared criterion
Four archetypes were driven by exposure: they are the person who explains it when the
report is wrong. Two were driven by identity: they want to look like the person who built
the system. The criterion did not split.
“If a learner can skip the compliance quiz and still get marked complete, my audit trail is worthless.”
Enablement Lead, 45
04 · Where the sequence stalls
Every gate cleared until verification
The six findings above describe what the archetypes said. The gate ladder
places them in order and shows where the journey stopped. Each gate carries the paired event
underneath it: the belief change the design is trying to produce, and the barrier standing
in the way of it. Where the gate cleared, the barrier was overcome. Where it held,
the shift never happened. The pairs are gated, so a gate that does not clear closes
everything downstream, and the barriers below the stall are untested rather than absent.
A gate the sequence never reached can still be assessed, because every
archetype described what it expected to find there. Those expectations are marked
projected and drawn with a dashed rule: they are a forecast from the anticipated
barriers in section 05, not an observation. They carry real weight, since a barrier named
in advance by five or six archetypes is the best available signal about what happens next,
but they are not evidence that the gate failed.
The headline is not how many gates cleared. It is which one held, and how
little relief the gates below it offer.
01
Legitimacy
Is this real and worth my attention?
Shift
“Is this a scam?”→“This is legitimate”
Barrier
Beliefs and expectations
Cleared
02
Relevance
Is this for someone like me?
Shift
“For everyone”→“For someone like me”
Barrier
Prior knowledge and mental models
Cleared with friction
03
Mental model
Does this understand the work the way I do?
Shift
“Expecting X”→“Understand it is Y”
Barrier
Workflow incompatibility
Cleared
04
Capability
Can I do this myself?
Shift
“Too complicated for me”→“I can do this”
Barrier
Task complexity, choice overload
Cleared
05
Evidence
Can I verify it rather than trust it?
Shift
“Trust us”→“See for yourself”
Barrier
Trust and perceived risk
Blocked · all archetypes
06
Reversibility
Can I correct it when it is wrong?
Shift
“Permanent decision”→“I can change my mind”
Barrier
Poor error recovery, loss aversion
Not reachedProjected: At risk
07
Identity and reputation
Does adopting this make me look competent?
Shift
“This makes me look careless”→“This is how I want to be seen”
Barrier
Role and identity
Not reachedProjected: Splits by register
08
Authority
Am I the one who gets to decide?
Shift
“I need approval”→“I have authorisation”
Barrier
Access and dependencies
Not reachedProjected: At risk
09
Commitment
Do I move real work into this?
Shift
“Trial mode”→“Production use”
Barrier
Existing habits and routines
Not reachedProjected: At risk
Why
Psychological shifts
A shift and a barrier are the same event seen from two directions. The shift is the belief change the design is trying to produce. The barrier is what stops it.
Coding one without the other gives you a wish list or a complaint list. Coding them as pairs gives you a diagnosis: not what users disliked, but which belief failed to form, and what prevented it.
Read left to right, top row then bottom. Gates 01 through 04 are the
comprehension gates and they cleared without difficulty across every archetype. Nothing here
is a difficulty problem, and nothing here is a desire problem. The sequence stopped at
Gate 05, the first gate on the top row that asks for proof rather than understanding.
The projections matter as much as the stall. Clearing Gate 05 does not
release the sequence, because the three gates behind it are all forecast at risk on evidence
the archetypes volunteered before they got there. Reversibility and Commitment are the
product’s to fix. Authority mostly is not. Gate 07 is the only place the panel splits, and
it splits along emotional register rather than along capability, which makes it the least
safe projection in the set.
05 · Barriers to progress
Nine barriers encountered. Only one of them is about difficulty.
Every friction point was coded against a thirty-barrier taxonomy organised on
the five terms of the behavioural model: Person, Environment, Motivation, Ability, and
Prompt. What matters is not the count. It is where the count concentrates, and which
dimensions stayed quiet.
Because the sequence stalled at Gate 05, barriers are reported in two sets.
Encountered barriers fired at a gate the archetypes actually reached.
Anticipated barriers sit at gates 06 through 09, which nobody got to. The archetypes
described those in advance, which is real evidence about the segment but not evidence about
the experience. The two sets are never pooled.
Encountered barriers, by dimension
Person
3 of 6
Three of six Person barriers were encountered. This is where the study concentrates.
Environment
2 of 6
Partially observable in a single session.
Motivation
1 of 6
One of six. The struggling moment is strong and the payoff was believed.
Ability
1 of 6
One of six, and it is decision load rather than difficulty.
Prompt
2 of 6
Attention landed, but on the wrong next action.
Barrier
Dimension
Gate
How it appeared in this study
Seen in
Trust and perceived risk
Person
05
Would not connect real data or real people until they could see how completion is calculated and who can change it.
“A nice-looking generated app means very little if I can’t tell why certain statuses exist or how completion gets recorded.”
Learning & Development Manager, 41
6 of 6
Prior knowledge and mental models
Person
02, 03
Category language forced each archetype to translate into training objects themselves, and they named the translation as work.
“I’m the one mapping inventory management into role-based onboarding tracker. If you want me to sign up fast, don’t make me do that much work.”
Partner Education Manager, 39
6 of 6
Workflow incompatibility
Environment
03
Prerequisites, role-based assignment, recertification windows, exceptions and overrides. The logic the product did not visibly model.
“Can you handle: sales reps must complete modules 1 to 3 before certification, unless they transferred internally and already completed equivalent training last quarter?”
Training Operations Lead, 37
6 of 6
Attention on the wrong action
Prompt
05
The flow emphasised generation and publishing. Verification was never the cued next step. Inferred from verbatims, not from attention data.
“That’s the dangerous moment in AI products. It’s asking me to sit back and trust the machine. But this is exactly when I want transparency.”
Training & Compliance Manager, 49
6 of 6
Beliefs and expectations
Person
01
Prior experience of demos that looked tidy until a hard question was asked. Every archetype arrived pre-loaded with that pattern.
“Polished website, enterprise logos, then three demos later it turns out their reporting logic falls apart on basic audit needs.”
Training Operations Lead, 37
4 of 6
Low outcome expectancy
Motivation
05
Doubt that a generated system would produce reporting that survives an audit question.
“I’ve had tools give me very pretty charts on top of messy logic, and then I’m still manually validating numbers before I send anything to leadership.”
Learning & Development Manager, 41
4 of 6
Prompt lacks specificity
Prompt
04
The single open input asked the archetype to word a request correctly before they had learned anything about the product.
“It creates pressure to word things perfectly. Do I need to write like a normal person, or like I’m briefing a systems analyst?”
Learning & Development Manager, 41
4 of 6
Choice overload
Ability
04
Theme palettes and generic component blocks appeared at the moment structural confirmation was wanted.
“Are you building my app, or are you handing me a box of parts and calling it AI?”
Training Operations Lead, 37
3 of 6
Competing priorities
Environment
01
Evaluation was explicitly bounded. A Friday afternoon, twenty minutes, one contained test.
“I’ll poke at it for 20 minutes and see if it embarrasses itself.”
Training Operations Lead, 37
3 of 6
Anticipated barriers, at gates never reached
These are the barriers the archetypes named in advance,
at gates the stall prevented them from testing. They are the clearest signal available about
what would happen next, and they are not observations.
Barrier
Dimension
Gate
How it was anticipated
Named by
Poor error recovery
Ability
06
Fear of an uncorrectable mistake. The publish affordance read as a way to ship something wrong at scale.
“If a trainee swears he finished his module but the system says he’s at zero, I need to go into the back end and see exactly what triggered that.”
Learning Program Manager, 34
6 of 6
Loss aversion
Motivation
06
The known-bad spreadsheet felt safer than an unproven system.
“If I still have to maintain the real truth in spreadsheets because I don’t trust the generated app, then this is just another shiny extra tool.”
Partner Education Manager, 39
6 of 6
Existing habits and routines
Person
09
Every archetype proposed keeping the current process running in parallel rather than replacing it.
“Somebody ends up back in a spreadsheet just for now, which turns into forever.”
Partner Education Manager, 39
6 of 6
Role and identity
Person
07
Archetypes refused the builder identity the product offered. They wanted to remain accountable operators.
“I don’t need to bring ideas to life. I need a system that won’t embarrass me in front of my VP or compliance lead.”
Learning & Development Manager, 41
5 of 6
Access and dependencies
Environment
08
Progress routed through an IT lead, single sign-on, or an HRIS connection before real data could enter.
“I need to see how I can connect this to our HRIS. If I have to manually invite every new hire one by one, the automation doesn’t mean much.”
Enablement Lead, 45
5 of 6
What did not fire, and why that is the finding
Four barriers that dominate most onboarding studies are absent here. Task complexity
and knowledge or skill gap never appeared: nobody found the product hard to use.
Unclear personal value and weak struggling moment never appeared either:
every archetype arrived with an acute, specific pain and understood exactly what the
product would do about it.
That absence is what makes the diagnosis specific. Of the nine encountered barriers, one
sits in Ability, and it is decision load rather than difficulty. The four that fired across every archetype are about consequence, comprehension mismatch, and misdirected
attention. An onboarding programme that responds by reducing steps, shortening time to
value, or adding guidance is optimising the dimensions that were never blocking.
06 · Interpretation
The barrier is consequence, not difficulty
Person held constant
Not the barrier
The diagnosis
M and A both present
Avoidance dominates
All five laws were checked against the evidence. Four of them are
satisfied, which is what makes the fifth reading so specific.
Read against the Five Laws, the diagnosis is unambiguous. The 1st Law holds the
environment constant by design, so this is a controlled read on the person. The 2nd Law
is not where the failure sits: nothing here is a comprehension problem. Under the 4th
Law, motivation was present, since the struggling moment is strong and everyone believed
the payoff, and ability was present, since nobody found the product hard. The prompt
landed too: attention went where the design intended.
What suppressed the behavior was the third law. The anticipated consequence of a wrong
record is severe and personally borne, and that consequence is evaluated before the
behavior is attempted. No amount of added motivation or reduced friction moves an
operator past a gate that is held shut by anticipated blame.
In Make It Toolkit
terms, this segment arrives with Make it Aversive already fully
engaged, and engaged against the product. Nobody needs to be shown the cost of getting it
wrong. They carry that cost into the first screen. Which is why the strategies that would
normally open an activation flow, making it easier, faster, more attractive, do very little here, and why the counter-levers
are Make it Obvious and Make it Empowering: reveal the reasoning, then hand
over control of it. Under the 5th Law this segment is running almost entirely on
extrinsic avoidance, and avoidance of blame is a stronger force here than any reward the
product offers.
Which reframes the design problem. The work is not to make onboarding easier. It is to
make error recovery and logic inspection visible before commitment is requested. For
this segment, verification is not a late-stage enterprise concern. It is the activation
moment.
07 · The archetypes, and what the method supports
Six modeled archetypes, and the limits of what they can tell you
What a synthetic archetype represents
These six are not six recruited individuals. Each is a modeled archetype generated from a
much larger body of population, attitudinal, behavioural and language data. The individual
profile is what gives that archetype internal coherence across an interview: a stable
operational context, personality structure, vocabulary, priorities, and emotional response
pattern.
The analytical unit is therefore the reaction produced by a configuration within the
defined audience, not an isolated opinion. Convergence carries weight when the same
criterion appears across materially different configurations, and particularly when they
reach it by different reasoning paths. This does not turn six interviews into a
projectable survey. It supports population-informed diagnosis. It does not estimate what
share of the market holds a belief, and it does not predict conversion lift without
behavioural validation.
The panel was generated rather than recruited, and we say so plainly. Attention
prediction in this class of platform is validated against recorded human gaze. The
reasoning content of an interview is not.
Digital tooling proficiency was the one variable
deliberately controlled, spanning levels 2 through 5 so the non-technical operator
configuration was represented rather than screened out. The remaining four dimensions were
measured as descriptors.
Danielle Brooks
Learning & Development Manager, 41
Chicago, IL
Digital tooling3
Decision authority4
Regulatory rigor4
Workflow complexity4
Data integration3
O 4C 5E 3A 4N 3
Miguel Alvarez
Training Operations Lead, 37
San Antonio, TX
Digital tooling2
Decision authority3
Regulatory rigor5
Workflow complexity4
Data integration2
O 3C 5E 4A 4N 2
Priya Raman
Enablement Lead, 45
Seattle, WA
Digital tooling5
Decision authority5
Regulatory rigor3
Workflow complexity5
Data integration5
O 5C 4E 3A 3N 3
Tasha Greene
Learning Program Manager, 34
Atlanta, GA
Digital tooling4
Decision authority3
Regulatory rigor4
Workflow complexity3
Data integration4
O 4C 4E 5A 5N 2
Evan McAllister
Training & Compliance Manager, 49
Columbus, OH
Digital tooling3
Decision authority5
Regulatory rigor5
Workflow complexity5
Data integration4
O 2C 5E 2A 3N 4
Marisol Vega
Partner Education Manager, 39
Fresno, CA
Digital tooling2
Decision authority4
Regulatory rigor3
Workflow complexity4
Data integration3
O 4C 4E 4A 5N 3
Each archetype carries two profiles. The five bars are operational
context. The radar is the OCEAN personality profile that drives the emotional layer of the
interview: how fast wariness escalates, how long persistence holds, whether an unfamiliar
interface reads as curiosity or as confusion.
Two patterns are worth reading off the radars. Conscientiousness sits at
4 or 5 for all six, which is the mechanism behind the most uniform behaviour in the
study: every archetype offered a bounded pilot rather than walking away. And the two poles
are visible in shape alone. Evan carries the lowest Openness and the highest Neuroticism in
the panel and produced the most skeptical account. Tasha carries the lowest Neuroticism with
the highest Extraversion and Agreeableness and produced the most enthusiastic one.
A note on reading the context bars. Only digital tooling proficiency was
controlled in recruiting, and only it separated the panel behaviorally. The other four
dimensions clustered in the upper range and did not predict any reaction. Descriptors
that are measured but not controlled cannot be read as segmentation cuts, and we do not
report them as such.
Behavioral profiles, not just role profiles
Each archetype carries an OCEAN personality profile alongside its operational
context. This matters more than it sounds. Onboarding failure is rarely a comprehension
failure. Two operators can read a screen identically and diverge completely on whether
they proceed. Trait-driven emotional transition is the variable that separates them:
how fast wariness escalates, how long persistence holds under friction, whether an
unfamiliar interface reads as curiosity or as threat, whether failure gets attributed
to the tool or to the self.
In the Five Laws frame this is the P term of B=f(P,E). Concept testing holds the
environment constant by design, which makes the study a controlled read on the person.
Without a behavioral profile layer, the person term is empty and every archetype
returns the same reaction to the same screen.
Research basis
Chain of feeling
An emotional narrative runs in parallel with the reasoning. OCEAN governs
how it moves.
Openness · 2 to 5 across the panel
Decides whether a new design registers as curiosity or as confusion. This is the
widest spread in the panel, and the archetype at 2 produced the most skeptical
account in the study.
Conscientiousness · 4 or 5 for all six
Drives perseverance before giving up. This is the twenty-minute evaluation budget:
bounded, but genuinely offered. It is also why every archetype proposed a pilot
rather than walking away.
Agreeableness · 3 to 5 across the panel
Can produce self-blame rather than anger at the interface. That supplies a mechanism
for the most-cited pattern in this study: fearing you will look gullible rather than
concluding the product is bad. In this panel the pattern appeared at every level of the
trait, so we report it as a mechanism the trait can supply, not one it predicted here.
The emotional layer is generated by design. We read it as directional
evidence about where friction concentrates, never as measurement of felt experience.
Analysis frame
Every friction point was coded as a paired event. A psychological shift and a behavioral
barrier are the same moment seen from two directions: the shift is the belief change the
design is trying to produce, the barrier is what stops it. The pairs are gated. A gate
that does not clear closes everything downstream, so a missing late-stage shift is not
evidence of a late-stage problem. It is evidence that the sequence stalled earlier.
What this study supports, and what it does not
Open any line for the reasoning behind it.
This study can support
Use these directly
Locating the barriers and the stall gate
Where a first-run sequence loses the operator, which belief failed to form, and which barrier prevented it.
Projecting the gates below the stall
Whether the gates the sequence never reached would clear if the stall were fixed, forecast from barriers the archetypes named in advance. These are marked as projections, and they are the study's most useful output for roadmap sequencing.
Comparing across audience configurations
How reactions differ across the modeled capability spread and both emotional registers, and which findings survive that variation.
Identifying recurring decision criteria
Criteria that appear across configurations without being named in the instrument. Verification was never asked about and emerged in every interview.
Prioritising what to change and test next
Generating design changes and selecting what to put in front of real users or measure in product data.
It does not independently establish
Needs other evidence
Conversion lift or revenue outcomes
The study produces a prediction about activation. It does not measure one, and no commercial journey was observed.
Market incidence or prevalence
It does not estimate what share of operators hold a belief. Six archetypes are configurations, not a sample frame.
Causal effects of personality traits
Conscientiousness sits at 4 or 5 across the whole panel, so its contribution cannot be isolated. Trait patterns illustrate range, they do not demonstrate causality.
Segment differences on uncontrolled dimensions
Only digital tooling proficiency was controlled in recruiting. The other four context dimensions clustered in the upper range and did not discriminate.
Preference between the three stages
The stages are consecutive rather than competing, and run in the same order for every archetype. Any apparent preference for later stages belongs to the ladder, not the designs.
Read with care
Method caveats
Emotional arcs are directional only
The generation method produces an emotional narrative by design, and the instrument asked about confidence at nearly every step, so a well-formed arc was the guaranteed output. Arc shape is not reported as a design result.
Attention data was not collected
This study used the screen-sequence method but did not generate per-archetype attention maps. Screen-level attribution rests on what each archetype said about each screen, and the Prompt-dimension barriers are inferred from verbatims rather than observed.
Convergence was examined before it was reported
The criterion converged more strongly than the explanation: archetypes reached the same requirement, inspect and correct the logic, through different accountability contexts, examples and consequences. That increases confidence in the criterion. Separately, three archetypes produced near-identical phrasing around a named colleague, and that similarity is treated as a possible generation artifact rather than a finding.
What separates the two emotional registers is open
The two enthusiastic archetypes are also the two highest on tooling proficiency, and OCEAN alone does not cleanly separate them from the rest. Both candidates are reported rather than one being chosen.
Recommended next
Validate the activation prediction against product data, and run a human-participant wave against a revised instrument that drops the confidence framing in favour of decision and past-behaviour questions.
Next step
Where does your onboarding stall?
Most onboarding work optimizes the gates that already clear. The stall is usually
somewhere else, and it is usually a consequence problem rather than a friction problem.
The onboarding audit finds the gate that holds and names the barrier holding it.
syntheticusers.com
