🧪 Pricing Experiments

The Pricing Lab lets you test different rates and track whether raising (or lowering) your price actually improves your income — not just your revenue per job.

Pricing Lab in action

A/B pricing experiment showing control vs variant revenue and statistical significance

Animated demo — GigAnalytics tracks sessions per rate variant and shows significance as sessions accumulate.

What is a Pricing Experiment?

A pricing experiment records a period during which you charged a specific rate (Variant A) and compares it against another period or rate (Variant B). GigAnalytics tracks:

Revenue per hour — the primary metric
Transaction count — volume / conversion proxy
Total net revenue — absolute income
Average transaction value — per-job revenue

How to Run a Pricing Experiment

Go to Pricing Lab in the sidebar.
Click New Experiment. Give it a name (e.g., "Raise Upwork rate from $85 → $110").
Select the income stream you're testing, a start date for Variant A, and optionally a switch date to Variant B.
Import your transactions as usual. GigAnalytics will automatically bin transactions into Variant A or B based on their date.
After at least 2 weeks of data per variant, open the experiment to see results.

What Is Measured

Metric	Formula	Why it matters
Revenue/hour	net_revenue / billable_hours	Primary. Accounts for volume changes.
Tx count	COUNT(transactions)	Did volume drop when you raised prices?
Avg tx value	net_revenue / tx_count	Higher rate but fewer jobs — is each job worth more?
Total net revenue	SUM(net_amount)	Did total income go up or down?
Conversion proxy	tx_count / days_in_period	Jobs per day — crude but useful for active platforms.

Statistical Significance — What We Show

GigAnalytics does not run formal significance tests (t-tests, p-values) on your pricing data. Here's why:

Freelance income is highly variable and non-normal (a few large jobs dominate)
Most users have < 50 transactions per variant, making p-values misleading
External factors (seasonality, platform algorithm changes) confound results

Instead, we show a practical significance estimate:

practical_significance = |variant_b_metric - variant_a_metric| / variant_a_metric * 100
-- "meaningful" threshold: ≥ 15% change with ≥ 5 transactions per variant

A result flagged as "meaningful" means the difference is large enough to likely matter in practice, not just noise. Always combine this with your own judgment about market conditions.

Pricing Lab Bucket Bars

The Pricing Lab also shows a revenue-per-hour histogram across your transaction history, bucketed by price range. This answers: "At what price point do I earn the most per hour?"

Buckets are auto-sized to your rate range (e.g., $0–$50, $50–$100, $100+)
Bar height = revenue per hour for all transactions in that price bucket
Sweet spot bucket = highest revenue/hour with ≥ 3 transactions
Bars are highlighted when bucket is the sweet spot

Tips for better experiments

Run each variant for at least 2–4 weeks before drawing conclusions
Test one variable at a time (rate only — don't also change your service scope)
Compare the same platform against itself (don't mix Upwork vs. direct clients)
Log your time during the experiment period — without time data, revenue/hour can't be computed
Use the "Notes" field to capture why you ran the experiment