Last month, a client reached out. They ran a weekend top-up campaign where Group A’s click-through rate was 40% higher than Group B’s, and they were ready to scrap Group B entirely. We pulled the logs. Group A used a short link; Group B used a full URL. Half of Group A’s short links timed out after multiple redirects. The clicks went out, but Group A’s registration completion rate was actually 15% lower than Group B’s. If they had only looked at clicks, they would have permanently killed the version that actually converted.
Testing Starts with a Decision, Not a Variation
Which version should the team use for the remaining audience or the next similar campaign: A or B? That decision determines what to test. If the goal is registration, registration completion matters more than clicks. If the goal is win-back, return events are the focus. Sender ID and opening lines are worth testing, but only if both versions are approved and registered for the market. Results may reflect trust, registration, or filtering—not just the copy itself.
Third-Party Benchmarks: CTR Ranges Are Wide
Data compiled by EZ Texting shows that SMS marketing click-through rates typically range from 21% to 35%. Omnisend’s analysis of 246 million campaigns in 2025 gives an average CTR of 12.39%, with CTR climbing from 18.02% in October to 23.92% in December due to BFCM and year-end promotions. These numbers point to one thing: the “normal range” for CTR fluctuates significantly, and using a global average as a KPI will mislead decision-making. Your baseline depends on industry, audience, timing, and message type.
How to Control Variables
“A shorter, event-focused opening line will increase completed registrations among players who viewed the event page but did not join.” That sentence contains the audience, the variable, and the outcome. Changing sender ID, offer, timing, and CTA all at once might improve results, but it won’t tell you which change caused it. Split the audience before launch; removing underperforming versions based on early results biases the comparison. Guard against concurrent campaigns too: if one group receives another promotion during the window, the team won’t know which touchpoint influenced the result.
Encoding and Character Sets Also Affect Results
Thai messages using UCS-2 encoding are limited to 70 characters per segment, and Vietnamese Unicode encoding also reduces per-message capacity. Keep encoding comparable during testing, or document these differences as part of the hypothesis rather than attributing results solely to wording. Compliance is also unavoidable. Thailand’s NBTC further tightened international A2P SMS rules in October 2025 and implemented KYC registration, while Vietnam’s Decree 91/2020/ND-CP strictly controls spam messages. Short links, multi-step redirects, and domain-brand mismatches increase filtering risk.
Which Metrics Decide the Winner
The primary metric drives the decision, while supporting metrics reveal whether it caused harmful trade-offs. Show results by country and player segment. Global averages hide local problems. International SMS is affected by local carrier filtering, time zones, and sender registration. A winning version is only a market-specific learning.
| Metric | What It Helps Answer | Why It Can’t Stand Alone |
| Delivery Rate | Did the message reach the route? | It says nothing about relevance |
| Click-Through Rate | Did the call to action generate interest? | The destination may still fail |
| Registration Completion Rate | Did the message support the intended action? | It may not show long-term value |
| Deposit or Return Event | Did the player continue the journey? | Other channels may influence it |
| Opt-Out Rate | Did the message create unwanted pressure? | A low rate doesn’t prove high satisfaction |
Technical status and business success must be viewed separately. Alerts should focus on patterns: spikes in country-level failures, sender-specific filtering, or Webhook error bursts. One failure is normal; a sudden wave is not.
Frequently Asked Questions (FAQ)
Q1: In A/B testing, the click-through rate is high but registration completion is low. What’s wrong?
Common causes include slow landing pages, login friction, a mismatch between the message promise and the destination, or short-link redirects timing out. Evaluate the CTA across the full journey.
Q2: Why should an A/B test change only one primary variable at a time?
Changing sender ID, offer, timing, and CTA simultaneously might improve results, but it won’t tell you which change caused it. A useful hypothesis includes the audience, variable, and expected outcome.
Q3: Third-party data shows a wide CTR range. Which one should I use as a target?
EZ Texting gives a range of 21% to 35%, while Omnisend reports an average of 12.39%. The difference comes from sample, industry, and season. Comparing against your own historical baseline is more reliable than applying any external average.
Q4: Can promotional SMS be retried automatically after a failure?
Be cautious. Retries for authentication or critical notifications can improve the experience, but auto-retrying promotional SMS can increase fatigue. Retry strategy should consider message type, user action, and opt-out status.
Data Sources
Correction Contact: If you find factual errors or improper data citations in this article, please contact service@laaffic.com or +63 9637902308.
