What is Statistical Significance & Smart Traffic Routing?
Auto-Stat Sig & Smart Traffic Routing
A hands-off way to wrap up tests and send more visitors to the winner—built into every Fibr campaign.
1. What this feature does
Fibr’s Auto-Stat Sig engine monitors every live experiment continuously from the moment you hit “Start.” As data rolls in, it:
Calculates significance using the model best suited to your test configuration.
Declares a winner once one variant holds the highest probability of outperforming the rest after a minimum of 100 sessions per variant (to avoid low-traffic bias).
Shifts most traffic to that winner while keeping a small “safety drip” on the other variants so you can catch unexpected changes.
Sends you an outcome email summarising the uplift and the new traffic split.
No spreadsheets, no manual re-balancing—Fibr keeps the maths and the routing on autopilot.
2. Why you’ll love it
Refreshing dashboards and guessing whether a lift is “real”
Continuous significance checks that only declare a winner when a variant has the highest probability of success
Forgetting to end a test, letting losers burn budget
Traffic is automatically re-allocated to the winner; lagging variants receive just enough traffic for ongoing validation
Manually re-allocating traffic
Routing flip happens automatically—no more midnight traffic tweaks
Early spikes causing knee-jerk decisions
Continuous monitoring waits for reliable data (≥ 100 sessions/variant) before acting
3. How it works
Live monitoring starts instantly – As soon as your campaign goes live, Fibr begins crunching data in real time.
Smart significance model – The engine smartly chooses the appropriate model for your test (single-variant A/B or multi-variant).
Winner detection – When one variant achieves the highest probability of outperforming the rest (after the 100-session threshold), Fibr crowns it the winner.
Traffic flip – The winner receives the majority of incoming visitors; non-winners share a single-digit percentage so you can spot regressions.
Outcome email – A concise email tells you which variant won, the uplift, and the new traffic split.
4. Using the feature
1. Launch a campaign
Create your variants, choose a primary goal, click Start Test.
2. Let Fibr run the numbers
The engine tracks conversions continuously—no action required.
3. Review the email
When a winner emerges you’ll get a summary with uplift and new traffic allocation.
4. Keep optimising
The test stays on autopilot; fire up your next hypothesis whenever you’re ready.
5. Best practices
Give tests breathing room – Avoid pausing too soon; the engine needs steady data.
Focus on one conversion goal per campaign – Mixed goals dilute confidence.
Trust the safety drip – Leaving a few percent on other variants guards against seasonality shifts.
Archive finished campaigns – Keeps reporting snappy and dashboards clutter-free.
6. FAQ
Q: What if I pause a campaign? Pausing stops data collection. When you resume, significance calculations pick up exactly where they left off—unless you edit a variant, in which case the engine resets.
Q: We drove 10k sessions in one day—does the engine over-sample? No. Fibr’s stats layer scales for high-volume traffic and applies the same probability logic without capping sample size.
Q: When does calculation begin, and how do you handle low-traffic bias? Calculations start immediately after the campaign goes live, but the engine waits until each variant has logged at least 100 sessions before it’s eligible to win. This avoids false positives from tiny samples.
You focus on new ideas; Fibr’s AI handles the heavy lifting.
Last updated