Next Steps Based on Campaign Results

Once your campaign reaches a clear conclusion—🟢 Positive Impact, 🔴 Negative Impact, or 🟡 Inconclusive—it's time to decide on your next strategy based on the data.
This guide provides recommended next steps to maximize Monetai's effectiveness for each result.

---

1. 🟢 Positive Impact

When your campaign's positive impact has been statistically proven, the next step is to expose the promotion to more users to maximize its impact.

Recommended Strategy: Gradual Traffic Expansion
Even with positive results, it's more stable to expand traffic gradually rather than increasing it to 100% all at once. This is to finally verify that the performance remains consistent with a larger user group and to minimize any unexpected variables.

Expansion Example

• Step 1 (Current): 10% of total traffic (Baseline 5% / Monetai 5%)
• Step 2: 20% of total traffic (Baseline 10% / Monetai 10%)
• Step 3: 50% of total traffic (Baseline 25% / Monetai 25%)
• Final Step: 100% of total traffic

Want to expand your campaign? The Monetai team is here to help.
Contact us through your preferred channel below, and we will quickly change the settings to your desired traffic.

• Contact: Monetai Slack Channel or support@monetai.io

2. 🔴 Negative Impact or 🟡 Inconclusive

If your campaign is Inconclusive or has a negative impact, it's an important signal that your Monetai strategy or SDK integration needs improvement.
Follow the steps below to diagnose the cause and try to improve the results.

2-1. Advancing your predict() calls

• Step 1: Check the number of predict() calls

  • Problem Diagnosis: If the AI model has too few opportunities to predict a user as a 'non-paying user', the chance for the promotion to be shown also decreases, and the campaign's effect may not appear properly. Are you currently calling predict() only 1-2 times per session?
  • Solution: Increase the prediction and promotion exposure opportunities by adding the predict() function to various screens where a user might contemplate a purchase (e.g., key detail pages, upon re-entering the app). We recommend adding it to as many relevant points as possible.

• Step 2: Change the timing of predict() calls

  • Problem Diagnosis: Calling predict() too early (e.g., right after app launch) or too late (e.g., right before abandoning checkout) can be ineffective.
  • Solution: It is most effective to call predict() at the 'critical moment' when a user is hesitating to purchase, after they have sufficiently recognized the product's value. Try changing the call timing and running the campaign again.

Still no improvement after trying the methods above?

Optimizing the count and timing of the predict() function is important, but the AI model's performance ultimately depends most on the quality of the data it learns from.
If you've tried the methods above and performance has not improved, we recommend proceeding to the next step, 2. Advancing User Behavior Events to provide the model with richer data.

2-2. Advancing User Behavior Events

• Step 1: Segment user behavior events

  • Problem Diagnosis: The events you are currently logging might be too general (e.g., app_open) or have little direct connection to the purchase decision. For the AI model to more precisely grasp the user's 'purchase intent,' more specific behavioral data is needed.
  • Solution: You can provide richer context to the AI model by logging more granular "micro-events." In particular, it is good to log the user's detailed behavior up until the moment right before the predict() function is called.
    • Example: Instead of just logging a single 'click_bestseller_tab' event, log several segmented events like 'enter_product_list' → 'click_product' → 'scroll_product_details'.

• Step 2: Request a model retrain with the improved data

  • Problem Diagnosis: If you have added or changed user events, it is essential to have a process to reflect the improved data in the AI model.
  • Solution: After adding new events, you must ensure the AI learns the latest data by rebuilding the model.
    • Please contact our support channel or email (support@monetai.io) to request a model rebuild.