Split Testing

Split tests (A/B tests) let you distribute traffic across multiple paths, landers, or offers to find the best performer. Voluum supports splits at three levels — you can stack them in the same campaign.

Three Levels of Splits

Level	What You're Testing	Where to Set It
Path-level	Different funnel flows (e.g., direct-to-SP vs. quiz-first)	Default paths section
Lander-level	Lander variants inside one path (e.g., JPG vs WEBP optimisations)	Landers block on the right when a path is selected
Offer-level	Different offers/variants inside one path	Offers block on the right when a path is selected

---

Real Example — APPlovin - United States - Quiz US - Ledisa GLP-1

Here's the full traffic breakdown of one live campaign — two active paths, each with its own internal lander split:

CAMPAIGN TRAFFIC (100%)
│
├── PATH 1: Standard Sales Page — weight 60
│   │   (60% of total traffic)
│   │
│   ├── Default                     weight 40 →  40% of path  →  24% total
│   ├── Optimized first image JPG   weight 10 →  10% of path  →   6% total
│   └── Optimized first image WEBP  weight 10 →  10% of path  →   6% total
│
└── PATH 2: Email Collect Sales Page — weight 40
    │   (40% of total traffic)
    │
    ├── Email Collect (before transformation)   weight 30 →  75% of path  →  30% total
    └── Email Collect v2 (before profile res.)  weight 10 →  25% of path  →  10% total

A third path — Balance — exists in the campaign config but is disabled via toggle, so it receives no traffic right now.

Path-level split (Default paths)

Path	Weight	Share	Status
GLP-1 (Standard Sales Page)	60	60%	Active
GLP-1 Email Step Collect (Email Collect Sales Page)	40	40%	Active
Balance	10	—	Disabled (toggle off)

Lander-level split — Path 1 (Standard Sales Page)

Lander	Weight	Share of path
Global - Quiz 18p - Ledisa GLP-1 (default)	40	66.67%
Global - Quiz 18p - OPT JPG - Ledisa GLP-1	10	16.67%
Global - Quiz 18p - OPT WEBP - Ledisa GLP-1	10	16.67%

Main quiz gets 2/3 of this path's traffic; 1/6 each to image-format variants (JPG vs WEBP) to test load / conversion impact.

Lander-level split — Path 2 (Email Collect Sales Page)

Lander	Weight	Share of path
Global - Quiz 18p Email Collect - Ledisa GLP-1	30	75%
Global - Quiz 18p Email Collect (before profile results) - Ledisa GLP-1	10	25%

Two variants of the email-collect quiz — one shows the email step after the profile, the other before — tested 75/25 to see which placement collects more emails without killing conversion.

Offer-level split

Both paths route to a single main offer (one Ledisa GLP-1 sales page variant per funnel), so there's effectively no offer-level split here — but you can add one using the same approach.

Weights don't need to sum to 100

Voluum uses weights as ratios, not percentages. 40/10/10 → 66.67% / 16.67% / 16.67%. 1/1/1 gives an even split. Weights are normalised inside each level — path-level weights split total traffic, then each path's landers/offers split that path's share.

---

Setting Up Splits

Step 1 — Open Campaign Destination

Open the campaign → Destination tab.

Step 2 — Path-Level Split (Default paths)

1. In the Default paths section, click "Add default path" for each funnel variant you want to test
2. Name each path clearly (e.g., GLP-1, GLP-1 Email Step Collect, Balance)
3. Set the weight on each row
4. Use the row toggle (green = active, grey = paused) to pause a variant without deleting it

Step 3 — Lander-Level Split (inside a path)

1. Click on a path on the left — the right panel shows its settings
2. In the Landers block, click "Add lander" for each variant
3. Set weight next to each lander
4. For A/B: two landers at equal weight (e.g., 50/50 or 1/1)
5. For multi-variant: weight to taste (e.g., 40/10/10 to give main lander the majority)

Step 4 — Offer-Level Split (inside a path)

Same process as landers, but in the Offers block:

1. Click "Add offer" for each offer you want to test
2. Set weight — typically one main offer (100%) with backup offers at weight 0 until you activate them

---

Traffic Distribution AI

Each path has a Traffic distribution AI section with Weight optimization and Smart rotation toggles.

We never use Weight Optimization

Leave both toggles off — always. At Acentecom we run manual weights on every campaign.

Weight optimization auto-shifts traffic to whatever variant is winning in real time, which sounds good but in practice:

• Locks in a false winner after a handful of early conversions (statistical noise)
• Starves the other variants of traffic before you have enough data to decide
• Makes split tests un-reproducible and reporting unreliable

Stick with manual weights. When you decide a winner, you move the weights. That's the workflow here — don't enable AI optimization.

---

Split Testing Tips

Best Practices

• Wait for statistical significance — target 100+ conversions per variation before picking a winner
• One variable at a time — only change one thing between variants (headline, image, CTA)
• Don't stop tests too early — day-of-week and time-of-day patterns skew short tests
• Track both CTR and CR — a higher CTR lander isn't always the winner (look at EPV / total revenue)
• Keep losers at weight 0 instead of deleting — easier to re-enable for follow-up tests