Learning design case study · Aaron Gomez

Nexus
Rush

A 12-minute branching simulation that takes new sales reps from “what's nexus?” to winning over a skeptical CFO.

See how it was built
Plays in your browser · no login · about 12 minutes
academy.tallyline.com/nexus-rush
Role
Sole designer & developer
Audience
New Account Executives
Format
Scenario-based web simulation
Seat time
12 min · 5 modules
Tools
Figma · Claude · Claude Code

From a sales problem to a playable game, in six moves.

ADDIE, run in fast loops
  1. 01AnalyzeFound the real performance gap.
  2. 02DesignSet measurable goals and a course map.
  3. 03Storyboard & scriptMapped every screen and branch.
  4. 04Collaborate with AIDrafted fast, decided carefully.
  5. 05BuildShipped a real, accessible web app.
  6. 06Test & evaluateThree test rounds, four levels of evidence.
01Needs analysis

New reps were pitching before they were listening.

Tallyline's new AEs knew the product cold. They were losing deals in the first ten minutes of the call. Before designing anything, I'd find out why. Tallyline is fictional, so this is the needs analysis I'd run, with illustrative findings.

Our new AEs can demo every feature. They just can't get a CFO to admit she has a problem.
Sales enablement leadTallyline (fictional) · the kind of kickoff quote that frames the analysis
How I'd gather the data

A plan, not a record. This scenario is fictional, so there were no real stakeholders or SMEs to interview. On a real project, this is the evidence I'd collect before designing anything.

  • 3Stakeholder interviews with sales managers and enablement
  • 40Recorded first calls to review and code
  • 12New AEs to take a short nexus pre-assessment
  • 6Months of CRM data to pull on deal velocity by tenure
68% of first calls pitched the product within five minutes, before a single pain was named. Illustrative · call audit, n = 40
3 in 10 new AEs could correctly call nexus for a state when given its threshold. Illustrative · pre-assessment, n = 12
21 days from first call to qualified opportunity for new AEs. Tenured reps: 12. Illustrative · CRM, 6 months

Is training the fix?

Gap analysis · not every gap is a training gap
GAPCan't tell if a prospect has nexus
CAUSEKnowledge. Rules taught as law, never as decisions.
TRAIN · state-by-state practice
GAPPitches before finding a pain
CAUSESkill. Shadowing showed good calls but never let reps practice one.
TRAIN · branching call
GAPStalls at “we've been fine”
CAUSESkill. No shared model for objections, so reps reached for scary stats.
TRAIN · ACRC reply builder
GAPNexus notes get lost after the call
CAUSEEnvironment. The CRM has no field for it.
NOT TRAINING · sent to Sales Ops
02Objectives & outline

One business goal. Three things every rep must do.

Every screen traces to a measurable objective. Every objective traces to the number the sales team already watches.

Business goal

Cut the time from first call to qualified opportunity for AEs in their first 90 days.

21days today
14days target
Measurable course goalWritten in ABCD format

Given a prospect's state sales data and a live discovery conversation, new AEs will determine economic nexus and surface the prospect's tax exposure before pitching, scoring 78% or higher on an unaided transfer scenario.

A · Audience B · Behavior C · Condition D · Degree

Performance objectives

O1BLOOM · APPLY

Decide whether a prospect has economic nexus in a state, using that state's sales and transaction thresholds.

O2BLOOM · APPLY

Run discovery questions that surface a prospect's sales tax exposure before pitching a solution.

O3BLOOM · ANALYZE

Diagnose the concern behind an objection and respond with Acknowledge, Clarify, Reframe, Confirm.

Course map

Merrill's First Principles · support fades as skill grows
Intro · 1 minMission briefingACTIVATIONA message from your manager sets the stakes.
Support
M1 · 2 minLearn the conceptDEMONSTRATIONFive idea cards, a job aid, and a senior rep's worked example.
Support
M2 · 3 minThe nexus mapAPPLICATIONFlag 9 states for Ridgeline, with hints on.
Support
M3 · 3 minThe discovery callAPPLICATIONTwo branching decisions. Dana's trust reacts to each.
Support
M4 · 2 minThe objectionAPPLICATIONDiagnose “we've been fine,” then build an ACRC reply.
Support
M5 · 2 minTransfer scenarioINTEGRATIONNew prospect, new states. Hints off. Scored.
Support · none
Wrap · 1 minDebriefFEEDBACKScores by objective, the move to make next time, and a badge.
Support
O1 · O2 O1 O2 O3 O1 · O2 · O3 Reports all three
03Storyboard & script

Mapped on a canvas before a line of code.

I storyboarded the whole experience in Figma as one path across a canvas. That path became the product: the game's camera pans the same route.

Interactive · switch pages in the file
Nexus Rush / {{ figTitle }}
Share
Pages
Layers {{ ly }}
01 Landing
02 Briefing
03 M1 · Concept
04 M2 · Nexus map
05 M3 · Discovery call
06 M4 · Objection
ACRC
07 M5 · Transfer
08 Debrief
FLOWThe game's camera pans along this exact path, frame to frame.
SME: re-verify every threshold before launch.
Trust meter reacts to every choice.
Hints switch OFF from here on.
M2 · Nexus map · Desktop 1440
Hints on1
2
3
4
  1. 1Hints switch. Learner controls support here. Locked off in Module 5.
  2. 2Tile map. Every state the same size, so small states are as easy to hit as Texas.
  3. 3Threshold bars. Sales vs. the line, drawn. The decision is visual, not mental math.
  4. 4Two buttons, instant why. A binary call, then feedback that explains the rule.
COLOR
Night Tally Sky Gold Cream Ink Canvas
TYPE
AaPublic Sans
Headings & body
AaIBM Plex Mono
Labels & data
COMPONENTS
Start → Continue → 120 XP ● Nexus — No nexus
CAST · 22 LEARNER AVATARS + 5 CHARACTERS
SCRIPT · M3 DISCOVERY CALL · DECISION 1 OF 2Trust starts at 50
DANA

“We're growing fast, and honestly sales tax isn't on my radar. We charge it in Colorado and our accountant handles the rest.”

REP SAYSTRUSTDANA REPLIESCOACH FEEDBACK
A“Great. Let me show you how our platform automates filing.”−20“I think we're fine for now.”Pitched before finding a problem. Dana has no reason to listen yet.
B“Where are most of your online orders shipping these days?”+15“All over, honestly. Texas and Florida have really taken off.”An open question about her business, not your product. Unlocks the multi-state pain.
C“Did you know you could owe back taxes in dozens of states?”−25“That sounds like a sales tactic.”Fear first puts a CFO on guard. Let her find the risk in her own answers.

Rules I scripted by

  1. 01Every wrong answer is a mistake real reps made in the call audit.
  2. 02Dana reacts in character first. The coach explains after.
  3. 03Feedback says why, not just right or wrong.
  4. 04Plain language. Short sentences. No tax jargon without a definition.
04Collaborating with AI

AI drafted.
I decided.

Claude was my collaborator at every phase. It made me faster. It never made the calls about accuracy, learners, or tone.

RESEARCHClaude
AISummarized Wayfair and drafted a state-by-state threshold table.
MEChecked every threshold against state revenue sites. Cut the ones in flux.
SCRIPTINGClaude
AIGenerated a dozen wrong-answer variants for each decision.
MEKept the ones real reps actually say. Rewrote all of them in Dana's voice.
PROTOTYPINGClaude
AITurned the Figma storyboard into a clickable build in hours, not weeks.
MEDirected every screen, interaction, and avatar, one revision at a time.
BUILD & QAClaude Code
AIRefactored, fixed edge cases, and flagged accessibility issues.
METested by keyboard and screen reader myself. Approved every change.
A REAL REVISION · M4 REFRAME LINEfrom the working session

Write a Reframe line for Dana's “we've been fine” objection.

AG
CLAUDE · DRAFT

“Audits are rising every year, and penalties can be severe.”

MY NOTETrue, but it's fear. A CFO hears pressure. Use her own numbers instead.
SHIPPED

“Since 2018, states can require collection based on sales alone. From what you shared, that likely applies in four states beyond Colorado.”

The rejected draft didn't go to waste. It became a wrong answer in Module 4.

Guardrails

  • No real customer or learner data in a prompt.
  • Every tax fact traced to a primary source.
  • Every line read aloud in character before it shipped.
  • Accessibility checked by hand, not just by tool.
05Building the game

A real web app, not slides with buttons.

Built with modern web tools, so it runs in any browser, loads fast, and treats keyboard and screen reader users as first-class players.

Tech stack

DESIGNFigmaStoryboards, wireframes, UI kit
PROTOTYPEClaudeStoryboard to clickable build
DEVELOPClaude CodeRefactor, polish, edge cases
FRONT ENDReact + JavaScriptComponents and game state
STYLE & MOTIONHTML + CSSLayout, animation, reduced motion
GRAPHICSSVG web componentOne avatar engine, 27 characters
STANDARDWCAG 2.2 AAKeyboard, screen reader, contrast

How it's put together

Content lives in plain data, separate from the interface. An SME can update a state's threshold without touching a single screen.

1Content dataStates, dialogue, choices, feedback
2Game stateTrust, flags, XP, hints, pains found
3Interface8 screens on one pannable canvas
content · states.js1 of 51
{
  id: 'TX',
  name: 'Texas',
  sales: 450000,
  orders: 3800,
  ans: 'X',  // no nexus
  why: "$450K is under Texas's
        $500K threshold."
}
The why field is the feedback learners see. Content and explanation travel together.

What makes it work

It's your call, literallyLearners pick a name and one of 22 avatars. Both follow them into every conversation.
Choices have consequencesDana's trust rises and falls with every line. Pitch too early and she closes off.
A model, not a scriptLearners build a reply step by step: Acknowledge, Clarify, Reframe, Confirm.
Practice on a real mapA 51-tile map with real thresholds. Learners flag each state and see the rule behind it.
Support that fadesHints are on while learning and locked off for the final test, so the score reflects real skill.
  • Full keyboard play
  • Screen reader live regions
  • Never color alone
  • Respects reduced motion
Accessible by designBuilt in from the first wireframe, not bolted on at the end.
06Testing & evaluation

How I'd test it with the people it's for.

Two rounds I'd run with real SMEs and new reps, plus an accessibility round that's already done. Kirkpatrick's four levels define how we'd know it worked.

ROUND 1Planned · 2 tax SMEs
Content accuracy
WHAT IT WOULD CATCHThresholds that changed after the draft was written, like a state dropping its transaction rule.
ALREADY BUILT INThe job aid says “check it's current,” and the game reflects Illinois dropping its transaction rule in 2026.
ROUND 2Planned · 5 AEs, think-aloud
Usability
WHAT I'D WATCH FORReps skimming the worked example, or missing the hints control.
ALREADY BUILT INThe example reveals one line at a time, and hints are a labeled switch.
ROUND 3Done · keyboard + code
Accessibility
FOUNDMap states relied on color. New dialogue lines weren't announced.
CHANGEDEvery state got an icon and label. Every conversation got a live region.

Evaluation plan

Kirkpatrick levels 1 → 4
LEVEL 1 · REACTIONDid it feel worth 12 minutes?Readiness rating, built into the debrief.Target · avg 4 of 5
LEVEL 2 · LEARNINGCan they do it without help?Transfer scenario with hints locked off. Scored by objective.Target · 78% to pass
LEVEL 3 · BEHAVIORDo they do it on real calls?Managers score three behaviors on the rep's next real calls.Target · 2 of 3 by day 30
LEVEL 4 · RESULTSDid the business move?Days from first call to qualified opportunity, tracked in the CRM.Target · 21 → 14 by day 90

Next iteration

  • Report scores via xAPISend Level 2 results straight to the LMS.
  • Manager dashboardMake Level 3 checklists one tap on a call.
  • Voiced dialogueLet reps practice listening, not just reading.
Your turn

Your first discovery call is in 48 hours.

Meet Dana, find the exposure, and earn your Nexus Navigator badge.

Open in a new tab ↗
Designed and built by Aaron GomezInstructional & learning experience designer
PortfolioLinkedInEmail
Tallyline, its people, and all figures in this case study are fictional, created for this portfolio sample. Tax thresholds in the game reflect published state rules at time of build.
Nexus RushEsc to return to the case study
New tab ↗