Optro Scenario Planning

Same risk, two rooms, no shared math

You're the Risk Manager, and you've spent two weeks building a scenario: mapping causes, scoring controls, pressure-testing every consequence you can think of. Then you walk into the boardroom, and the CFO asks one question. "What's the dollar exposure?" Everything you've built is correct, but the language doesn't carry. Two weeks of work, dead on arrival, because the room speaks in dollars and you showed up with a qualitative model.

That gap was the problem. Risk Managers reason in ambiguity because their job lives in it. Executives reason in quantified ROI. Same risk sitting in front of both of them, two completely different vocabularies, and no shared math to bridge the two.

The gap between them wasn't knowledge, it was syntax.

The brief was "make scenario planning work." The way I read it: give risk a language the boardroom already speaks.

Research insights, on a sprint

I led design end-to-end, owning the interaction model, visualization, simulation, and reporting. PM, UX research, and Lead Eng fed into every major decision. I inherited four months of customer research from a prior phase and an accelerated timeline. The job was to filter what was already there and gather more in the same motion, testing on limited evidence and updating the model with every pilot session.

I filtered everything through one question: what are the most important pieces for problem identification? Three patterns surfaced fast, and they shaped every decision downstream.

Five directions explored, and why four didn't hold up

I researched how data-dense information is presented across the GRC space. I looked at the metaphors competitors and adjacent products lean on, then pressure-tested the five most common against how Risk Managers actually reason.

• Spider / Radar

  Visual clutter obscured the relationships between cause, control, and consequence.

• 5×5 Risk Matrix

  Too high level. The subjective labels lost the nuance Risk Managers care about.

• Gantt Timeline

  Over-indexed on process. Risk trade-offs happen simultaneously, not sequentially.

• Fishbone Cause

  Couldn't integrate qualitative and quantitative narratives in one read.

• Bow Tie

  Chosen

  Mapped cleanly to the way Risk Managers already think on whiteboards.

The bow tie was the one that clicked. Causes feed an event on the left, consequences fan out on the right, controls clamp both sides. It was a familiar shape with room to score every node.

Wizard vs. Canvas, and the third option I pushed for

The next question was how to present the bow tie. Before settling, I prototyped both ends of the spectrum to see what each one revealed in practice.

Engineering didn't have a preference here. They wanted a clear direction and the right answer for the user. The prototypes alone weren't conclusive, but the pattern felt familiar from other data-dense workflows I'd worked on: Risk Managers need the whole bow tie in view to think, and a single decision in focus to act. No one mode gets you both. So I pushed for a third option that tried to combine the strongest piece of each.

Wizard

• ·Step-by-step guidance
• ·Splits attention
• ·Cramps the canvas

Canvas

• ·Full picture at a glance
• ·Easy to get lost in
• ·Data entry breaks the view

Canvas + Wizard

Chosen

• ·Canvas for the overview
• ·Wizard only for heavy lifts
• ·One decision at a time, in context

Generating scenarios faster

The bow tie was the right shape. But Risk Managers told me the bottleneck wasn't drawing it. It was starting it. Every scenario began as a blank page, and getting from blank to draftable took hours.

Two principles drove the V2 design:

• No blank page. Risk Managers start from AI-sourced suggestions, never a blank canvas.
• Risk Manager in the driver's seat. Three generation modes match the Risk Manager's appetite for control, from one node at a time to the whole bow tie in one shot.

1Most user control

Item by Item

The Risk Manager generates one node at a time, in place. Best when the scenario is half-known and the AI is filling specific gaps. Slowest in raw output, but the user is in command of every piece.

2Balanced

Plan Mode

The AI proposes a structured plan; the Risk Manager checks what to apply before anything lands on the canvas. The second axis of control lives here (Library Only, Mix, or Hypothetical Only), so the Risk Manager scopes the AI's imagination to what the scenario calls for.

3Most AI control

Build Mode

One-shot generation of the entire bow tie. Fastest path from blank to draftable. The Risk Manager edits after the fact rather than directing the AI in flight, useful when the scenario is well-known and the goal is a starting point fast.

Meeting users on the AI trust curve

The three modes weren't a progression. They were entry points. Risk Managers vary in how much AI they want in their workflow, and the most trusted product is the one that flexes to where the user is, not where we want them to be.

1. 1

   Item by Item

   “I'll drive. Suggest one node at a time.”

2. 2

   Plan Mode

   “Propose a plan. I'll review before it lands.”

3. 3

   Build Mode

   “Build the bow tie. I'll edit after.”

Inside any mode, the Risk Manager can sharpen the AI's suggestions with a Focus Prompt, a short keyword or theme that biases what the AI proposes. Focus gives guidance without removing control. The story behind each item stays clear.

Prototypes in the same week as the interview

AI coding tools changed what V2 could be. V1 was almost entirely Figma: the bow tie canvas, the Wizard prototype, the simulation and analysis flows. V2 lived in Replit, and the speed honestly caught me off guard.

Clickable prototypes landed in pilot calls within the same week as the interview that surfaced the need. Iteration loops I'd never had access to before. Customers shaped the product in real time, and I rebuilt between sessions.

The curveball that almost killed the simulation

The UI was only half the work. The math underneath had to earn the same trust. Enter Monte Carlo, and the three-step trap that comes with it.

1. 1

   All good

   Cause 1

   Likelihood ##%

   Prevention Control 1

   Strength ##%

   Mitigation Control 1

   Strength ##%

   Consequence 1

   Exposure $####

   All good

   Risk Managers expect every object to have data behind it.
2. 2

   Curveball

   Lowest

   Median

   Highest

   Curveball

   Risk Managers think in 1–5 scales, not percentages. Exposure ranges feel natural to them; likelihood and control strength don't.
3. 3

   Dead on arrival

   Best Case

   Likelihood ##%Exposure $####

   Most Likely Case

   Likelihood ##%Exposure $####

   Worst Case

   Likelihood ##%Exposure $####

   Dead on arrival

   Shaky confidence in the inputs means shaky confidence in the outcome. If the data feels wrong, the simulation does too.

Step one was fine: Risk Managers already expect every object to have data behind it. Step three was a known property of the simulation, best-case, most-likely, worst-case outputs.

Step two was the curveball. Monte Carlo needs ranges, and ranges need percentages. Exposure ranges felt natural. Risk Managers price out dollar bands all day. Likelihood and control strength didn't. Force them into raw percentages and the inputs would feel wrong; once the inputs feel wrong, the simulation is dead on arrival.

How I solved the curveball

I kept the label on the front and the percentages underneath.

Likelihood to happen this year

• 1. Very UnlikelyVery unlikely to occur in the next 12 months
• 2. UnlikelyUnlikely to occur in the next 12 months
• 3. PossiblePossible occurrence in the next 12 months
• 4. LikelyLikely to occur in the next 12 months
• 5. Extremely LikelyAlmost certain to occur

Pick from the 1–5 scale Risk Managers already use

Likelihood to happen this year

3. Possible▾

Possible the risk may occur in the next 12 months

One label selected

Generated range

Best

30%

Most Likely

40%

Worst

50%

The simulation runs on the range. The user never sees a sigma.

Range generated for Monte Carlo

Users pick from a familiar 1–5 scale, the language they already use. Each label maps to a generated range that feeds the simulation. The math gets richer inputs; the user never feels like they're guessing a single "right" answer.

The lesson I took from this: a small change at the input layer made a much bigger change at the output. The 1-5 scale felt like nothing to the user, but it gave the simulation richer data than a single percentage ever could.

The unlock: an output both rooms could read

This is where the original problem finally resolved. Risk Managers and executives had spent years talking past each other. Same risk, two vocabularies. The analysis view became the room where the languages met.

Getting there meant partnering closely with our Lead Engineer, the team's in-house Monte Carlo expert. We worked in lockstep for weeks: I'd bring a layout, he'd pressure-test whether the visualization was honest to the math. He'd surface a distribution property, I'd find a way to make it readable. The analysis page was the product of that loop, not a handoff.

Inputs the user understands

Design

Familiar 1–5 scales, plain-language labels, and verify-over-input patterns. The Risk Manager never feels like they’re guessing.

Simulation users can trust

Lead Engineer

Monte Carlo engine, probability distributions, and range math. The numbers underneath hold up to scrutiny.

Analytics page that tells the story

Shared

Percentile tables, S-curves, and leading indicators. An output both the Risk Manager and the CFO can read in the same room.

The team landed on three things so the simulation could be defended by Risk Managers and understood by executives, in the same view:

• Percentile tables for confidence. "P50 is $175K, P90 is $210K." Auditors and analysts speak this.
• S-curve graphs for credibility. The shape of the distribution, not just the headline. Executives finally saw the story, not a single scary number.
• Leading and lagging indicators for direction. Which way the risk was moving and what was moving it.

Risk Managers stopped agonizing over a single number. Executives stopped second-guessing the math. Two audiences, one view. That was the moment it started working.

What customers said

1. “

   This brings credibility to my job.

   Risk Manager

2. “

   I have been looking for something like this for years.

   CyberRisk Manager

3. “

   You don't know how hard it is to get executives to understand these things. This changes that.

   Risk & Audit Manager

4. “

   I need this. Now.

   Audit Manager

From insight to decision

Reviewing the simulation isn't where the Risk Manager's job ends. They still have to walk into the room and propose action to the executive team. Pulling that proposal away from the scenario that prompted it would have been disorienting, so the Response side of Scenario Planning lives inside the same flow. Action plans stay rooted in the analysis that produced them.

Reporting that started a conversation

The same V1/V2 pattern showed up on the reporting side. V1 was about getting the data into one place. V2, the part I spent the most time on, had to set the course of the boardroom conversation.

Three rules that guided V2:

• Reports start at the source. No copy-pasting between tools. Analysis to artifact in a single click.
• The right format for the right room. Slides for executives, an email for the quick update, a Word doc for the audit team.
• Curated data, not data dumps. Risk Managers shape the story before the meeting, not during it.

V1 got every input into one place. V2 gave Risk Managers a way to walk into the room with a finished artifact, not a pile of data to explain.

$1.4M in pipeline

The brief was "make scenario planning work." The pipeline told us whether we did.

• $1.4M in pipeline from market response
• $240K ARR converted from pilot users
• EMEA expansion unlocked by the capability
• Forrester recognition in the GRC space

Where I'd push harder

1. Validate sooner

   At one point I caught myself pixel-pushing, seeking perfect spacing and color before I'd validated the approach. The polish was burning runway I didn't have.

2. Collaborate more frequently

   When I got stuck, a colleague's perspective almost always showed me I was zoomed in too far.

3. Test interactions before visuals on data-dense screens

   I let myself polish before validating whether users could actually edit data efficiently. Next time, interaction prototypes come first.

Same room, shared math

The CFO still asks about dollar exposure. Now the bow tie answers in a language the room already speaks.