Sponsored By

Featured Blog | This community-written post highlights the best of what the game industry has to offer. Read more like it on the Game Developer Blogs.

8 tips for traffic analysis

Let’s be honest: everyone buys traffic, and the first thing we care about is its price. But large volume and low price do not guarantee the success. To get the high-quality but cheap audience we have to analyze the traffic, and the sooner is the better.

Eugene Matveev, Blogger

December 2, 2015

7 Min Read

Often, especially in small companies, people work with traffic in the same way. They receive a budget for acquisition, they buy traffic, they mark that the influx of registrations started, so the company is delighted and moves on to other tasks. But suddenly, this traffic does not bring the money, and it turns out that the marketing budget is wasted. Reasons for failure may be both monetization problems of the product and the low quality of the traffic. After all, according to various estimates, the share of fake (fraudulent) traffic in the overall structure of paid registrations now ranges from 20% to 60%.

To avoid such situations, you have to analyze the traffic from the very beginning. And we'll tell you how to do it.

Use relative metrics
Such metrics as Installs Cost, Installs, Revenue, of course, are important for understanding the scope of the campaign. Though, they tell nothing about the traffic quality. To evaluate if the quality of your traffic is any good, you need to operate the relative metrics in terms of user:

  • ARPU (Average Revenue Per User);

  • CPI (Cost Per Install);

  • Paying Share;

  • Retention;

  • ROI (Return On Investment).

These metrics describe the quality of the traffic and allow you to compare several channels of acquisition.

Use the cohort analysis
Record the date of registration or the first visit of users for each traffic source, and then study the retention metrics and monetization metrics after a day, two days (one week, two weeks, a month) from the registration date.

For example, let's take source A, fix a week interval of user registration and calculate metrics for each cohort:

  • cumulative 1-day ARPU (how much one user brought an average in a day of life in the project);

  • cumulative 2-days ARPU;

  • cumulative 7-days ARPU;

  • cumulative 28-days ARPU;

  • and so on.

Similarly, take the cumulative revenue and divide it by the cost of acquisition of a single user to get the X-days ROI metric. Keep track of this indicator, and you will notice the moment of payback of the traffic: your ROI will exceed 100%.

When comparing multiple sources of traffic, ROI is a key indicator. In addition, we recommend to count the following retention rates for each cohort: 

  • 1-day retention;

  • 7-days retention;

  • 28-days retention

  • and so on.

These metrics are useful to compare the sources cited by one point - in fact, it is wrong to compare the efficiency of sources if they were active during a different time.

Set the target events
Metrics like retention are easily forged. If the client is focused on, say, a 1-day retention, the traffic source could simply log in the application the next day or write the corresponding bot.
Therefore, set the additional events, that will show you that the user is properly motivated.

the screenshot is taken from devtodev system

These events might be the passage of the particular, but not the first level, the entrance into the store, clicking on any not the most obvious element of the menu, etc. Consider these events for each traffic source together with the above metrics, and you will find the channel where the retention is high, but the target action is not performed. If this is a frequent situation, then it would be better to reject this channel.

“Slice & dice” the data
It is possible that the bots operate out of a single IP, or they are registered in one hour or come from the same site. If you analyze the data in all of these cuts, especially in combinations, you will have all the chances to notice suspicious deviations.

It's not only about dealing with bots: if you see that the traffic from a particular country or a particular site does not pay off, you are entitled to request a partner to turn off such sub-source.

the screenshot is taken from devtodev system

Develop a procedure for evaluating the traffic
In general, the traffic analysis is not a one-time task. Try to produce it regularly: evaluate the incoming traffic and optimize algorithms for analysis. Set daily and weekly traffic evaluation procedure. Make your specialist start the day not only with a cup of coffee but also with the quality analysis of traffic in the last N days.

If you collected a set of metrics for analysis, set the target events and "sliced & diced" traffic on several parameters, you could occupy your traffic manager for a half an hour in the morning. You may apply the monitoring system to every metric with the help of statistic mechanisms like confidence intervals, the rule of three sigmas. This is the way to increase the chances of finding abnormal variations in the traffic.

Work with reliable partners
But do not be afraid to experiment. 

In general, traffic acquisition can be compared to the formation of an investment portfolio. Both are highly remunerative and, shall we say, conservative instruments. Start your portfolio with proven traffic partners like Google, Facebook, or those with whom you already have a relationship. There will come a time when someone offers you a lower CPI. If you already have a stable and predictable source of traffic, you may take a risk because the new partner might appear to be better. Minimize the risk due to the diversification of the sources.

If you regularly evaluate the quality of traffic, switch off poorly performing subsources and experiment with new partners sometimes, then you step by step learn how to determine the next month number of users and how much you will have to pay for it.

Set the limits
Nevertheless, don't forget to set limits for each partner. Firstly, you get the upper threshold of the budget for acquisition. Secondly, you may avoid the situation where untested partner suddenly invoices you unexpectedly large sum.
Of course, these limits may vary for different partners. Set them in proportion to your confidence in the source of traffic.

Deduct the whales from the analysis
Whales are those users who bring the largest amounts of money. It is possible that the whole pot from one source of traffic is created by a single user, but the rest of users have rather low conversions. It would be better to make sure that it doesn't work this way.

Now, let's solve the exercise.
There are three partners: A, B and C. We paid each partner  $100 for  the acquisition. Traffic from the partner A brought us $15, traffic from the partner B – $20, and traffic from the partner C – $25. 

Which one is better? I would like to say that it is C. But the correct answer is “I don’t know”. To make a decision, you have to clarify a few points:

  1. How many users came from each source? Focusing on absolute metrics, we can not compare the quality of these channels. We need relative metrics, and for this it is necessary to know the number of users.

  2. What is CPI for each source? Without knowing the cost of acquisition, we can not calculate ROI, and ROI is a key indicator when comparing sources.

  3. How much did time pass from the moment of user registration for each source? Perhaps a campaign with partner C started a month ago, and with the other partner the campaign took place only for one week?

  4. Even if all partners started their campaigns at the same time, is that enough time to talk about the differences? For example, if only 2-3 days passed, then A and B could still catch up and overtake partner C.

  5. Are there whales in the analyzed traffic? What if one user paid those $25, and all the other N-1 users did not bring a penny?

I hope this material would be useful to you as after all, we all buy traffic and it's nice to learn how to do it without overpaying.

P.S. I would also like to invite you to a free webinar where we will give a few tips on how not to overpay for the traffic. The webinar will be held on December 8, 2015, at 20:00 Moscow time. Sign up now!

Read more about:

Featured Blogs
Daily news, dev blogs, and stories from Game Developer straight to your inbox

You May Also Like