Executive Brief · Decision Memo

Prediction markets: scale only after proving demand, operational safety, and unit economics.

ToShift Markets senior leadership
FromSaren Abgaryan
DateAugust 6, 2026
ReDecision on scaling the live prediction markets product

Recommendation: approve a 90-day validation program before scaling Shift's existing client-branded prediction markets platform. Use the program to prove committed client demand, a workable legal and partner model for the first country, reliable prices and correct payouts, and sustainable profit after operating costs. Keep a Shift-operated exchange, Shift-controlled outcomes, borrowed funds, broad US sports or politics, EU retail access, and clients taking the opposite side of customer trades out of scope. Return on day 90 with a scale, narrow, or stop decision.

Why the decision is about scaling, not launching: Shift's public materials show that the product is already live. The responsible next decision is therefore whether the evidence supports scaling it, not whether Shift should begin building it.

Three facts drive this:

Ninety days should answer four questions with evidence: Will clients commit? Is the operating model legal in each target market? Can customers trade at fair prices and receive the correct payout? Will committed revenue cover the cost of running the product around the clock?

ASK Approve validation program
TEAM Product · Engineering · Risk · Operations · Sales · Legal
DURATION 90 days, with reviews on days 30 and 60
READOUT Scale / narrow / stop

Scope boundary: validate the client-facing platform only. Do not approve a Shift-operated exchange, Shift deciding outcomes, broad US sports or politics, EU retail customers, borrowed funds, or clients taking more trading risk.

Five-minute decision path: scan the market signals, compare the options, review the stop rules, and finish with the 90-day ask. The rest provides supporting product, trading-system, operating, and business-case detail.

01

Why now: large-scale demand inside financial apps is demonstrated

13.6B
Robinhood event contracts in Q2 2026
Company-reported record; more than 10× year over year
$156M
Robinhood Q2 event-contract revenue
More than 10× year over year and above Robinhood's $100M crypto revenue in the quarter
$100M+
Coinbase annualized prediction-market revenue
Current quarterly pace expressed over a year, not actual Q2 revenue; revenue and contracts both more than doubled from Q1
200+ exchanges
75+ brokerages
Launched by Shift
Separate company-reported counts

The opportunity is not to create another consumer destination. It is to make Shift the technology and distribution partner that lets existing financial platforms add event contracts safely, quickly, and under their own brands.

02

Why Shift is well positioned

03

The mechanics and where trust can break

A typical event contract asks a yes-or-no question and pays $1 for the selected outcome or $0 otherwise. A YES price of 63¢ roughly signals a 63% market probability and limits the buyer's loss to 63¢. Matching that order with a 37¢ NO order reserves the full $1 payout before trading.

The challenge is making the full process trustworthy. Each contract needs an official source, trading cutoff, correction and cancellation rules, and a dispute process. Access and prices must remain valid, funds must be reserved, and the regulated partner's result must trigger a single, documented payout that matches the client, partner, and Shift records.

CAPABILITIES SHIFT ALREADY HAS NEW TRUST REQUIREMENT Create client selects markets · location rules apply List Trade YES 63¢ + NO 37¢ = full $1 payout reserved YES and NO orders matched Stop trading trading stops at cutoff: no trades after result is known Confirm result partner-defined review period official source · fixed rules · two-person approval Pay pays $1 / $0 reserved funds released event occurs Maximum loss is known before trading: no borrowing, forced sale, or negative customer balance.
Fig. 1: The full process is the product. Trust requires clear rules, an official result, one documented payout, and records that agree across every party.
04

Who should own each part of the service

Shift's public materials indicate that the main components are assembled. The program must now prove the highlighted capabilities across clients, countries, partners, market conditions, and payout failures.

Client trading app shown under client's brand client brand & domain 200+ exchanges · 75+ brokerages SHIFT PLATFORM Order matching existing trading engine Event contracts full payout reserved Accounts · funds · identity checks balances · wallets · onboarding Administration · fees · reporting client controls and records Market selection location and category rules Outcome review and disputes official source · review period · responsibilities · evidence pay $1/$0 Fully funded trades need no borrowing, margin calls, or forced sales. Regulated markets Kalshi · Polymarket · aggregators live prices and trading capacity Market makers supply prices and capacity orders ID checks · funds send orders trades · results prices runs in production today capabilities Shift must prove external partner
Fig. 2: Shift should own the client controls and technical coordination. Regulated partners should remain responsible for market access, trade processing, custody of funds, live prices, and the official result.

The default approach should be clear. Route orders automatically to external regulated markets and tie each displayed market to one exact partner contract, because similar contracts can have different rules. If a client takes the opposite side of customer trades, called B-book internalization, better-informed customers may trade most aggressively as the result becomes predictable. Require separate legal approval, offsetting capacity, sufficient reserves, strict exposure limits, and executive approval.

05

Four strategic choices, one clear fit

Option A · bridge only

Connect to one external market

Send client orders from a branded interface to one regulated partner. This is fast and useful for learning, but makes Shift dependent on that partner's markets, prices, results, reliability, and terms.

Client fitHigh
Time to valueFast
DifferentiationLow
DecisionBridge only
Option B · do not advance

Shift-operated consumer exchange

Shift would attract consumers and own listings, prices, trading capacity, monitoring, trade processing, customer funds, and results. This duplicates regulated partners, starts without customers or capacity, and underuses Shift's distribution advantage.

Client fitMedium
Time to valueSlow
Regulatory burdenVery high
DecisionNo
Option C · recommendation

One platform across regulated partners

Give clients one connection and operating experience across regulated partners. Shift manages contract presentation, location and category rules, routing, customer controls, records, and reporting. Partners handle regulated trading, custody, live prices, processing, and results.

Client fitVery high
Time to valueFast
DifferentiationHigh
DecisionAdvance

Sequencing: start with one regulated market during the validation program, but design the contract, control, and record model so a second partner can be added without rebuilding the product. Option A is the bridge; Option C is the destination.

Fourth option, later: build data and analytics only after Shift has reliable cross-partner information and a strong history of correct outcomes and payouts. It should not distract the first program.

06

Five ways this can fail, and when to stop

Failure modeEarly warningControl and stop rule
Unclear legal responsibility No written agreement assigns responsibility for market creation, customer access, trade processing, customer funds, outcomes, payouts, or country-specific support. Use a regulated market and intermediary; block disallowed countries and categories. Stop: no written approval covers the first client, country, and categories.
Poor pricing or insufficient trading capacity The bid-ask spread, meaning the buy-sell price gap, is too wide; there is too little trading capacity; prices disappear or become stale; or orders are rejected. Use two credible price sources, route externally, block stale prices, and pause unhealthy markets. Stop: price availability falls below 98% or the typical spread exceeds 4¢.
Unclear outcome rules Independent reviewers disagree after applying the contract's official source, cutoff, correction, cancellation, and dispute terms. Tie each market to one partner contract, lock the rule version before trading, save evidence, and require two independent approvers. Stop: an important outcome requires rewriting rules after trading starts.
Insider knowledge or outcome manipulation Connected accounts, event links, unusual profits, or trading by people who may know or influence the result. Restrict people connected to the event, monitor linked accounts, limit positions, and assign an investigator. Stop: any unresolved market-integrity incident.
Payout or trading-risk mismatch Partner trade records, reserved funds, payouts, and Shift records do not match, or the client cannot fund or offset its exposure. Fully fund trades, process each payout once, reconcile records, and own every exception. Stop: any customer-fund shortage or unresolved difference.
07

The business case must clear a revenue floor

Robinhood's Q2 figures imply approximately $156M ÷ 13.6B = 1.15¢ revenue per reported event contract, demonstrating revenue potential inside a financial app, not forecasting Shift's earnings. Shift's base model should combine implementation and recurring platform fees with a small share of revenue per contract. The base case should not rely on profit from taking the opposite side of customer trades or interest on customer balances.

Illustrative path to $2 million in annual recurring revenue

Three clients at a $250,000 annual minimum contribute $750,000; 0.2¢ per contract across 625 million annual contracts adds $1.25 million. This is an illustration, not a forecast. The 90-day program must validate client commitments, partner charges, contract volume, and Shift's revenue share.

Proposed requirements before scaling

These are decision hypotheses, not industry standards. Reset them using partner quotes, price history, and actual client economics.

Primary success measure: profit contribution from each active client after risk and operating costs, not trading volume.

08

The ask: 90 days, four questions

WorkstreamQuestion it answersOutput
Committed demandWhich clients will commit budget and a minimum payment, name commercial and technical owners, and agree to a launch date?15 structured interviews; at least 3 signed client commitments; and one client selected for the closed pilot, with a defined end user, event category, and country.
Right to operateWho is responsible for customer access, market creation, order handling, trade processing, customer funds, official results, payouts, support, and disputes?A written legal operating model; an approval table for each country and event category; signed distribution rights; and terms with the regulated market, intermediary, and price provider.
Fair prices and trustworthy payoutsCan customers trade at fair, available prices and receive correct payouts during normal operations and failures?A test connection; 20 clear markets; a replay of 60 days of real price data; independent review of 30 rule sets; 10,000 simulated trades, including price-feed, partner, and payout failures; and no unexplained record differences.
Healthy economicsWill committed revenue cover partners, data, compliance, support, engineering, and a team accountable around the clock?Signed partner terms; a complete cost model; a closed client pilot; an incident-response simulation; and a scale, narrow, or stop recommendation based on the agreed requirements.

Kill criteria: we stop and say so if:

  • Fewer than 3 clients commit with named commercial and technical owners;
  • No written legal model or signed distribution path covers the first client, country, and categories;
  • Price availability is below 98%, the typical buy-sell gap exceeds 4¢, or partner rejections exceed 0.5%;
  • An outcome requires rewritten rules, customer funds are short, or partner and Shift records do not reconcile;
  • No credible path reaches 65% gross margin, recovers acquisition and implementation costs within 18 months, generates positive client contribution within six months, and funds an accountable 24/7 operating team.

If the early evidence supports continued investment, the first customer pilot remains deliberately narrow:

Included in the first version

  • One closed pilot, designed with the client
  • Economic data releases, central-bank decisions, weather, and broad crypto events with clear official sources
  • 20 selected markets, not a long list of low-demand markets
  • Fully funded trades, external routing, and no borrowed funds
  • One exact partner contract for each displayed market
  • Locked rules, two-person outcome approval, each payout processed once, and continuous record matching

Not included in the first version

  • A Shift-operated exchange, trade processor, or consumer app
  • US sports, politics, subjective outcomes, or EU retail customers
  • Unapproved or user-created markets
  • Borrowing, shared margin across products, or a Shift-created outcome source
  • Treating similar contracts from different partners as interchangeable
Q

Questions to answer before expanding beyond the first pilot

Executive close · What is live, what remains unproven

The launch proves that Shift can assemble the product. It does not yet prove that Shift should scale it.

Public information does not yet show adoption, committed revenue, customer pricing, payout reliability, legal responsibility, support costs, or profit per client. The next 90 days should prioritize evidence, not more features: committed demand, fair prices, official outcomes, correct payouts, records that agree across partners, written legal boundaries, and an accountable team available around the clock.

Decision requested: name one accountable leader across product, engineering, risk, operations, commercial, and legal. Authorize the team to secure at least three client commitments and select one client for the closed pilot; select one regulated market, one intermediary, and one price provider; obtain legal advice for the first country; and return in 90 days with a scale, narrow, or stop recommendation.

Submission notes, per the case study brief

AI tools used
I used Claude Code and OpenAI Codex to support the preparation of this brief. The recommendation, boundaries, thresholds, and final judgment are mine.
Time spent
Approximately one to two hours, including analysis, primary-source verification, strategic review, drafting, and final editing.
Next, with more time
I would interview clients to understand what they need and would pay; confirm written distribution rights and commercial terms with regulated partners; obtain legal advice for each launch country; and test the platform using historical prices and realistic order volumes. I would also independently review the rules for 30 sample markets; simulate incorrect payouts and mismatched records; rehearse how the team would handle a serious incident; and calculate expected profit for the first group of clients.