Kwai Ad Launch Troubleshooting Guide: Post-Launch Optimization & Issue Resolution

After your campaign goes live, maintaining stable performance and addressing unexpected issues becomes crucial. This comprehensive troubleshooting manual provides practical guidance to help you identify, analyze, and resolve common problems that may arise after launch, ensuring your ads continue to deliver strong and consistent results.

1. Platform Traffic Bidding Logic

Question: What is the bidding and ranking logic for information flow ads?

Answer: Regardless of the bidding method used, the system ultimately ranks ads based on eCPM (effective Cost Per Mille). Ads with higher eCPM values are prioritized and displayed first.

Bidding Logic Formulas:

2. Recommended Account Structure for Initial Launch

Guidelines:

• For deep events (e.g., add-on purchases, paid conversions):
  • Avoid creating too many ad groups to prevent scattered conversions.
  • Focus on 1–2 deep event groups to maintain conversion concentration and help develop stable, high-performing groups.
• For shallow targets (e.g., clicks, views):
  • A 1:1 structure (one campaign, one ad group) is recommended, as it is easier to scale and gain volume.
• Creative Updates::
  • Update no more than 10 sets of creatives per day to ensure new materials receive cold-start exposure support.
  • If you have a large number of creatives, distribute updates evenly across days to maintain consistent performance.

3. Cold Start / Learning Period

How to Launch During the Cold Start or Learning Period

Background

When launching new accounts, campaigns, or creatives, the system needs time to identify the right conversion audience and align the estimated click-through rate (CTR) and estimated conversion rate (CVR) for each unit. During this stage, the main goal is to accumulate conversion data (clicks, behaviors, and conversions) to help the system’s machine learning model become more accurate.

Principle

The system initially assigns new units an estimated performance based on market averages and multi-dimensional data, including both market benchmarks and historical account performance. Currently, the estimated value is equivalent to the market average.

The three most critical factors in bidding are:

• Bid amount
• Estimated CTR
• Estimated CVR

If the estimated CTR and CVR are low or average, it will be difficult to outperform competitors unless the bid is sufficiently high. As a result, the campaign may not achieve the desired exposure.

Optimization Recommendations

• Targeting: Use broad targeting during the learning phase.
• Bidding:
  • Set bids 30%–50% higher than your regular daily bid during the learning period.
  • Adjust bids once every two days.
  • After the learning period, gradually reduce bids to your target (psychological) price.
  • Each adjustment should not exceed 10%.
• Creatives: Launch with at least 6 creatives to provide sufficient data for optimization.

CPA Cost Verification Cycle

• If a single ad group generates fewer than 10 conversions per day, extend the observation period before evaluating CPA performance.
• If a single ad group generates 10 or more conversions per day, monitor CPA daily for optimization.

Why Is There No Volume After the Cold Start?

Two main factors can cause a drop in traffic after the cold start phase:

• CTR/CVR Performance Below Market Average
  • When a new campaign is launched without prior delivery data, the system estimates CTR and CVR and bids based on those estimates.
  • Once the campaign accumulates around 20 conversions, the system builds a model using the actual CTR and CVR.
  • If the actual performance is below market average, the campaign’s exposure and traffic volume will decrease.
• Excessive Bid Reductions
  • Each time the advertiser adjusts the bid, the system recalibrates based on the new value.
  • If the bid adjustment is too large or too frequent, it can cause model instability, leading to reduced traffic and inconsistent delivery.

4. No Volume

What should I do if a new campaign or ad group fails to gain volume?

Common Phenomena:

• Phenomenon 1: The bid for the ad campaign is sky-high, but the traffic volume remains low.
• Phenomenon 2: The click-through rate (CTR) and conversion rate (CVR) of the creatives are higher than the market average, yet the exposure volume is only a few dozen impressions.

Data Discrepancy Explanation

When exposure is extremely limited, the system’s background data may show artificially high CTR and CVR. However, this is surface-level data and not statistically reliable. At this stage, the system has not yet entered the model optimization phase, so these values are not factored into machine learning adjustments.

Label Targeting Issues

• Many campaigns use overly strict targeting settings, such as age, gender, or device model.
• Currently, targeting precision has a significant impact on delivery volume.
• Labels like age and device model can filter out more than 20% of potential users in the market.
• The targeted audience includes only users who have explicitly set their age or gender in the app.
• Overseas apps differ from domestic ones — users often have fewer profile settings and many do not log in at all.
• As a result, when these labels are applied, the system automatically excludes users without defined attributes, leading to a substantial reduction in available audience size and, consequently, a significant drop in delivery volume.

5. Expansion: How to Increase or Scale Up Volume

i. Broaden Targeting

Principle:

Keep only essential targeting parameters open. You only need to set the target country. For social networking campaigns, you may additionally specify gender and age if necessary.

Examples:

• Example 1: An e-commerce advertiser initially restricted targeting by operating system language. During the expansion phase, it is recommended to remove the OS language restriction and set it to unlimited, which resulted in a 30% increase in daily spend.
• Example 2: A customer initially limited delivery to specific time periods. During the expansion phase, switching to full-day delivery led to a 120% increase in spend the following day.

ii. Allocate Budget Strategically

• Increase the budget for ad groups or campaigns that are already generating stable volume.
• Maintain the new budget level for 1–2 days to allow the system to stabilize.
• Avoid frequent bid or budget adjustments, as they can disrupt the learning process and reduce delivery efficiency.
• Reasonably allocate the budget under the account
• Increase the running volume unit budget, stabilize for 1~2 days, and avoid frequent price adjustments.

6. Bidding / Price Adjustment: How should I bid/price adjustment?

• You can refer to the industry benchmarks synchronized monthly by the sales/CS manager to make bids
• If it is a new account/new ad group, it is recommended to increase the daily bid by 30%-50% during the cold start period
• It is recommended that the price adjustment should not exceed 10% each time, and it is advisable to adjust it at most once a day.

7. Creativity

i. What Is the Appropriate Frequency for Updating Creatives?

• Update creatives 2–3 times per week.
• When iterating, refer to high-performing creatives within the same account and make improvements in visual performance, content, or messaging.
• If an account already has 5–6 well-performing creatives, but new creatives with good data fail to scale further, consider:
  • Opening a new account
  • Testing the potential creatives in the new account
  • Setting a slightly higher bid to help them gain traction

ii. How to Update Creatives for Better Volume

• Add new creatives to mature ad groups so they can benefit from the existing ad model and cold-start traffic support.
• If there are no ad groups with strong historical performance, create a new ad group for delivery.
• This prevents new creatives from being affected by historical low-quality data and allows them to receive fresh cold-start exposure.

iii. What Is the Creative Decay Cycle and How to Identify It?

• The platform’s creative decay cycle is typically 7–10 days.
• Ensure that creatives are updated at least once per week to maintain performance.
• Creative Decay Self-Check: If a creative’s CTR or CVR drops by 10%–20% compared with its historical average, it is considered decayed and should be optimized or replaced promptly.

8. Others FAQ