In this article, we outline why increased targeting and throttling may actually decrease a campaign’s performance and we offer tips on what you can do to optimize your campaign instead.

First, let’s define what ‘performance’ means. In programmatic advertising, the goal is to win bid requests at the most favorable cost that will produce the most favorable results. Simplified, it means that Moloco Cloud DSP will give you the best return on ad spend (ROAS).

Our shared goal is the highest possible value for every dollar spent.

For the sake of our discussion here, let’s say that the desired result, or value, is user acquisition (UA). Moloco Cloud DSP, through its sophisticated machine learning engine trained on your first-party data, will respond to bid requests that will have the highest probability value (your UA KPI event) at the lowest price.

In a hypothetical scenario, let's assume your campaign is running and your goal is a ROAS of 10% by day “x”. For every dollar you spend, you receive .10 dollars in value by day “x”. Moloco Cloud DSP responds to bids that have a high probability of achieving or exceeding your KPI at a favorable price within your budget.  

When you are using targeting in your ad campaign, you may be tempted to tighten the targeting even further. The thinking here is that by only focusing on very strict targeting variables, the resulting performance (generally measured by return on ad spend, or ROAS) will be increased. However, this may not always be the case - in fact, with an effective model, manual targeting is likely to damage the performance. Here’s why.

Visually, Figure 1 represents a campaign where there is no manual targeting. The best bid requests from the available options, represented by the blue, orange, and green dots, will be bought. You can see that these are low cost but high value bids within the solid red lines.

Let’s say that targeting is applied and we no longer buy blue dots, only the orange and green dots are bought. In order to spend the same budget, a wider net has to be cast. That means we have to buy from more expensive inventories as well as less valuable inventories, represented in Figure 2 below. Subsequently, the ad campaign’s return on ad spend performance will be reduced.

Figure 1. No manual targeting                          Figure 2. Tightened targeting

An all too common example is where advertisers tighten the targeting in an attempt to achieve a lower CPI by excluding (throttling) certain publishers. Let’s see how that can play out in ad campaign performance with machine learning.

Throttling publishers example

If the goal is to achieve 10% ROAS, you may be tempted to throttle or pause any publishers with low ROAS. This might be a smart tactic if you are buying from networks that offer pub-level bidding as the most granular optimization strategy. However, it can be a harmful approach if you buy through Moloco, as we utilize user-level buying strategies. In a user-level buying strategy, if we throttle an entire publisher because they have a lower ROAS it limits the scope of users and potential whales.

For Moloco campaigns, the performance is not the reflection of the publisher, but the reflection of individual bid-requests. Let’s assume that some users in the learning phase have not converted to purchase. This will in turn be a negative sample, and will be ingested in model training, effectively de-targeting users with similar a combination of signals. Moloco’s user-level bidding can also find good users in poor-performing publishers: Even if Publisher A has a lower ROAS than Publisher B, Moloco could still identify users on Publisher A that can drive a higher ROAS. For these reasons, throttling publishers is seldom a good strategy for ad campaign performance.

The specific machine learning models that we use in customer campaigns take many different variables into account including your own first-party data. And, our machine learning engine is constantly (every hour!) updating based on live transactions as they happen so your campaigns are getting smarter every day.

As a growth marketer, you are looking for ways to optimize your campaign’s performance. Now that we discussed how targeting/throttling may not help you with your campaign’s performance, let’s shift to  things you can do to optimize your ad campaign’s performance. Here are a few tips successful advertisers use.

Campaign Optimization Tips

1. Optimize your creative

Creative size formats

Depending on the region, Moloco Cloud DSP enforces uploading creative ads in various size formats. There are also optional formats and you should take advantage of supplying all the possible sizes as it gives you more opportunities to showcase your ad. In short, the more variations, the more opportunities to maximize campaign performance.

Use evergreen creatives

Ensure you have evergreen creatives (ones that have historically performed well as part of your creative rotation, especially when doing A/B testing)

Monitor and adjust

Look at the creative groups’ performance for the late funnel metrics that are important to you. This may be IPM, CTR, CVR and CPI for example. Remove those that are not performing well compared to other creative groups.

Add new creatives

Continually add new creatives to replace underperforming ones and monitor/adjust after at least two weeks.

Use video formats

Videos typically perform well compared to static ads and you can experiment with different video lengths. Our experience shows that videos up to 30 seconds usually perform best.

If you need assistance with creatives for your campaign, Moloco Studio can help.

2. Use A/B testing on your creatives

A/B testing can help you discover your best performing ads.

Monitor your ad’s performance after the minimum number of impressions (usually after at least 2 weeks) and remove or refresh any ads that are not performing well.

3. Expand to new geos

Find new opportunities in other countries or regions.

4. Consider a weekly budget vs a daily budget

While a fixed daily budget ensures that the daily spend does not exceed the daily budget, a weekly budget allows more flexibility in spending the set budget more efficiently. For example, a weekly budget setting may allow higher spending on some days when there are favorable bidding opportunities and less on other days when the bid opportunities are not as favorable while still not exceeding the fixed weekly budget.

‍