Articles on: Upsells & Revenue

A/B Testing & Upsell Analytics

Purpose

Answer: “How do I measure and improve upsell performance?”

⭐ Why This Matters

Without measurement:

You guess instead of optimize
Discounts erode margin silently
High-performing surfaces go unnoticed

A/B testing and analytics turn upsells into a predictable revenue channel.

⚙️ How A/B Testing Works

Each test splits traffic into:

Control (A) → original offer
Variant (B) → modified version

A/B Testing dashboard overview

You control the split and the bar for declaring a winner.

Traffic split — how visitors are divided between Control (A) and Variant (B). Default 50/50, but you can change it; it does not have to be even (e.g. 70/30 to limit exposure to a risky variant).
Confidence level — how statistically sure Account Editor must be before naming a winner. Default 95%. Higher = more certain but slower; lower = faster but riskier.
Locked per session — once a visitor is bucketed into A or B, they stay there for a consistent experience.

📌 A 50/50 split learns fastest, but an uneven split lets you protect revenue while still testing.

🟥 MEDIA NEEDED → IMAGE: A/B test setup showing the adjustable traffic-split control and the confidence-level field (default 95%) — add before publishing

🎯 Success Metric & Countdown

Each test optimizes for one metric and shows time remaining.

Success metric — pick one: Conversion rate, Revenue, AOV, or Click-through rate. This is the number used to decide the winner.
One test per surface at a time, and the surface needs at least two active offers.
Days-remaining countdown — the test card shows how long until the test is scheduled to conclude, so you know when results are due.
Apply winner — sends the winner live and the loser to draft, once your chosen metric clears the confidence level.

📌 Don't end a test early just because one variant looks ahead — wait for the confidence level to be met (and ideally ~14 days of data) so the result isn't noise.

📊 Metrics Tracked

Views
Conversions
Revenue
Click-through rate
Average Order Value (AOV)

Account Editor automatically calculates winners once significance is reached.

🧪 Creating an A/B Test

Step 1: Hypothesis

Define one change only.

Example:

“Increasing discount from 10% to 15% improves conversion.”

Step 2: Metric Selection

Choose one:

Conversion rate
Revenue
AOV
Click-through rate

Test metric selector

Step 3: Surfaces

Tests run per surface.

Rules:

One test per surface at a time
Requires at least two active offers

Step 4: Duration

📈 Reading Test Results

You’ll see:

Control vs Variant
Winner badge
Projected gain

Test result card with winner

🏆 Applying the Winner

When ready:

Click Apply
Winner goes live
Loser moves to draft

Analytics automatically update.

📊 Analytics Explained

Key Views

Total upsell revenue
Conversion funnel
Surface performance
Geography & device split

📊 Per-Offer "Offer Statistics" & the Conversion Funnel

Open any offer's Offer Statistics dialog to see how it performs on its own.

The Offer Statistics dialog reports:

Revenue
Impressions
Conversions
Conversion rate

It also shows a 4-stage conversion funnel:

Impressions → Clicks → Add to cart → Conversions

Reading the funnel tells you where shoppers drop off: lots of impressions but few clicks means the headline/offer isn't compelling; clicks but few add-to-carts means the product or price needs work.

🟥 MEDIA NEEDED → IMAGE: Per-offer "Offer Statistics" dialog with Revenue/Impressions/Conversions/Conversion rate and the Impressions → Clicks → Add to cart → Conversions funnel — add before publishing

📈 Dashboard Breakdowns

Beyond a single offer, the Analytics dashboard breaks performance down by:

Surface performance — which surface converts best
Geographic performance — performance by country/region
Device performance — desktop vs mobile
Customer segmentation — how segments respond

📌 Use these to decide where to invest: e.g. if Order status outperforms Thank-you for repeat buyers, shift effort there.

🧪 Real Merchant Scenarios

Scenario A — Discount test wins, AOV drops

Interpretation:

More conversions, lower margin

Action:

Test smaller discount next

Scenario B — Checkout surface underperforms