Is DV360's automated bid strategy having a Monday?

Hey Raj Kumar,

Oh, tell me about it. It's not just you; it feels like DV360's automated bid strategies have been having a whole week of Mondays lately, not just one. I've seen similar erratic behavior with some of our recent campaigns, and it's definitely frustrating when the system seems to be fighting against your goals. Your snippet perfectly illustrates the kind of rollercoaster ride some of these strategies are on right now!

While the "having a Monday" phrase is quite fitting for programmatic, let's look at a few technical reasons why you might be seeing such inconsistent bidding, especially with a Target CPA strategy:

1. Recent Significant Changes: Any major modifications to your line item (budget, targeting, creatives, conversion tracking, or even the CPA target itself) can effectively reset or heavily influence the learning phase of the machine learning algorithms. Even if it's an existing strategy, a big enough shift can throw it off.
2. Data Volatility or Low Conversion Volume: If your line item is getting very few conversions, or if conversion volume is highly inconsistent day-to-day, the algorithm struggles to find stable patterns. Similarly, sudden changes in user behavior (e.g., seasonality, new competitive landscape) can make the data noisy.
3. Unrealistic Target CPA: A Target CPA of $15 might be too aggressive or too conservative for the current market conditions or your targeting.
   • If it's too low, DV360 might underbid significantly to try and hit it, resulting in very low bids and impression volume.
   • If it's too high (relative to actual market value), it might overbid to secure impressions that it thinks will convert, leading to high bids and potentially blowing through budget.
4. Bid Strategy Constraints: Check if you have any minimum or maximum bid caps set within the bid strategy. These can sometimes create "walls" that the algorithm hits, forcing it to behave erratically when trying to stay within bounds while optimizing for CPA.
5. Conversion Lag: If there's a significant delay between an impression/click and a conversion being recorded, the algorithm is optimizing on delayed data, which can lead to misinformed bidding decisions in the short term.

Here’s what I’d recommend for some immediate campaign optimization and troubleshooting:

• Review Change History: Go through the change history for the affected line item and its insertion order. Pinpoint any recent changes that correlate with the onset of the erratic bidding.
• Analyze Conversion Trends: Look at your conversion rate and volume over the last 7-14 days. Are they stable? Are there any significant drops or spikes that might be confusing the algorithm?
• Adjust Target CPA Incrementally: If you suspect your $15 Target CPA is off, try adjusting it by 10-20% up or down, then monitor for 3-5 days. Avoid drastic changes.
• Check Bid Caps: Ensure your min/max bid caps (if any are set) are reasonable and not overly restrictive, preventing the algorithm from finding the optimal bid range.
• Pacing Settings: Verify your pacing is set to "Standard" if you're trying to evenly distribute budget. "Front-loaded" can sometimes lead to more aggressive early-day bidding.
• Audience & Targeting Overlaps: Ensure you don't have significant audience overlaps or conflicting targeting criteria across different line items that might be cannibalizing performance or confusing the learning process.
• Patience and Observation: After any change, give the bid strategy at least 3-7 days to re-learn and stabilize. Continual, rapid adjustments can perpetuate the erratic behavior.

It's all about providing the machine learning algorithms with clear, consistent signals and enough data to learn effectively.

This is really thorough, thanks! I've seen some conflicting advice online about how quickly to adjust the Target CPA when things are going sideways tho — some guides say be more aggressive. Is it always best to go incremental like you suggested, or are there situations where a bigger shift is needed?