How to convert freemium to premium users

Introduction

 

Edit: I made the website Manastats.com, where i publish more infographics about gaming!

Thank you for reading my post! As stated other places this post will be fairly long (7000 words), so if you just want to gather some quick information, check the freemium infographic or the Reddit post. Another option is to skip to the last part of this post, where you can read all my hypotheses and conclusions. The last option is simply to use the table of content and find the answers you need from there.

I’ve spent hundreds of hours writing my master thesis to figure out one simple thing: “Why do people buy in-game virtual items that have no functionality at all?”. I decided that my case should be based around freemium MOBA games.  The idea is, that if you can figure out why people buy virtual items they don’t need, then it might be possible for businesses selling actual services like Spotify, Dropbox and Skype that does things you actually need to upsell freemium to premium users. The point of my thesis was to discover which factors makes people buy freemium skins, what would happen if these factors were changed, and if it was possible to put people into groups only based on their buying behavior.

It is not a coincidence that the freemium business model is getting so popular. Both Dota2 and League of Legends have a huge fanbase, and are both in the top 5 of most played games. League of legends is actually more played than the rest of the top 5 combined.  League of legends make shitloads of money. Some phone games might even gross more money than League of legends.

Definitions

If you know what the freemium model is, and what a virtual good is, then you can skip this part.

The freemium model:

According to Seufert (2014) the freemium model consists of 3 components:

  1. “A price point of $0 renders the product accessible to the largest number of people.” (p. 3)

The product is free. The aim of this is to make more people use it.

  1. “Some users will not engage with the product beyond the free tier of functionality.” (p. 3)

In this blogposts we will also discover how many of the respondents buy in-game items. Wall Street Journal claims than in businesses such as Dropbox and Skype the number of users that pays for the product are less than 5%. (http://www.wsj.com/articles/SB10000872396390443713704577603782317318996)

  1. If the product is extremely appealing to a group of users, and the product presents the opportunity to make large or repeat purchases, a portion of the user base may spend more money on the product than they would have if the product had cost a set fee. Thus, the revenue fulcrum, or the crux of a product manager’s decision to develop a freemium product, is the potential to maximize scale, paid engagement, and appeal to the extent that the total revenue the product generates exceeds what could be expected if the product cost money.” (p.3)

Virtual goods

Virtual goods as a term has been confused as to what it includes or does not include: what are the difference between a virtual item and a digital item? Is an MP3 file and a color theme in a game(skins) in the same category? Fairfield (2005) definition have seen much use lately, and that is the following:

They [Virtual items] are rivalrous. If one person owns and controls them, others do not. They are persistent. Unlike the software on your computer, they do not go away when you turn your computer off. And they are interconnected. Other people can interact with them. This kind of code I term “virtual property.” (p. 1049-1050)

That means that if one person has the item others do not, because they are rivalrous. This definition excludes things like MP3 files: even though it is illegal by the law (copyright protection), I can give a MP3 file to a friend, and we both have the file. The next thing Fairfield mentions is, that virtual items are supposed to be persistent. If the file is lost easily, or if they do not exist for a certain time, it cannot be virtual goods.  As we discover later, one of the key reasons players don’t buy skins, is that they are afraid to lose them.

Previous research?

I was not able to find a single academic paper researching freemium interactions in free games. There is done little research in both LoL and Dota 2 however, and in the freemium model in 3d social worlds like secondlifte.com. If want to research it yourself, I would suggest “Lehdonvirta”, “Guo and Barnes” and “Hassouneh and Brengman”.

What factors make people buy skins in a freemium game

The 1-million-dollar question.

To discover these factors, I was looking into research and found 3 factors I wanted to work with in this context:

  1. Customization: Lehdonvirta (2009) and Guo and Barnes (2011) found this was important under the hedonic attributes, and is very relevant to MOBA games. Fun fact, in World of Warcraft, one of the most used enchants in Vanilla (the first WoW game) was “Beastslayer”, it was completely useless (mostly), but the red glow was very appealing to people. People do a lot for cosmetics.
  2. Attention-craves: This category is based on the shopper type Status seekers and hedonic self-expressionist shoppers, that mainly shop for community acceptance (Hassouneh and Brengman (2011). So yeah, this category is pretty obvious: some people like attention, even though they don’t want to admit it.
  3. Community-oriented shoppers: This is completely new since a lot of MOBA games is pouring a lot of money in the e-sport scene, people feel like they can support e-sport by supporting the games. A random user in League of Legends asked on the internet (http://forums.na.leagueoflegends.com/board/showthread.php?t=2839199), and found that supporting Riot was the biggest factor of all as to why people bought skins. This also led me to conclude that e-sports work opposite that of normal sports. In normal sports you support the teams, not the organizations.
  4. Habit shoppers: I wanted to test if buying skins simply could be a habit: the whole system is designed so it is really easy to buy skins combined with low costs (Hence the name, Microtransactions). It wouldn’t be insane to conclude that this process could be a habit.
  5. Social influence: this factor consists of 2 things: firstly, can there be pressure from a group of people to buy skins? – secondly, I wanted to see if playing a lot with friends would that have an effect on buying behaviors.
  6. Hedonic motives: These motives are things like “It is fun to buy skins”, “it is entertaining to buy skins”, and so on. This was a backup; if all other things failed I could at least prove that people would buy skins for hedonic reasons.

Testing the theory before the survey was distributed

I wanted to take a pilot test to see if people did buy skins for the reasons I found, therefore they got presented for an open textbox, before they saw the categories. Out of 41 answers, the answers were in the following categories:

  • AC: One in this category.
  • Price: Five in this category.
  • Customization: 20 answers in this category.
  • Community oriented: Four answers here.
  • Habit: Two answers.
  • Hedonic: Nine answers.
  • Other: Nine answers were in this category, mainly because of valueless answers: “no comment” “shit’s pretty, yo” and “Skins = WINZ”.

Freemium Analysis in MOBA games:

Distribution

The survey was distributed on the following sites: Reddit.com, Hardwareonline.dk, the official forums for DotA 2 (playDotA.com/forums/), and League of Legends (leagueoflegends.com/en/community/). To give incentive to take the test there was a chance to win a small reward of 4×10 dollars gift cards to buy skins in the MOBA games.

After removing the partly completed answers there were 692 respondents left. The number of samples exceed the 5% sampling error (384 samples) and are a bit less than the +3% error (1067).

outliers

The results:

case

How many Percent of people buy skins?

Out of 692 participants, around 10% had never bought cosmetic in-game, which is quite impressive. This means that 90% did buy in-game cosmetics, which will be debated later. Cronbach’s alpha is .851. (>.7) which is considered good. That means 90% of all players in league of legends had bought skins in the last 3 months. This shit is amazing. A couple of years ago Wall Street Journal assumed that the normal rates would be 1-5%. Dropbox is 1%, Skype is converting around 10%, and Spotify is kings with around 30% conversion.

What is the age of the average MOBA gamer?

Frequency Valid Percent Cumulative Percent
Valid Under 15 years old 16 2,3 2,3
15-19 years old 279 40,3 42,6
20-24 years old 271 39,2 81,8
25-29 years old 102 14,7 96,5
30-34 years old 18 2,6 99,1
35-39 years old 4 ,6 99,7
40 years old or older 2 ,3 100,0
Total 692 100,0

This information match earlier polls, and Riots earlier infographics.

Why do people buy skins?

Before reading the table below you should know that Customization is shortened “custom”, Social influence is shortened “SI”, Attention Craving is shortened “AC”, Community shoppers “Co”, Habit “HA”, and Hedonic motives is shortened “hed”.

The scale was the following

1= Not important
2= Slightly important
3 = Moderately important
4 = Important
5 = Very important

It is worth to note that a thing cannot be different degrees of “not important”, this means that when the value exceeds 2, it is already slightly important.

To paint the right picture, each item was measured with 3 different statements.

 

I buy skins because…
Statement Tag Value
Customization
It will enhance my own gaming experience Custom1 3,11
It will make me use a champion more Custom2 2,63
Sometimes I dislike the classic skin Custom3 3,05
It will make the champion/hero look like i want Custom4 3,82
Social
I shop to have fun with my friends SI1 2,15
People whose opinions that I value, prefer that I use skins SI2 1,59
People who are important to me think that I should use skins SI3 1,56
Attention craving
I shop to make people think I am better with that champion,

and not a free player

AC1 1,78
I shop to show my new skins to others AC2 2,29
I shop to separate myself from others who use that champion AC3 2,47
I shop to make people notice me AC4 1,71
Community-oriented shopper
I shop to support my favorite team Co1 1,99
I shop to support E-Sport Co2 1,99
I shop to support the game-companies (Riot or Valve) Co3 2,2
Habit
It is a habit to buy skins Ha1 1,72
It is easy to buy skins Ha2 2,55
Of Impulsive buying Ha3 3,38
Hedonic motives
Skins are fun Hed1 3,8
Skins are enjoyable Hed2 3,86
Skins are very entertaining Hed3 3,55

case1

The scenario might unfold much like this: First, the student misbehaves. Then the teacher approaches the student and reprimands him or her for misbehaving. Because the student finds the negative teacher attention to be reinforcing, he or she continues to misbehave-and the teacher naturally responds by reprimanding the student more often! An escalating, predictable cycle is established, with the student repeatedly acting-out and teacher reprimanding him or her.” (http://www.jimwrightonline.com/htmdocs/interventions/behavior/ncrft.php)Customization is the only scale that has a good correlation on .485 with Hedonic motives, which can explain why these two categories have the highest mean value. Another interesting thing is that CUSTOM and ATTENTION have a good correlation: it makes sense that people like to customize their characters, and enjoy the attention they are getting for it. Another high correlation is Attention and Habit. Hassouneh and Brengman (2011) states that status seekers are intermediate/heavy shoppers (p. 331), which could indicate that it is a habit for them to buy virtual items. It also makes sense that things that give attention can become a habit. One example is attention seeking students in lower level classes:

Social influence scores over 2 “Slightly important” in question 13 in importance when buying skins in MOBA games, was not accepted. Two out of three items scored below 2 “slightly important”. Although it was predicted it would be a low score, it wasn’t predicted it would fail. One item made it.

(Customization scores over 2 “Slightly important” in question 13 in importance when buying skins in MOBA games, was accepted, and scored second highest.

(Attention Craving scores over 2 “Slightly important” in question 13 in importance when buying skins in MOBA games, was accepted.

Community-oriented shoppers scores over 2 “Slightly important” in question 13 in importance when buying skins in MOBA games, was partly accepted. One item scored below 2, and two items scored over 2. Supporting Riot/Valve, and e-sport in general, seems like a good motivation to buy virtual items, but supporting favorite teams does not seems like a strong motivator. This is quite opposite that of normal sports where in, for example soccer, the teams themselves sell a lot of merchandise. Teams like Real Madrid and Manchester united sell 1.4 million Jerseys a year (http://www.therichest.com/sports/soccer-sports/top-10-highest-selling-club-soccer-jerseys/?view=all) and probably more tickets to their games. On the other hand, soccer fans don’t seem to be eager to support the sport or organisations like FIFA.

Habit scores over 2 “Slightly important” in question 13 in importance when buying skins in MOBA games, was also accepted.

The last thing to mention is that the only negative factor that seems to have influence, was the item: “The prices on the skins” with a mean value on 5.27 (see analysis part 2).

Gifting is riots real money maker

As seen on the gifting table, 75,4% says they are gifting skins to friends in-game, meaning that 3 out of 4 who play games do it.

Here we see that people who gift in general use more money than people who do not give gifts to friends in-game. Of the 16 people that spent over 40 dollars a month on freemium games, 15 of these people gift friends, while only one does not. We see that, in the categories from 10 dollars+, people who gift spend more money on the game in general.

Demographic differences:

case2

As seen above: North America seems to enjoy the hedonic motives more than the Europeans. They agree more on these two items: “Skins are enjoyable” (difference .27), and “Skins are very entertaining” (difference .37). It should be noted, that finding two differences in 20 items will partly accept hypothesis H2, except the hedonic motives, which will be noted in the future model.

We see that Hed item 3: “Skins are very entertaining”, is highest in the group with 15-19. It is .224 higher than 20-24 and .530 higher than group 25-29. We can also see that group 20-24 are .306 higher than group 25-29. This pattern shows that the younger people are the more entertaining skins are for them. Which is an interesting finding. Venkatesth (2012) found something similar: “Some interesting results are that the effect of hedonic motivation on behavioral intention is stronger for younger men” (Venkatesh et al. (2012) p.15).

Different groups of people segmented by buying behavior

By using the data from above I made a cluster analysis to segment all people into 8 different groups (Also called clusters) based on their answers. These groups are mixed and calculated by the value they put in each. If you, for example, answered a high value in Social and hedonic motivation you belong to group 1, but if you have a high value in social influence, habit and Community oriented shopping you belong to group 2. Group 1, 2 and 8 are heavy shoppers.

Group 1: Strong values in Social, a solid value in hedonic, and a good value in custom. This value is close to what Author describe as a “Social Shoppers“

Group 2: The highest value in Social influence. (Note that only 2 clusters out of 8 have high values in social influence). The main difference here is, that this group have high numbers in CO and HA also, and the highest value in custom shopping. This fits the group:  “Social Shoppers“.

Group 3: Shopping for habit, fun and customization. This is a perfect example of a “habit shopper” that only shops for habit, but still have strong values in the other fields, but all groups have that (except 5).

Group 4: The only cluster with a high AC, matching the shopping type: what the auther will call a “Status seeker”

Group 5: This cluster has none of the six reasons for shopping, this is an interesting group, and might be shopping for other unknown reasons, even though there are no indication from pretests or the inspiration sources.

Group 6: this group only shops to support either the game, esport or esport teams, and is a prime example of “community oriented shoppers”.

Group 7: The only group that only shops because of enjoyment and customization. These people fit the “Relaxation/gratification shoppers”. They shop for fun and “They mainly buy avatar-appearance items (i.e., clothes, body parts, hair, skin etc.).” (Hassouneh & Brengman (2011) p. 330).

Group 8: This groups is a mix of cluster 6 and cluster 3. They are community oriented shoppers, and also habitual shoppers. A mixture of both.

A last thing to note is cluster size. I accepted everything higher than N>30. We see there is 3 big groupgs 7,6 and 5 and rest is around 39 to 61.

Analysing cluster types

Since we are looking at relatively small groups, we are happy with a significance of 0.005 or higher. The following segmentation bases are significant:

  • Howsocial: “How social would you describe yourself in the game?” (How social is flipped, so the scale matches the others.)
  • Timesweekwithfriends: “How many times a week do you play with friends you made online?”
  • Timeweek with online friends: “How many times a week do you play games with “real life” friends?”
  • Helpsiteuse: “How many times a week do you use webpages that help you in game?”
  • Negative social: “Which of the following factors if any, are likely to hold you back from buying virtual items in freemium games? – Social influence from friends or others”
  • Spent on virtual items: “How much money do you spend on buying virtual items each month in freemium games?”
  • Shopping frequency: “In the last 3 months, how many times did you purchase virtual items, on average per month?”
  • Reputation riot: “What is your opinion on the following game producers?– Riot”

case4
In the first box we see the values for each cluster. Green indicates it is higher than the mean value, white indicates it is the same, and red indicate is lower. To be green or red, it has to be over/under 0.20 points higher. We see that cluster 5, besides having low value in every mixed-scale, also mostly has low value all over the place, and especially in shopping frequency and money spent.
The marked numbers are shopping frequency and money spent.
 

We see that group 2 has a green line all the way, and also an orange on the negative factor. Group two, four and especially group eight, have increased values in frequency and spent on virtual items.

Before we looked at the segmentation bases here are the game items that are significant for League of legends:

  • Enjoyment 1: “How much do you enjoy the game?”
  • Enjoyment 2: “How exciting is the game?”
  • Enjoyment 3: “How much do you enjoy purchasing inside the game?”
  • Enjoyment 4: “How satisfied are you in general with your shopping through the game?”
  • Enjoyment 5: “How much do you enjoy your skins after your bought them?”
  • How long before purchase league: “How long did you play the game before you bought your first “skin” for real money?”

There are no significant data on DotA 2, which might be because of the “small” sample group off 133/138 respondents. (I tried with only 2 clusters (bigger groups), but would still get the same amount of items significant.)

case5
What we see here are group five again doing very bad overall. At this point we can make the cluster number 5 a DUMMY and use it as worst case scenario. Another thing is, in that the last box “how long before purchase?“, the higher the number is, the worse it is.

REPUTATION a factor?

The reputation with Riot is a bit tricky because DotA2 players might not have such a good reputation with Riot as League of legends players. Therefore I made a version with only League players. Here we see the reputation with Riot:

Report
Mean
CLU8_1 Reputation_RIOT
1 5,31
2 5,68
3 5,24
4 5,00
5 4,91
6 5,91
7 5,20
8 5,72
Total 5,34

Not surprisingly 5 also does really bad here, while group 8, 6 and 2 has a very good relationship with Riot. However, all groups in general does very well, reputation wise, with Riot: a 5 on the 7-point scale is “Quite good”.

My theory Combined with existing theory.

With the numbers from the previous 3 figures, I’ve made a figure like the one from Hassouneh & Brengman (2011)’s theory with my views, and the old theory views (can be found on the CD, called “segments”).

Type of products and shopping perceptions are not measured because it is only about one product: “skins”, and my questionnaire already had 26 questions. To start measuring what shopping perceptions the respondents had, would take up extra space.

Main motivations shopping behavior

 

 

Main motivations frequency Money Spent Assumptions
Social shoppers SI Frequent Normal Heavy shopper
Social shoppers SI  – CO – HA Very frequent Much Heavy shopper
Habit Ha Frequent Normal none
Status seekers AC Very frequent Much Intermediate shoppers
Dummy None Do not shop often Not much None
Community shoppers CO Frequent Normal None
Relaxation/ Gratification Seekers Custom- HED Frequent Normal Heavy shopper
Heavy community shoppers Ha – CO Very frequent Very much None

The assumptions are from Hassouneh & Brengman (2011). Where there is no assumption, there is a new shopper type.

The two first clusters are different kinds of social shoppers: they both have in common that Social influence is High. They want to express themselves through shopping (customization) and that they shop for freedom (in this case high hedonic values, because it is exciting and fun). However, in theory, these kind of people should be heavy shoppers. Cluster 2 is heavy shopping (both frequency and money spent), and also enjoying the game more (see the enjoy1, 2, 3, 4,). Both are very social. The thing that differs these two groups are the fact that the second group is more extreme. They are community shopping, and at the same time they also shop to support e-sport and Riot, and they shop more out of habit. That explains the heavier shopping, and enjoying the game more. The first one is more relaxed, and only shop for the social aspect. So in these 2 clusters we found 2 different kind of social shoppers.

The third cluster is Habit Shopper. Shopping is very “normal” with normal frequency and money spent. None of the social factors seems to bother them: they score a median value in all of them, including the negative social factor item. This group does not use help sites, and nothing special in enjoyment and reputation either. This is interesting, since the data says that they are buying out of pure habit. None of the other factors are really important to them. The only factor that is not close to median is their shopping frequency (and also negative social factors that are unlikely to make them stop purchasing). This group proves, that habit buying exists in freemium games.

In cluster 4 the idea from attention craving came from “status seekers”, and it is the only field with attention craving. In the article it states that they are intermediate shoppers. I found that they shop more frequently than some groups, but not as much as social shoppers. They still spend “much” money, which fits the theory really well. In the social aspect they are playing a lot with real friends, and it is likely that their friends have some control of their purchases (negative social). This makes sense since they buy to impress people (shopping for gratification/compensation) (from Hassouneh & Brengman (2011) p. 12). They are using help sites a lot, maybe that is what they are compensating for. Help sites in League of Legends often tell you how to play, build and so on, this could indicate that they don’t play that well. They enjoy aspect 3, 4, 5 but their general reputation with Riot is bad.

Cluster 5 is our DUMMY. These people don’t like any motivational factors, don’t shop often, don’t use much money, don’t play much with real and online friends, and aren’t social in general. They do not use help sites or enjoy the game, they do not have a good reputation towards Riot, and they play a long time before making their first purchase.

This group is exciting because they potentially prove four theories: The 1st one is that the more social you are, the more you buy. The 2nd is that motivational factors actually make you buy more. Also, they are proving that enjoyment and reputation is bad => not spending a lot of money. And the last thing is, that this group play a long time before they make their first purchase. Someone might think that these people just pressed the worst possible choice every time they could in the questionnaire because they had a bad day, or don’t like anything. However, it is significant.

Moving back to cluster 6 is the community oriented shopper. What we see here is a group that do not use a lot of money or shop frequently. They shop primarily to support their favorite team, E-sport or the game company. Enjoyment is high in all 5, and reputation is very good. These respondents simply just want to give back for playing a game they enjoy for free.

Cluster 7 reminds us of Relaxation/ Gratification Seekers: Their main shopping motivations are shopping for freedom, and fun seeking. They get freedom from customization, and one of the items from hedonic motivations are “fun”. These guys are not heavy shoppers, as the theory would have them to be, their shopping patterns are normal. Their values are close to their mean values in all others. They simply enjoy the fact that they can buy whatever they want, when they want to, free from any RL restrictions. Hassouneh & Brengman (2011) found that this group: ”Mainly buy avatar-appearance items” (p. 330). Even though it was measured in a social world where you can buy everything you can in the real world, this cluster fit this shopper type the best.

Cluster 8 is the mix of cluster 6 and 3, and the fact that they both have a habit of shopping, and want to support the game, make these people spend the highest amount of money of all clusters. Their mean is 3,33, while the closest competitor is below 3. The mean for all groups is 2.64. This group also have the highest shopping frequency of all clusters. They generally use “much” time with online friends, and use help sites “much”. They are “very enjoying” in some of the enjoyment categories, and have a good relationship with Riot. They might enjoy the game very “much” and want to give back to Riot by buying skins. We see that they also play with online friends, and not so much with real friends, so the theory about social aspect have positive correlations with buying come into play here.

How can Riot trick people into buying more skins?

Below you see which factors that can be increased to make gamers spend more on virtual items, and have a higher shopping frequency. Customization was ruled out because it wasn’t significant. Other than that we see that age correlate badly with hedonic motives, as mentioned earlier, and “good” with habit. Since social influence was really bad, I replaced it with “how many times a week do you play with friends you made online?” It showed a moderate effect on shopping behavior.

case6

Disclaimer: the RSquared value is pretty low, which means that these things have an effect, but can only explain very little of the actual number, I believe this is due to bad questionnaire design. – I should have asked for intent, instead of past actions. – The results still works, though.

To analyze the results from the factor analysis we will use these brackets:

  • <0.01 no effect
  • <0.10 small effect
  • 10-0.20 some effect
  • 20-0.30 moderate effect
  • 30+ high effect

These brackets are an expression of the effect. As long as they pass 0.01, they have passed the hypothesis, but it is also worth noticing how strong the effects are, which is the reason I made these brackets. The hypothesis will be explained in the findings at the end of this paper. For now, we are only looking at how strong each path is.

The strongest factor was “Habit”, which had high effect: .36 to frequency and .31 to “spent on virtual items”. This is interesting because in the frequency analysis Habit was not such an important factor when people were buying.

Hedonic motives had high/modern effect .31 to frequency and .27 to “spent on virtual items”. Hedonic motives scores high in everything, and something could indicate, that to sell virtual items in a game, you have to make it fun and entertaining. We also again see that “age” have a negative covariance with hedonic motivations.

Attention craving is the third biggest with 2 moderate effects to the items. As mentioned earlier, “habit” and “AC” are closely correlated, however the covariance in this figure is negative, which seems a bit off.

“Playing with online friends” also scores moderate with .22 to frequency and .20 to money spent, which was what previous theory suggested, so these findings are a perfect match with the theory. As mentioned earlier, playing with real friends was not significant, but there was a correlation between these two items.

The last construct “community-oriented shoppers”, had some effect: .12 to frequency and .10 to money spent.

The difference between Dota 2 and LoL.

There is not enough data to see if there is a significant difference between Dota 2 and LoL gamers, but the data suggest the following:

By looking at that table we see that DotA 2 scores are overall higher than LoL, the only paths that have lower value is AC -> spent on virtual items (.20 > .15) and Playing with friends (.24>.10), while the money spent is the same. DotA 2 have some high values at the habit construct -> money spent, and Frequency on 57 and 54. With LoL having 35 and 29, it seems like a really big difference. This might be due, again, to the low numbers of respondents that play DotA 2.  These numbers should be investigated more before making any conclusions of the differences between MOBA games. However, of course it fails the hypothesis that there is no difference between MOBA games in this thesis.

List of Findings

Here is a list about the results, which will be discussed on the next page.

List of hypotheses
H1 There are more than 5% of the players in freemium game that make in-game purchases. Supported
H2 There are no differences in demographic factors (Except age*) in gaming communities. partly supported
H2.1 Age is negatively correlating with Hedonic motives. Supported
H3 Respondents that play more with real friends spent more money on virtual items. Not supported
H3.1 Respondents that play more with online friends spend more money on virtual items. Supported
H4 Respondents that play more with real friends buy virtual items more frequently. Not supported
H4.1 Respondents that play more with online friends buy virtual items more frequently. Supported
H5.1 Social influence scores over 2 “Slightly important” in question 13, in importance when buying skins in MOBA games. Not supported
H5.2 Customization scores over 2 “Slightly important” in question 13, in importance when buying skins in MOBA games. Supported
H5.3 Attention Craving scores over 2 “Slightly important” in question 13, in importance when buying skins in MOBA games. Supported
H5.4 Community-oriented shoppers scores over 2 “Slightly important” in question 13, in importance when buying skins in MOBA games. Partly supported
H5.5 Habit scores over 2 “Slightly important” in question 13, in importance when buying skins in MOBA games. Supported
H6 Social influence  has a positive relation to frequency and money spent  on virtual items in freemium MOBA games. (Replaced with “played with online friends”, which was positively related to frequency and money spent.) Significant
H7 Customization has a positive relation to frequency and money spent on virtual items in freemium MOBA games. Not supported
H8 Attention craving has a positive relation to frequency and money spent on virtual items in freemium MOBA games. Supported
H9 Community-oriented shopping has a positive relation to frequency and money spent on virtual items in freemium MOBA games. Supported
H10 Habit has a positive relation to frequency and money spent on virtual items in freemium MOBA games. Supported
H11 Respondents that have high enjoyment values buy virtual items more frequently. Partly supported
H11.1 Respondents that have high enjoyment values spend more money on virtual items. Partly supported
H12 The same models that are used for Virtual social worlds can be used for freemium MOBA games. Supported
H13 There is no difference between the game types (DotA 2 and LoL). Not supported
H14 It is possible to segment consumers in freemium MOBA games. Supported

Discussion about the hypotheses

It is worth noticing that H3, H4, H6, H11 and H11.1 are “not supported” because they are “not significant”. If a study like this will be conducted in the future, they might still check these hypotheses.

H1 was supported. We saw that 90% of the respondents, 623 out of 692, have bought virtual items. This number seems really high, which makes me skeptic. I made sure to say in the distribution part that it was not a requirement that they purchased. When asking my network, 14 out of 14 people said that they had purchased in-game items.

H2 was partly supported. Out of 20 items there were two hedonic items that had a difference. These items were, that North America had bigger mean values in Hed2 and Hed3. This item is still marked as partly supported since it is expected that out of 20 items, there are some differences.

H3 and H4 are not significant, however H3.1 and H4.1 are significant, and the effect is as predicted: that the more social a person is in a game, the more often they buy, and the more money they spend. While H3 and H4 are not significant, there is a significant positive correlation between playing with real friends and online friends, so there is something there for future research.

H5.1 social influence is the only item below two, out of all six factors. Guo and Barnes (2009) found Social influence to be quite important. “Social influence was the most frequently mentioned issue in this study. There were 24 statements made related to social influence. Its importance index is 2.03.” (p. 89). The reason this did not pass, might be because of the design of the questionnaire: when the items to influence was borrowed my UTAUT2 model, I predicted they scored low due to the nature of the items. If some other items had been used, it might have passed.

H5.2, H5.3, and H5.5 were all supported.

H5.4 was partly supported. The item about supporting an e-sports team was not >2, but the two other items, about supporting companies and e-sport in general, were above 2. Which, as mentioned, is quite the opposite of real sports.

H6 was not significant and had to be dropped. Again, we see the weak social influence, which is counter-evidence to what it is supposed to be. It got replaced with time spend with online friends, which had a positive effect on frequency and money spent.

H7 was not supported. It simply had really low factor loadings: maybe the category was too broad. It was one of the most important factors for buying.

H8, H9, H10 were all supported.

H12 was supported. By using the UTAUT2 framework this paper had a lot of expected results with 9 hypotheses supported, and 3 partly supported, this framework is doing a good job. It had to be modified, but the theory presented by Guo and Barnes (2009), and Hassouneh and Brengman (2011) from virtual worlds, are fitting for freemium MOBA games.

H13 was not supported. There was a big difference in some of the numbers. Overall DotA 2 scored highest, however, as described in the analysis, it might be because of the low numbers of DotA 2 respondents.

H14 was supported. It was possible with a cluster analysis to segment eight groups. The groups fit with some of the categories Hassouneh and Brengman (2011).

10. Conclusion

In this thesis we proved that five of the six factors were important when buying virtual items. The only factor not important was social influence. However, if we change the question items, this might be important too. In the factor analysis we proved that four out of six factors had a positive correlation with frequency of purchase and money spent. Social influence and customization did not make it due to not being significant. However, by replacing social influence with playing with online friends, there was a positive effect on purchase behavior.

What we can conclude is that community oriented shopping exists. Players want to support e-sport and the gaming companies. Shopping for attention and being social also increase sales, so MOBA games might want to make a bigger effort in making people like each other in the game, which will also reduce the amount of people that are “toxic”: a term made by RIOT describing players that are bad losers, and when they lose they turn negative, racist and angry. (http://forums.na.leagueoflegends.com/board/showthread.php?t=3854624) These are a huge problem in MOBA games.

We see that it can become a habit to buy virtual items because it is easy and “cheap”. MOBA games can exploit that by making promotions to players that have not bought an item in a long time. For example: say a player used to spend 20 euros a month on skins in the MOBA game, and have not done so in three months, then there might be a promotion to make them buy again and increase the habit.