How we afford
to pay your customers' cashback.

The short answer

We collect subscription fees from your customers ($10-$40/month depending on tier). We also sell anonymized, aggregated spending pattern data to institutional research buyers—satellite imagery companies, quantitative hedge funds, real estate investment trusts, and urban planning firms.

No. And we mean that — this platform was built specifically because we were sick of services that prey on small businesses with fees buried in the fine print. No email fees. No hosting fees. No "premium support" fees. No per-transaction processing fees. No monthly minimums. No setup fees. No cancellation fees. No annual renewal fees. No feature unlock fees.

That's it. But to understand why anyone would pay for that—and pay well—you need to understand a $7.2 billion industry hiding in plain sight.

$7.2BAlternative data market (2024)

Grand View Research¹

29.3%CAGR through 2030

Grand View Research¹

78%Hedge funds using alt data

Greenwich Associates²

$1.1MAvg. annual spend per fund

Alternativedata.org³

The parking lot trade

In 2018, a hedge fund paid a satellite company to photograph Walmart parking lots across America. Every day. For months.

By counting cars in each image, they built a model that predicted Walmart's quarterly revenue—weeks before Walmart reported it. When the earnings came out, they were ready. The trade reportedly generated returns exceeding $100 million.

This wasn't illegal. It wasn't insider trading. It was just better data.

Alternative Data Market Growth

Global market size in billions USD, 2019-2030 projected

Why satellite companies need transaction data

Here's the thing about satellite imagery: it tells you where people are, but not what they're doing.

A satellite can see that a parking lot is full. But it can't tell you if those customers spent $15 or $150. It can't tell you if they're buying groceries or returning items. It can't distinguish between a lunch rush and people waiting for the DMV next door.

Satellite imagery companies need "ground truth" data to validate their models. Without knowing actual transaction volumes at specific locations, their parking lot counts are just… parking lot counts. The value is in correlating physical presence with economic activity.

That's where we come in. And that's where our name comes from: OrbitBack. The data that flows back to the satellites to make their imagery actually useful.

Satellite Data Accuracy: With vs. Without Ground Truth

Revenue prediction accuracy for retail locations

Source: Internal validation study across 847 retail locations, Q3 2024

Why our data is different

You might ask: why not just buy transaction data from payment processors or POS systems?

They do. And it's deeply flawed.

The key difference is selection bias. When you buy POS data, you're only seeing businesses that use that specific system. It's like trying to understand national voting patterns by only polling people who shop at Whole Foods.

Our data comes from customer behavior—wherever they spend, whatever payment method they use. The customer links their bank account once via Plaid, and we see their transactions across all merchants they've opted into.

Data Coverage Comparison: POS vs. Consumer-Side

Percentage of total retail transactions captured by data source type

*Consumer panel captures transactions at ANY merchant where panelist shops, regardless of POS system used

Source: Company estimates based on POS market share data (Nilson Report, 2024)⁵

The Plaid advantage

When a customer signs up, they connect their checking account or credit card through Plaid—the same secure bank connection used by Venmo, Robinhood, Coinbase, and over 8,000 other financial apps.

This single connection gives us something no POS system can provide: a complete financial fingerprint.

Income ranges. Spending categories. Rent vs. mortgage. Subscription services. Savings patterns. Credit utilization. Geographic mobility. Brand affinity across dozens of merchants. All without asking a single survey question—because it's derived from actual financial behavior.

Source: Internal validation study comparing Plaid-derived attributes to self-reported survey data across 12,400 matched users, Q2 2024

The mathematics of signal

For the quantitatively minded: here's why uniform customer sampling produces better predictive signals than merchant-side sampling.

POS data carries three uncorrectable bias terms: which merchants adopted the system (merchant selection), where those merchants are located (regional clustering), and what size businesses can afford enterprise POS (size bias).

Our approach carries one bias term—demographic skew toward cashback-seeking consumers—which is measurable and correctable because we know the exact financial profile of our user base from Plaid.

In plain terms: we know exactly how our sample differs from the population, so we can adjust for it. POS aggregators don't know what they're missing.

Signal Degradation by Data Source

Percentage of true market signal retained after bias effects

How the money flows

For the quantitatively minded: here's why uniform customer sampling produces better predictive signals than merchant-side sampling.

Revenue Allocation Model

How each dollar flows through OrbitBack

The unit economics work like this:

Customer subscriptions ($10-40/mo) and merchant platform fees ($5-20 per transacting customer) fund our operating costs: bankgrade AES-256 encryption, SOC 2 Type II compliance, PCI DSS Level 1 certification, Plaid integration, real-time fraud monitoring with <0.01% false positive rate, 99.99% uptime SLAs, multi-region redundant data centers, and the security engineering team to maintain it all.

Revenue from anonymized data sales to institutional buyers funds customer cashback rewards directly. Current data licensing revenue covers approximately 112% of cashback obligations, creating a sustainable funding surplus.

More customers = richer data = more data revenue = higher cashback = more customers. Geographic density compounds value exponentially—our internal models show a city with 10,000 OrbitBack users produces 8.3x more valuable insights per user than 10 cities with 1,000 each.

Data Buyer Categories

Revenue share by institutional buyer type, FY 2024

Source: OrbitBack internal data, fiscal year ending Q4 2024

What we will never sell

To be absolutely clear about what our data buyers receive and don't receive:

Names, emails, phone numbers, addresses, bank account numbers, or any personally identifiable information. Our contracts prohibit re-identification attempts with $10M+ liquidated damages clauses, and our data format makes it technically infeasible. We aggregate to zip code level minimum (avg. 30,000 residents), and no individual's transaction history is ever visible to buyers.

Why this matters for your business

You might be thinking: "I don't care about hedge funds. I just want a loyalty program."

Fair. Here's why this matters to you:

Because we make money on data, we can offer you something no one else can: a loyalty program where you pay nothing upfront, nothing monthly, and nothing for the cashback your customers earn. We only charge you when customers actually spend at your business—and even then, it's a fraction of what you'd pay for traditional customer acquisition.

Customer Acquisition Cost Comparison

Cost per acquired/retained customer by channel

Source: Industry benchmarks (WordStream, 2024⁷) and OrbitBack internal data

The satellite companies and hedge funds subsidize your loyalty program. That's the deal.

The privacy architecture

We take one thing seriously above all else: the trust your customers place in us when they link their bank account.

Our privacy architecture uses differential privacy techniques—the same mathematical framework Apple uses for iOS analytics and the U.S. Census Bureau uses for population data. Data is anonymized at the point of collection, not after.

ε = 0.1Differential privacy parameter

Lower = more private

72hrsMax deletion propagation

Usually < 24 hrs

30K+Min. aggregation size

Residents per zip code

0Re-identification incidents

Since founding

Customers can delete their data at any time. When they do, it's gone —from our servers and from any future data products. We maintain a real-time deletion pipeline that propagates removal requests within 72 hours (typically under 24). We publish an annual transparency report detailing exactly what data categories we've licensed, to what types of buyers, and in what format.

Still have questions?

We know this is a lot. The intersection of loyalty programs, satellite imagery, and quantitative finance isn't exactly dinner table conversation.

If you want to go deeper, we've published a technical whitepaper explaining our anonymization methodology, our data buyer vetting process, and the mathematical proofs behind our privacy guarantees.

Or if you just want to launch a loyalty program and let us worry about the rest—that works too.