How to start marketing games as a developer: choosing sources and budget

• “If your studio has $2-3 million to spend on marketing it's worth hiring an analyst to calculate the key metrics you should be targeting”
• “If you have a small budget you can and should try to mimic the audience for tests this allows you to quickly get results and then move on to more structured purchasing strategies”

What happens behind the scenes of your game will determine its trajectory for success. You could develop an incredible game, but players may never even know its there or try it out without sufficient marketing.

Each studio works within their own capabilities, and while some may have millions to spend on marketing, others may have less. So, how can you get the best possible marketing while keeping your own personal criteria and circumstances in check?

In this article, AppQuantum’s head of UA, Artem Sprygin, offers insight and advice on marketing for developers, from assigning budgets to what platforms to use for your campaigns.

Launching a product and connecting potential sources of interest depends strictly on the product itself. Before launching, you need to select the most relevant traffic sources. But how exactly do you choose these sources and what differentiates them? This directly depends on the game’s budget, payback period, and the amount of monthly budget that the developer can allocate for marketing.

Let’s delve into the numbers and analyse several scenarios in a Q&A format.

Remember, marketing is about testing and adapting to a constantly changing environment. What worked a few months or even weeks ago may no longer be effective. With this in mind, let's look at several scenarios.

We have a good product and $2-3 million for marketing: What do we do?

The first step is to connect an attribution system (MMP, mobile measurement partner), which will allow you to determine which channel the user came from, such as Facebook, Google, or organic. Use proven tools like Appsflyer, Adjust, Singular, etc.

Next, we need to find the best sources for our target markets. Often, you may not know which market to choose initially or which one will become the primary market.

If you didn't develop the project with market fit research in mind, rely on intuition and data from MMP indices. Always look at current indices from main MMPs (e.g., AppsFlyer and Adjust). If unsure which market the game will succeed in, start with the main top markets. While anomalies can happen, don't count on them initially.

For example, start with the USA, then other Tier-1 English-speaking countries: Great Britain, Canada, Australia, New Zealand. Then Tier-1 Asian countries: Japan, Korea, Taiwan - this is a standard set.

What platforms should you use for campaigns in the USA?

Select at least three top advertising platforms. If it is Android in the US, then likely:

• Google

• Facebook

• AppLovin/Unity

iOS:

• Applovin/Unity

• TikTok

• Facebook/Google (if you’re familiar with SKAD/SKAN)

For hybrid projects featuring both iAP and iAA in various proportions, AppLovin is worth considering if you use their mediation, Applovin Max. This allows for ad revenue transfer and running Blended/Ad ROAS optimisation.

You must also gather the minimum required data on payers to optimise ROAS campaigns. Without them, optimisation of such campaigns is impossible; the minimum threshold is more than ten unique payers on day zero.

Thus, the third source highly depends on the percentage of iAP in the project and what kind of mediation you use, if any. Mediation gives access to all types of campaigns. Unity Ads and Iron Source will work well if you have Level Play Unity installed. If using Max, then use AppLovin.

How to plan the launch of new traffic sources and select target geos

We have the main intended markets and three main sources. Now, allocate your budget.

Start with a design test that includes geographic and creative optimisations.

If the project has organics from previous tests, great. If not, purchase a minimum number of installs to obtain primary metrics. Always watch for anomalies in your geos. For instance, if you notice that retention in Mexico reaches targets and is similar to the USA, it is worth testing new creative hypotheses in these markets. This is a general approach to selecting geos when specific data is unavailable.

In most cases, you’ll have to test creatives in countries that are close to T1 in metrics but more cost-effective in terms of testing.

If your product has a large share of iAP% Revenue, you'll need to use a different strategy. Regardless, you must start with install campaigns to analyse your metrics. Once you have collected enough data, you can choose In-App ROAS optimisation, which will become available after training the networks.

During your design test, ensure the Learning Phase is accounted for as the necessary budget for the transition to ROAS optimisation. Afterwards, if your product continues to have a significant share of iAP% Revenue, you can start working on different types of optimisation: Value, In-App ROAS, Purchase, and Unique Purchase.

For iAA or Hybrid iAA+iAP revenue based products, use Ad ROAS, combined with repeatable events for ad views and Hybrid ROAS optimisation.

Video Networks and Google offer AdROAS optimisation out of the box, but Facebook does not. The choice of optimisation goal for ROAS depends on the share of iAP and iAA% Revenue in the product.

What about geos for creative testing?

This depends heavily on the type of creative being tested. If you don't have pre-tested creative, choose a country whose user behaviour closely matches your primary market.

For example, if your main market is Japan, choose somewhere like Thailand. Look at time zones or the necessary behaviour. The USA is often compared with Mexico, the Philippines, or Brazil due to similar time zones and user behaviour metrics (e.g. retention rate, but not solvency). This helps determine the correct IPM.

Why test in Mexico, the Philippines, or Brazil? To understand your creative's IPM. If you need metrics beyond IPM, include those in your creative test, such as retention on a specific day or the average number of ad views.

Run such a test on a slice, with a benchmark creative based on relevant metrics, typically the top first purchase test. Conduct iterative tests comparing new creatives to the benchmark. If a new creative outperforms the benchmark, launch it into active campaigns and monitor its performance. This is a basic technique for testing creatives.

Usually, these tests are conducted on Facebook. While some are moving away from this because social traffic differs from other types of traffic, Facebook is still a good choice for quickly obtaining a relevant audience for your test without excessive costs.

Now we've created a coverage map for optimisations, creatives, and geos. What's next?

Set up campaign structures to meet minimum thresholds for different types of optimisation. This information is available in source documentation for new developers.

Alternatively, you can contact the support to find out how much data - installs or unique purchases - is needed to pass the ROAS optimisation threshold on Facebook or other networks. They may say that for a 7-day threshold you need 50 unique paying users. From this, you can calculate the necessary figures to avoid mistakes in your analysis.

Ideally, launch a campaign for at least the chosen optimisation window and add two days. For example, for a D7 window, the first closed D7 cohort will only be in nine days. If you look strictly at the optimisation window, it may seem that the algorithm does not quite hit the mark, but it needs time to close the cohort to get data for optimisation.

Therefore, you should also look at the trends in the metrics and how CPM, IPM, Retention, and ROAS metrics are changing.

If you see a trend for improvement and the achieved KPI ROAS suits you, then you scale up. If that’s not working, look at what exactly is drawing down. Maybe the creative is not working well, maybe it’s the wrong geo, the type of optimisation, or maybe it’s a specific placement if we’re talking about Facebook. That is, you need to dive deeply into the specifics of the channel. For example, on Google, this means optimisation by geo, creative concepts, and video-only/non-video-only creative types.

We will definitely discuss working with Google Ads another time; there are many nuances that also need to be considered.

In networks such as Unity, Vungle, and Mintegral, optimisation occurs by geo, creative, and optimisation type. Usually, they’re used to launch singles for large geo clusters, but this depends on the strategy, and there is the possibility of optimisation at the publisher level. Just ask your managers for a White List (WL) of publishers with products in a similar genre that perform well on the channel. This way, you will increase the likelihood of showing ads to more targeted users.

If you optimise for CPI in Video Networks, you can add publishers with weak performance to the blacklist without any problems. But if you optimise for ROAS, adding to the blacklist can be painful. The ROAS algorithm, which is usually quite smart, will eventually understand that you should not send impressions there. Manually disabling publishers in ROAS can cause overtraining, which causes a decrease in efficiency, unnecessary spending, and wasted time.

In addition, it is worth reviewing which publishers are on the blacklist and giving them a new chance every three to six months since the market is constantly changing and nothing is set in stone.

How to measure the effectiveness of campaigns

It’s worth paying special attention to analytics to understand why your product is paying off. All products are different, with different tails, taxes, shares of IAPs, etc. All this must be taken into account.

If your studio has $2-3 million to spend on marketing, it's worth hiring an analyst to calculate the key metrics you should be targeting.

Focus on the average cost of attracting a user, retention, and LTV indicators on the 30/60/90th day at the start of the first advertising campaigns for the project. If the project uses a hybrid monetisation method, you can use a proxy metric.

We use our internal metric of the average number of ad views to the end (Reward Finish Per User of a cohort day). After understanding the rate of profit accumulation, you can gradually expand the payback window to avoid falling into a cash gap until day 180. It is important not to set targets immediately for the 360th day of payback - this is an unrealistic scenario and can lead to significant losses.

You can also use ROAS metrics, but you need to understand that they can fluctuate depending on the share of in-app purchases and advertising monetisation. On some traffic, everything will be fine, but on others, it won’t. It all depends on the sources, so you need reference metrics to understand that you are paying off.

At a minimum, check once a month to make sure that the plan matches reality because there can be fluctuations in both directions. You may underestimate or overestimate and have extra money, which most likely means you are ineffective in advertising spending.

Another scenario is when overestimation occurs. For example, LiveOps made a mistake in an update, and now the ARPU growth rate from the 14th to the 30th day falls. If you don’t make at least some simple dashboard that compares the accumulation of factual revenue to the predictor, it will hurt.

I believe this is also UA's task because UA is interested in fulfilling its P&L expenses, which it directs to the business. And we can use the entire resource pool. Creatives, analytics, and everything must work together.

And what if we have a $300K budget?

All stages remain, but with such a budget, most likely, the team may not have a dedicated analyst with ready-to-use in-house analytics. In any case, you need to develop KPIs according to which purchases will be made.

In a negative scenario, you need to monitor the accumulation of metrics and user behaviour yourself. To see if the money you receive really comes in the right form, be sure to carry out a reconciliation. Fraud is still a thing, and analytics covers not only metrics analysis but also traffic validation.

There is payment validation (store reports) from Google Play, App Store, and there is payment validation from AppsFlyer, etc. And these are always different. Even a large service can sometimes produce ambiguous data, and you should always be sure that you are not receiving fraudulent transactions.

What else should I pay special attention to?

Once again, focus on creatives. Creatives are a dominant part of success in UA. Therefore, the choice of channels closely correlates with the choice of creative concepts that you upload. Initially, when you don’t know anything, it’s great to test creatives on Facebook and take the most IPM creatives into battle = CTR * CTI * 1000 | (Installs * 1000/Impressions).

But as soon as you begin to understand that there is one concept on Google, a second on Facebook, and a third on AppLovin, you need to iterate and dig specifically into the catchy mechanics used in a particular creative in a particular network. This pays off because the audience and behaviour are different, and the creative concepts work differently.

So, when you have a million or less, you probably won’t have such variability in creatives. Most likely, these will be top Facebook performers, and you will promote them everywhere, and this is not very optimal.

If you need to bring a top creative from one network to another (because the tops are often different), remember that some sources react faster to changes but also burn out faster. This applies to SRN channels. If we talk about AppLovin and Google, they react to changes more slowly. Sometimes, a creative lives for a very long time, 3-6 months. But when it burns out, you need to sort out new concepts from these sources; you can’t get away with just minor changes.

On Facebook, the burnout time for creatives is much faster; on TikTok, it generally takes a week. In the latter, you must use your creative resources, which allow you to automatically slightly modify the creative so that it lives a little longer.

What if, on the contrary, there is a huge budget of $7+ million. Will anything change?

Most likely, there will be a team of writers making scripts, creative flow tests, and constantly rolling out iterations, as well as an analytics team. Then, you can iterate and quickly update your creatives in rotation.

But in other regards, the entire workflow remains similar.

We talked about the situation when the product is obviously good. What if the product is untested?

If a product is untested, it still needs to go through standard product tests. You look at a cross-section of key metrics such as CPM, IPM, CPI, ARPU Growth, CPB (Cost Per Buyer), PU% (Conversion to Paying User), OPP (Order Per Payer), AOV (Average Order Value), retention, and playtime. Then, based on these metrics, you decide whether the product is viable.

If it is viable, it goes through a second wave of benchmarks. Based on similar metrics, these benchmarks indicate that the product can thrive with a certain type of traffic. For this, you'll need a historical database of these cross-sections and supporting hypotheses to refer to.

If the product is not of high quality but the business goal is to promote products with weaker metrics (for example, as a PR case or for business value), you should use non-standard channels (Playtime/Offerwall CPE). Here, you find deep events where the RPI is higher than the predicted CPI you should pay.

You may encounter problems with the initial event chosen due to the high activity of such traffic and the lower share of iAP%. Then, there is the issue of fraud, which needs to be addressed through several iterations. Look for real payouts for achieving the target events needed for this purpose, adjust the funnel, and eventually, everything will start to work as intended.

Where can we get inspiration for creatives? What else can we try?

In fact, it’s pretty straightforward - conduct competitive analysis, track the distribution of traffic through analytical tools to specific creatives, and try to highlight the strengths of your projects while also focusing on existing trends.

Marketing is always about testing new hypotheses. Creatives, like knowledge, can burn out quickly, so it’s essential to maintain rotation and follow market trends. And getting started on ASO for your projects right away is always a good idea. We've had projects where just working with ASO boosted conversion to install by a significant 50-60%, but that's a story for another time.

The main thing is that if you have a small budget, you can and should try to mimic the audience for tests. This allows you to quickly get results and then move on to more structured purchasing strategies.

Remember, marketing has one primary goal - to generate project revenue with a clear and predictable revenue flow. In the first stage, you should not rely on a 360-day ROI; this can only be done if you have enough data to build a valid predictive model. Plan for a maximum ROI of 180 days, but even in this case, everything depends on the economics of not just your project but also your business.

Edited by Paige Cook