Viral Coefficient and Organic Uplift in Mobile Games: How to Calculate Them, What Realistic Values Look Like, and the Traps in Building Them Into eCPI

The viral coefficient (K-factor) is the number of new users each existing user brings in through invites or shares. If the average user sends 1.5 invites and 20% of the people invited install, K is 0.3. Acquire 10,000 players with ads, and those 10,000 bring in 3,000 more, who in turn bring in 900, and so on.

The number UA teams actually use for budget decisions is not the K-factor alone. When you run paid campaigns, organic installs rise too, even without a single invite. Downloads push up your store ranking, and people who saw an ad later search the store for the game by name. The ratio of organic installs that come along with paid installs is called organic uplift. The two numbers overlap in part, but they are not the same thing, and mixing them up makes your CPI look lower than it really is.

According to AppsFlyer, 14.1 billion of the 24.8 billion gaming app installs in 2025 were paid. The remaining 10.7 billion or so, about 43% of the total, were installs not attributed to ads. That 43% contains users who came through virality, users who came on the back of advertising, and users who would have come anyway. This guide, current as of October 2026, explains how to separate the three. The difference between blended CAC and paid CAC is covered in Why Your CAC Is Not Your CPI: How Game Studios Should Calculate Customer Acquisition Cost and Set an LTV:CAC Target.

What is the viral coefficient (K-factor), and how do you calculate it?

The K-factor is the average number of invites per user (i) multiplied by the install conversion rate per invite (c), and it tells you how many members of the next generation each generation of users produces.

The formula is K = i × c. Here i is the average number of invites or shares across all users, including those who never invite anyone, and c is the share of invite recipients who actually install. Tenjin's glossary uses the same formula and gives a mobile game example of 3 invites per user times a 25% conversion rate, for K = 0.75. A common mistake is to calculate i as the average among users who sent invites. If only 10% of users send invites and they send 5 each on average, i is 0.5, not 5.

When K is below 1, virality does not grow on its own. It shrinks with each generation and then stops. What it does instead is multiply the number of users you acquired with ads. Starting with 10,000 users at K = 0.3, each generation looks like this (illustrative calculation).

The sum over all generations is the starting user count ÷ (1 − K). At K = 0.3 that is 1 ÷ 0.7, or about 1.43x. At K = 0.6 it is 2.5x, and at K = 0.75 it is 4x. There is a catch. Tenjin's figure of "about 75 organic installs for every 100 paid users" counts only the first generation. Add the later generations and the same K = 0.75 yields 300 additional users. Conversely, if the second and later generations will not actually arrive within your calculation window, using only the first-generation figure is the safer choice. What decides this is cycle time, covered in the next section.

Why does cycle time matter as much as K?

Cycle time is the average time from a user's install to their sending an invite and the recipient installing, and with the same K, a longer cycle time sharply reduces the number of viral users who arrive within your payback window.

The generation-sum formula assumes infinite time. UA budget decisions usually have to resolve within 30 or 90 days, so the real multiplier depends on how many generations pass within that window. Starting from 10,000 users and varying K and cycle time gives the following (illustrative calculation).

When K is low, cycle time has little effect, because at K = 0.3 the first generation produces most of the contribution. The higher K gets, the more cycle time changes the result. For a game with K = 0.6, a 7-day cycle time delivers 92% of the theoretical maximum within 30 days, while a 30-day cycle time delivers only 64%. That is why timing matters when you design an invite feature. Prompt sharing while the player is still engaged, such as right after the first session or on the first victory screen. An invite sent as a player is drifting away stretches the cycle time and means there is no friend left in the game when the recipient arrives.

Churn works in the same direction. Retention is not part of the K-factor formula, but a game where few players stay long enough to send an invite will show a low i to begin with, because, as noted above, i is an average across all users.

Why is K above 1 so rare in games, and what values are realistic?

Few games sustain K > 1, and there is effectively no reliable public benchmark, so the starting point is measuring i and c separately with your own data rather than borrowing outside numbers.

K > 1 is a state where growth accelerates as the user base grows. But every player has a limited number of people to invite, and as more of those contacts already have the game, c falls. So K often spikes at launch or during a moment of buzz and then comes back down. Tenjin also describes K > 1 as rare and hard to sustain, and notes that seasonality, updates, and viral UGC can move K significantly.

The reason so few numbers are published is clear. On organic uplift, AppsFlyer says that an earlier analysis found the variance so extreme that no statistically valid single number could be released. K-factor varies even more from game to game, because the placement of invite features, the reward structure, how social the genre is, and the messaging environment in each country all differ. Ranges like "games are 0.3 to 0.7" that appear on glossary and calculator sites mostly have no verifiable source, so this guide does not cite them. If you only need a rough sense of where to start, Tenjin's bands (0.1 to 0.5 moderate, 0.5 to 1 strong organic contribution, above 1 viral) are a reasonable reference.

When you design invite rewards, be aware that raising K can lower user quality. A large field experiment at a mobile game company, which tracked 66,427 inviters and 97,181 invitees over 60 days (Mengzhou Zhuang of the University of Hong Kong and co-authors, presented in an April 2026 seminar), compared two reward schemes. One paid the inviter only if the invitee installed; the other paid for sharing alone. With high rewards, the conditional scheme produced 77.4% fewer invites, but its acceptance rate per invite was more than 11 times higher. Yet invitees acquired under the conditional scheme spent 30.3% less time playing and 76.6% less money. The researchers attribute this to social pressure created by the conditional reward, which boosts installs but reduces engagement after install. A mechanism that raises c does not necessarily bring in good users. K should be read alongside invitee D7 retention or payments, not install counts alone.

How is organic uplift different from the K-factor?

Organic uplift is the ratio of additional organic installs that come with each paid install, and it includes not only user-to-user invites but every indirect inflow that advertising creates, such as higher store rankings and more brand searches.

AppsFlyer models organic installs as "baseline + uplift × measured (paid) installs" and calculates uplift as (organic installs − baseline) ÷ paid installs. The baseline is the key. It is the organic installs you would have received without any ads: the share driven by ASO, existing brand awareness, and featuring. If you skip the baseline and divide all organic installs by paid installs, you count inflow that has nothing to do with ads as ad performance. AppsFlyer also cautions against treating uplift as a complete reflection of virality (K-factor), noting that the two concepts only partially overlap.

Measured invites (referrals tracked through deep links) can appear in your MMP as a separate, non-organic path. But if a player just mentions the game's name in a messenger and the friend searches the store to install it, that install is counted as organic. So within a single game, part of virality shows up as referrals and part is scattered into organic uplift. If you add the two numbers together, sort out the paths first so you do not count the same user twice.

Most published uplift figures are either old or come without methodology. TUNE wrote in 2017 that, on average, apps saw about 1.5 organic installs per paid install, but it did not disclose the sample or period, and it is unclear whether the baseline was removed. Mobile UA agency Upptic, in a March 2026 article, gave one organic install per five paid installs (20% uplift) as a cautious benchmark. The fact that sources diverge this much is itself the case for measuring your own.

How much do organic installs fall when ads are turned off?

One experiment with published methodology found that a full advertising shutoff cut organic installs by 20 to 30%. But the extra organic installs per $100 of spend came to around 7% of paid installs, far from the 1:1 people often expect.

Harang Ju, Michael Zhao, and Sinan Aral (MIT) analyzed a global ad shutoff run by a large US mobile game developer in 2018 and 2019 (arXiv 2504.16151, first posted April 2025, revised July 2026). They tracked six mature games on both Android and iOS for 500 days. The shutoff covered all publishers' campaigns, including search, social, and display, for about three weeks, from day 75 to day 98 of the data. Store featuring periods were excluded from the analysis.

There are three takeaways. First, paid advertising lifted organic installs rather than cannibalizing them. Second, store rankings are the most likely path for that effect: when the researchers added ranking to the model, the relationship between ad spend and organic installs disappeared. The interpretation is that ads raise download velocity, which lifts category rank, which in turn increases organic visibility. Third, the effect is smaller than many assume. Per $100, there were 32.3 paid installs and 34.5 total installs, so attribution missed about 7% of installs, which works out to an uplift of about 0.07 (our calculation). The study's limits remain: it covers mature games and a single company's data. In exploratory platform-level estimates, Meta ads showed a large organic spillover and Google ads close to zero, but the authors stress these are not causal estimates.

The practical message of this research is not the size of uplift but its direction and structure. Uplift does not grow in proportion to spend forever; it appears when enough downloads concentrate to move rankings. This is why the same budget can produce different uplift when spread evenly over a month versus concentrated in a short window right after launch.

How do you measure organic uplift?

The most accurate approach is an experiment that observes the baseline directly by turning ads off or splitting regions, and when experiments are not feasible, you can approximate it with daily-data regression and store console traffic-source splits.

AppsFlyer recommends a similar sequence: slice the data by country and OS, remove days with abnormal install spikes such as featuring, and then solve for baseline and uplift together. Channels that links cannot track, such as TV, outdoor, and influencers, should be separated out. When those channels are running, a rise in organic cannot be credited to paid digital ads. The details of experiment design (sample size, duration, geo-split tools) are covered in How to Design an Incrementality Test for Mobile Game UA: Holdouts, Geo-Lift, Sample Size, and Reading the Result.

For Android games, the store listing performance reports in Google Play Console are useful. They split traffic into Google Play search (searches for the app's name or a closely associated brand), Google Play explore (browsing, autocomplete suggestions, and category searches such as "racing game"), and ads and referrals. If explore traffic rose during a campaign, suspect ranking-driven uplift; if name searches rose, suspect brand searches by people who saw the ad, or word of mouth. From June to July 2026, Google switched the primary metric in these reports from completed acquisitions to unique clicks (user intent), so check that the metric definition matches before comparing with earlier periods. Completed installs remain available on the Grow users overview page and in downloadable reports.

How does uplift change eCPI and CAC?

Uplift-adjusted CPI is ad spend divided by paid installs plus the organic installs the ads created, and it sits between paid CPI and blended CPI. If you use it to raise bid caps, account for the error in your baseline estimate as well.

Suppose you spent $100,000 in a month for 40,000 paid installs, and the same month saw 30,000 organic installs. A baseline estimated from a period with no ads is 24,000 installs (illustrative assumption).

The three numbers serve different purposes. Use paid CPI to compare channels and campaigns. Use uplift-adjusted CPI to decide how many more players you actually get if you increase spend. Use blended CPI only to report company-wide acquisition efficiency. If you set bid caps at the $1.43 blended CPI, you are counting the 24,000 baseline players who would have come regardless as ad performance.

As the last row shows, uplift-adjusted CPI is sensitive to the baseline estimate. Raising the baseline by just 3,000 installs halves the uplift (to 0.075) and moves CPI from $2.17 to $2.33. That is why AppsFlyer recommends tracking direct ROAS and blended ROI side by side rather than folding uplift into projected ROAS. In practice, it is safer to discount uplift. Use the low end of your measurement, and assume diminishing returns as spend grows, since uplift only appears in ranges where rankings move. The roughly 0.07 observed in the ad shutoff study above is a reasonable anchor for such a conservative assumption.

Check the quality of uplift users too. Organic users often retain and convert to payers better than paid users, but there is no guarantee that uplift users who arrived through a ranking effect match baseline organic users. Applying the LTV of the whole organic cohort to uplift users inflates LTV along with eCPI. The basic eCPI calculation and common mistakes are in How to Calculate eCPI: Common Pitfalls and Optimization in UA, and genre and regional paid CPI reference points are in Mobile Game CPI Benchmarks 2026: Why One Average Is Useless.

Which players do you need to grow both virality and uplift?

The i in the viral coefficient comes from players who stay long enough to send invites, and uplift comes from installs concentrated enough to move rankings, so both numbers depend less on install volume than on the share of players who stay after installing.

As the K-factor section showed, i is an average across all users. The more players who leave on day one, the fewer chances there are to send an invite at all. A viral loop starts turning only when many players reach the point where social features such as guilds, friend matches, and co-op content unlock. Store rankings work the same way. If many installs are followed by quick uninstalls, the ranking effect does not last, and the uplift built on it ends quickly.

Playio is an Android-based community of 5 million gamers that matches players with games that fit their tastes, using genre preferences, play history, and in-game behavior data. Its default pricing is CPI, and it also supports CPE based on in-game actions such as reaching a playtime goal or a specific level. If you set the campaign goal at the level where social features unlock, or the step just before a player's first guild content, what you are paying for is not a single install but a player who stayed in the game long enough to be able to send invites. Reward structures that encourage daily play also help reduce uninstalls during install surges such as the first weeks after launch, holding on a little longer to the visibility that rankings create.

You can find more details here. (https://playioadsen.oopy.io/bizdeck)

Key takeaways

The viral coefficient K is average invites per user × invite conversion rate, and when K is below 1 it multiplies the users you acquired with ads by up to 1 ÷ (1 − K). That multiplier assumes infinite time, though, so you also need to calculate how many generations pass within a 30- or 90-day payback window, which depends on cycle time. In games, K > 1 is rare and short-lived, and there is no reliable public benchmark. Invite rewards can raise conversion while lowering invitee quality: in one field experiment, invitees under a conditional reward scheme spent 76.6% less. Organic uplift is the organic installs that ads create indirectly through store rankings and brand searches, and it must always be calculated net of a baseline. In an ad shutoff experiment with published methodology, organic installs fell 20 to 30%, but the extra organic installs per $100 were only about 7% of paid installs. Use uplift-adjusted CPI for spend increase decisions, paid CPI for channel comparison, and blended CPI only for company-level reporting, and discount uplift conservatively when you build it into your numbers.

For inquiries about Playio's advertising solutions, reach out at: [email protected]

Sources

• Tenjin, K-Factor glossary (K = invites per user × conversion rate per invite; mobile game example 3 × 0.25 = 0.75, about 75 organic installs per 100 paid users; bands 0.1–0.5 moderate, 0.5–1 strong organic contribution, above 1 viral; K > 1 rare and hard to sustain; shifts with seasonality, updates, and viral UGC; does not account for churn; distinct from organic uplift): https://tenjin.com/glossary/k-factor/

• AppsFlyer, The organic uplift multiplier in app marketing: Separating fact from fiction, Shani Rosenfelder and Adi Belan, updated 2025-11-27 (organic installs = baseline + uplift × measured installs; uplift = (organic − baseline) ÷ measured installs; only partially overlaps with K-factor; no statistically valid single number could be published due to extreme variance; slice by country and OS, remove outliers such as featuring, paused-spend, zero-organic country, and no-paid-country baseline tests; separate untrackable channels such as TV, OOH, influencers; track direct ROAS and blended ROI in parallel): https://www.appsflyer.com/blog/measurement-analytics/organic-uplift-multiplier-app-marketing/

• Harang Ju, Michael Zhao, Sinan Aral, Advertising Spillovers in Mobile Apps: Evidence from Ad Shutoffs and Store Rankings, arXiv 2504.16151 (v1 2025-04-22, v2 2026-07-10): https://arxiv.org/abs/2504.16151

• The same paper, v2 full text (500 days in 2018–2019, 6 games × 2 operating systems; all-publisher ad shutoff on days 75–98 across search, social, and display; featuring periods excluded; organic installs down 20–30% during the shutoff, about 27% at the 11-day bandwidth; per $100: 32.295 paid, 34.472 total, 2.177 organic (p = .51); CPI $3.10 on paid vs $2.90 on total; category ranking 2.3–3.3 times worse; spend-organic relationship disappears once ranking is included; platform estimates exploratory and not causal — Meta organic 9.519, Google −0.211; single large developer limits generalization): https://arxiv.org/html/2504.16151v2

• Shanghai Jiao Tong University Antai College seminar notice, Mengzhou Zhuang (University of Hong Kong), Unpacking the Incentive Conditions of Referral Programs: Evidence from A Large-Scale Field Experiment, 2026-04-29 (mobile game company, 66,427 inviters and 97,181 invitees over 60 days; under high rewards, the conditional scheme sent 77.4% fewer referrals, had over 11 times the acceptance rate per referral, and its invitees spent 30.3% less time and 76.6% less money; 222.9% more contact attempts): https://acem.sjtu.edu.cn/academic/95811.html

• TUNE, Paid and Organic: (Measurably) Better Together, Becky Doles, 2017-05-17 (about 1.5 organic installs per paid install on average; sample and period not disclosed): https://www.tune.com/blog/paid-organic-measurably-better-together-organic-uplift/

• Upptic, What is Organic Uplift? How to Measure Incrementality in Paid User Acquisition, Ethan Lofton, 2026-03-19 (cautious benchmark of one organic per five paid, 20% uplift; can exceed 1:1 at scale with strong product hooks; holdout, geo-lift, and baseline modeling methods): https://upptic.com/what-is-organic-uplift-how-to-measure-incrementality-in-paid-user-acquisition/

• AppsFlyer newsroom, State of Gaming for Marketers 2026, 2026-01-14 (24.8 billion gaming app installs in 2025, 14.1 billion paid, $25B UA spend, +3.8%, 9,600 gaming apps): https://www.appsflyer.com/company/newsroom/pr/gaming-marketing/

• Google Play Console Help, Understand and grow your app's user base (traffic sources: Google Play search = app name or closely associated brand searches, Google Play explore = browsing, autocomplete, category searches, ads and referrals; primary store listing metric switched to unique clicks from June–July 2026, with completed acquisitions still in the Grow users overview, statistics, and downloadable reports): https://support.google.com/googleplay/android-developer/answer/9859173?hl=en

• Own calculations: K = 1.5 × 0.2 = 0.3 generation totals, 1 ÷ (1 − K) multiplier, 30- and 90-day cumulative users by K and cycle time, all-generation sum for Tenjin's example (300 users), attribution gap from the paper's figures (34.472 ÷ 32.295 ≈ 1.067) and uplift of about 0.07, and the example of $100,000 spend, 40,000 paid, 30,000 organic, and a baseline of 24,000 (or 27,000): illustrative calculations and assumptions