Understanding programmatic advertising basics and the role of machine learning (ML) is crucial when launching your first in-app ad campaign. This baseline knowledge helps identify the differences between programmatic and traditional ad networks, and also helps marketers understand why ML is not a silver bullet that will solve their marketing challenges overnight. The power of programmatic lies in its ability to learn and recognize patterns on its own, a process that takes time.

At this point you’re probably thinking, “OK. But how long does this learning phase take before I can scale?” This is a great question that we address below by explaining the three phases a campaign needs to navigate to achieve greater reach, engagement and performance all while stabilizing spend.

Phase 1: Campaign Model Training

In the first phase of a programmatic campaign, relevant data about a target audience is fed into a ML model so it can “learn.” This learning includes analyzing the data and adjusting campaign details to drive a desired action. The ML models analyze data points and identify patterns to predict the likelihood that specific ad units, placements, and other campaign parameters will drive the desired action.

The more our models know about a target audience, the more accurately it can bid on impressions so that your ads reach the right eyes at the most efficient price. During the training phase, we use a variety of data points to optimize our ML models, including:

• Device specifications, including make and model.
• Geographic identifiers, including continents, countries and ZIP/DMA codes.
• Temporal metrics, like time of day and day of the week.
• Application data, including store IDs and categorizations.

• User-specific behaviors, including downloads, interactions, and purchases.


Model Training Mechanics

As mentioned above, our ML models use diverse datasets to identify patterns indicative of high-value customers. These patterns are based on predefined advertiser goals such as registrations and lifetime value (LTV). A few questions we try to answer in our model training phase are:

• What is the likelihood of a customer taking a desired action if we show them an ad?
• What ads should we show a specific user?
• If the user taps an ad, goes to a landing page and takes a desired action, how likely will they be to generate revenue?
• What is a user’s predicted LTV?

Every time a bid request comes in — in other words, an impression opportunity to show an ad to a user — our Nero platform decides whether or not to bid and how much to bid based on the answers to these questions.

An important note on model training timelines:

The model training timeline is not a fixed interval. Rather, it depends on the quantity and quality of data we’ve collected. Specifically, whether we’ve collected enough unique events (clicks, conversions, registrations, etc.) to enable accurate predictions.

To make the model training period go faster, we recommend that partners pass revenue data to our team through their attribution partner. The richer the data, the faster our models can optimize a campaign to hit an advertiser’s goals.

Phase 2: Optimizations

Our team starts tweaking, testing, and refining a campaign from day one. These initial optimizations involve minor adjustments that ensure campaigns are converting. For example, we’ll adjust bids to ensure we can gather as much user and conversion data as possible.

Once we have gathered enough conversion data, then we begin making more robust optimizations tailored to your specific goals. If your goal is to maximize ROAS, then our optimizations across different dimensions — such as rewarded/non-rewarded offers, App Store ID, app category, etc. — will focus on acquiring as many payers as possible.

Here are a few more examples of optimizations we make once we have collected enough user and conversion data:

• Adjusting bids to ensure we are hitting any CPA targets while also trying to hit daily budget caps.
• Optimizing towards inventory sources with high conversion rates.
• Adjusting bids based on the performance of certain ad exchanges and rewarded/non-rewarded offers.
• Tweaking creatives based on A/B tests and testing new concepts.

Phase 3: Scaling

Once we hit certain goals, we consider the campaign training phase to be over and we’re ready to scale. Determining when an ad campaign is ready to scale involves evaluating several key factors to ensure that the campaign has the potential to maintain or improve performance as it reaches a broader audience. Here are some of the more common goals we evaluate:

• CPA Goal: A target CPA is usually reached once we have enough unique conversions, causing CPAs to drop to a target rate. We consider a campaign out of the learning and training phase once we consistently hit an advertiser’s target cost-per-acquisition (CPA) or registration (CPR) goal.
• ROAS Goal: Once we reach and stabilize CPA goals, we start to evaluate weekly payer consistency to achieve advertisers’ ROAS goals.
• Event Goal: This is when a partner isn’t looking strictly at ROAS but also aims to achieve a high volume of a specific event.

For advertisers interested in achieving more scale, we slowly start increasing spend to new levels and let the campaign adjust to the new spend levels and inventory sources. Generally, we see performance fluctuate each time we increase spend since our models have to adjust to the increased budget and tap into more inventory.

Giving a campaign a few days to stabilize at its new inventory and spend level gives our CSMs time to make any adjustments needed to ensure we still hit ROAS goals. It also lets our team prioritize quality inventory and creatives as we scale up the campaign.

TL;DR

So to answer the question, “How long does this learning phase take before I can scale?” It depends.

Ultimately, the campaign’s ability to progress efficiently through these phases depends on the strategic management of data, optimizations, and scaling efforts tailored to specific campaign objectives and market conditions.

To learn more about unlocking the full potential of IAA through programmatic campaigns, you can contact the LifeStreet or check out some of our case studies.