Dashboard Design Guide

A North Star for the Dashboard Team

What Are Our Users Trying to Do?

Our users are professional investors — analysts and portfolio managers at asset management firms. Their job is to beat a benchmark (like the S&P 500) by finding investments that will outperform.

Term	What It Means
Alpha (α)	Returns above the benchmark. This is how investors prove their value. No alpha = no reason to exist.
Edge	An informational or analytical advantage that lets you predict outcomes better than the market. Edge creates alpha.
Consensus	What "the market" collectively believes. To generate alpha, you must believe something different from consensus — and be right.
Thesis	The specific reason an investor believes a stock will outperform. "I think margins will expand because..."
Price Target (PTA)	The price an analyst believes a stock will reach. Usually with scenarios: bull case ($120), base case ($100), bear case ($80).
Diffusion	The process of a thesis becoming recognized by the market. As others "catch on," the price moves toward your target.
Hedge	A position that offsets risk you don't want. If you're bullish on Apple specifically but don't want general tech exposure, you hedge the tech risk.

The investor's challenge: Find situations where you know something the market doesn't, filter for situations where beliefs transmit reliably to outcomes, size your bets appropriately, and exit before the opportunity disappears.

How Our App Supports Them

Our app is an alpha management system. It doesn't tell investors what to buy — it helps them capture, size, track, and learn from their investment beliefs.

Step	What the Investor Does	How the App Helps
1. Form Belief	Research a company, develop a thesis	Outside scope (AI tools can assist, but belief is human)
2. Apply Constraints	Check if opportunity passes policy gates (management quality, transparency, etc.)	Constraint capture and enforcement — filter before sizing
3. Capture Belief	Document price target, scenarios, confidence	Structured input forms capture beliefs in usable format
4. Size Position	Decide how much to invest	Optimization engine calculates sizing based on conviction + track record
5. Monitor	Track if thesis is playing out	Dashboards show thesis status, market movement, remaining upside
6. Exit	Sell when thesis is priced in or wrong	Alerts when "forward Sharpe" declines (upside exhausted)
7. Learn	Review what worked and didn't	Scorekeeping tracks accuracy over time; calibrates future confidence

Where Dashboards Fit

Dashboards support steps 2 and 5-7: constraint enforcement, monitoring, alerting, and learning. They do NOT support step 1 (forming beliefs) — that's human judgment, potentially assisted by AI tools outside the app.

Dashboards Show

Constraint status (does this pass policy gates?)
Current state (exposures, performance)
Thesis progress (is it working?)
Alerts (what needs attention?)
Track record (how accurate am I?)

Dashboards Don't

Tell users what to buy
Generate investment ideas
Make predictions
Replace human judgment

The Key Insight: Alpha comes from human beliefs and constraint policies. The app captures both and helps deploy them efficiently. Dashboards are the monitoring layer — showing constraint status, state, progress, and accuracy. They don't create alpha; they help preserve it.

The Absolute Return Problem: Why Risk Monitoring Matters

There's a second, equally important reason dashboards exist — rooted in how asset management actually works.

The Business Reality: Clients want both alpha (beating the benchmark) and absolute returns (actual money). Alpha alone isn't enough.

Here's the math that drives everything:

Component	Typical Amount
Alpha (stock-picking skill)	2-3% per year if you're good
Beta (market return)	7-10% long-term average
Client expectation	8-12% total return

The problem: If a PM hedges away all market risk, they're left with only alpha — maybe 2-3%. But clients expect 8%+. So PMs must intentionally keep market exposure (beta) to generate absolute returns.

The consequence: PMs are carrying systematic risk (market moves, sector rotations, style shifts) that they have no edge on. They can't predict these — they're just along for the ride. But they need the return.

This is why dashboards matter:

PMs are choosing to carry risk they can't predict
They need to monitor that risk actively: "What am I exposed to? Is it behaving? Should I adjust?"
Dashboards aren't about eliminating risk — they're about managing risk you're intentionally keeping

Not This

"Dashboards help you hedge away all risk"

This

"Dashboards help you understand and manage the risk you're choosing to keep"

This is arguably the biggest practical reason for Functions 4-5 (Hedge Identification, Hedge Execution). Users aren't trying to eliminate exposure — they're trying to monitor exposure they need but can't predict.

Concrete Example: Alpha Corruption

Alpha corruption = when market/factor movements (that you can't predict) overwhelm your stock-picking returns.

Dashboard Shows	What It Means
"You have 0.3 loading on Value factor"	PM has exposure to "value stocks" as a group
"Value factor is down 8% this month"	Value stocks are getting crushed
"Your alpha: +3%, Factor drag: -2.4%"	Stock picks are working, but Value exposure is killing the result
Decision	"Should I hedge some Value exposure to protect my alpha?"

This workflow — see loadings → see factor performance → see impact → decide to hedge — is what Functions 4-5 dashboards enable.

What Data We Have Today

Before designing dashboards, know what's actually available:

Data Type	Plain English	Source
Holdings	What stocks we own, how much of each, what % of portfolio	Our trading system
Price Targets + Scenarios	Analyst predictions: "I think it goes to $120 (bull), $100 (base), or $80 (bear)"	Captured in our app
Prices	Current and historical stock prices	Market data feed
Wall Street Estimates	What other analysts expect for earnings, revenue, etc.	FactSet (data vendor)
Market Factor Returns	How "the market," "growth stocks," "value stocks," etc. performed	Factor data vendor
Company Financials	Revenue, earnings, margins, balance sheet data	FactSet
Historical Predictions	What our analysts predicted in the past + what actually happened	Captured in our app

Derived Data We Can Compute

From the raw data above, these metrics can be calculated:

Metric	Plain English	Buildable?
Expected Price Target	Average of bull/base/bear targets weighted by their probabilities. "If I'm 20% bull, 60% base, 20% bear, what's my blended target?"	Yes
Expected Return	How much upside is left? (Expected target ÷ current price) - 1	Yes
Forward Sharpe	Expected return adjusted for risk. High = good risk/reward remaining. Low = thesis mostly played out.	Yes
Analyst Accuracy (Ω)	How good is this analyst's track record? Compares their stated confidence to actual outcomes over time.	Yes
Stock-Picking Return (α)	Return from picking the right stocks, minus market/sector moves. "Did we win because we're smart, or because tech went up?"	Yes
Market Exposures (β)	How much of our portfolio moves with the market, with growth stocks, with value stocks, etc.	Yes
Consensus Divergence	How different is our view from Wall Street's view? Big divergence = potential alpha (if right).	Yes

Practical vs Aspirational

~85% of dashboards are buildable with existing data. The remaining ~15% require new data capture (detailed thesis documentation, key driver identification, portfolio manager views). Know the difference before committing resources.

Phase	What's Included	Data Status
Phase 1: Build Now	Risk-adjusted upside remaining, stock-picking vs market return, analyst accuracy scores, comparison to Wall Street, basic performance/risk	Data exists today
Phase 2: Enhance	Hedge recommendations, thesis status ("on track" / "at risk"), implied market expectations	Needs additional logic
Phase 3: Aspire	Detailed thesis capture (key drivers, evidence checkpoints), advanced market-implied analytics	Needs new data capture

The Core Principle

Edge enters only through human beliefs AND human constraint policies. Everything else is infrastructure.

The app is an alpha management system, not an alpha generation system. Dashboards capture, display, and support decisions — they don't make the decisions.

Two Sources of Human Edge

Source	What It Is	Example
Red Nodes (Beliefs)	Non-consensus probability assessments	"I believe pricing power is stronger than consensus expects"
Gate Nodes (Constraints)	Policy-based filters that determine participation	"I don't invest in companies with poor management quality"

Red nodes generate alpha through belief mispricing. Gate nodes generate alpha through bet selection — filtering for situations where beliefs transmit reliably to outcomes. Both are human judgment. Both contribute to alpha.

What Dashboards Are For

Function	Purpose
Constraint Enforcement	Show policy gate status — does this opportunity pass admissibility filters?
Monitoring	Show current state — exposures, performance, thesis status
Workflow	Capture structured inputs — beliefs, constraints, scenarios, approvals
Alerts	Flag when attention is needed — limits, triggers, constraint violations, anomalies
Scorekeeping	Track accuracy over time — who was right, what worked

What Dashboards Are NOT For

Dashboards don't tell users what to believe.
They show data. They don't say "buy this stock" or "this sector will outperform." That judgment comes from humans.

Design For

"Where am I exposed?"
"Is my thesis on track?"
"How accurate have I been?"
"What needs attention?"

Don't Design For

"What should I buy?"
"Is this a good stock?"
"What does this news mean?"
"Generate alpha for me"

The Seven Dashboard Functions

All dashboards serve one of these purposes:

#	Function	Plain English
1	Consensus Extraction	"What does Wall Street think?" — Shows market expectations so users know what they're disagreeing with.
2	Consensus Monitoring	"Is Wall Street changing its mind?" — Tracks if analysts are revising estimates up or down.
3	Diffusion Monitoring	"Is my thesis working?" — Shows if the market is moving toward the user's view (price approaching target).
4	Hedge Identification	"What risks am I taking that I didn't intend?" — Shows unintended bets (e.g., accidentally heavy in tech).
5	Hedge Execution	"How do I offset that risk?" — Suggests ways to neutralize unwanted exposures.
6	Scorekeeping	"How accurate have I been?" — Tracks who was right, who was wrong, and why.
7	Core Analytics	"How are we doing?" — Standard performance and risk metrics. Table stakes — everyone has these.

Who Uses What

User	Primary Dashboard Needs
Analyst	Belief capture forms, diffusion tracking, accuracy review
Portfolio Manager	Aggregate views, optimization outputs, exposure monitoring
Risk Manager	Exposure dashboards, limit compliance, drawdown alerts
Leadership	Analyst rankings, system health, process metrics

Dashboards vs AI

AI is becoming the primary interface for exploration and queries. Dashboards remain valuable for:

Dashboards Best For	AI Best For
Ambient monitoring (wall screens)	Ad-hoc questions
Structured data entry	Natural language queries
Regulatory reporting	Cross-data synthesis
At-a-glance state awareness	Open-ended exploration

Design implication: Dashboards should be optimized for monitoring (always-on, ambient awareness), not exploration (AI handles that better).

Where the Value Really Lives

Dashboard displays are easy to replicate. What's hard to replicate:

Low Value (Easy to Copy)	High Value (Hard to Copy)
Visual design, layouts, charts	The calculations underneath the display
Standard metrics everyone shows	Our proprietary metrics (analyst accuracy, risk-adjusted upside)
Generic reporting	AI/ML trained on our historical prediction data

Prioritize the data layer. A beautiful dashboard on bad data is worthless. Solid derived data with a basic display is valuable.

Design Checklist

Before building a dashboard, ask:

Which of the 7 functions does this serve? (If none, reconsider)
Who is the user? (Analyst, PM, Risk, Leadership)
Does it support constraint enforcement? (Policy gates are alpha-generating)
Is the data available today? (Check Phase 1/2/3 — don't design for data we don't have)
What derived data does it require? (Can we compute it from existing raw data?)
Monitoring or workflow? (Optimize for the right mode)
Could AI answer this better? (If yes, maybe don't build a dashboard)
Does it generate proprietary data? (If yes, prioritize it — this builds moat)

Data-first rule: If the required data doesn't exist, the dashboard is Phase 3 (aspirational). Don't build UI for data we can't populate.

For full context, see: Alpha Capture: From Principles to Practice