Distribution has always been a problem. Even when an app like Draw Something hit 50 million installs in 50 days, “viral with no UA budget” wasn’t an accident.
The number isn’t the lesson, the mechanic is.
I sat down with Wilson Kriegel, who was COO of OMGPOP when Draw Something launched in February 2012. He told me the part most retellings skip. The company was less than a quarter from running out of money. They had let go of people. The launch saved them.
Less than a month later, Zynga acquired OMGPOP for $210 million. Wilson’s full account is on the Mobile UA Show podcast episode.
I also spent time at the Zynga era, working alongside the Words With Friends team after that game’s acquisition. Words With Friends is rarely called “viral”, but it has the strongest organics of any casual game I’ve worked on.
The reason is the same reason Draw Something blew up. An intrinsic social mechanic doing the distribution work. This post is what I learned from both, about what made hyper-viral apps actually work, and why 2026 isn’t 2012.
Page Contents
How Draw Something nearly didn’t happen
Wilson’s story isn’t a viral fairytale. OMGPOP started as a real-time matchmaking dating platform. It pivoted into casual real-time flash gaming. Then mobile arrived between 2010 and 2012, and the team had to find another bet.
By the time Draw Something launched, OMGPOP had spent four years trying different things. Wilson and the team had built engineering depth across genres. They knew what worked on Facebook and what worked on mobile. They had data infrastructure most one-shot teams didn’t.
The deliberate strategy across all that experimentation was non-obvious. Wilson described it directly: “We use our engineers, if you will, as a mechanic and our designers as a packing mechanism to try to grow our company versus purely relying on game mechanics that arbitrage a game to generate revenues.”
That’s not the language of “let’s go viral.” That’s the language of building distribution as a product capability, which is exactly what made Draw Something’s spike survivable when it hit. The Wikipedia entry tells the highlights, but Wilson’s account fills in the part you don’t get from press cycles.
What did the apps that “went viral” have in common?
The apps that went viral with little or no paid UA budget shared a single trait. An intrinsic social mechanic embedded in the gameplay or core experience. The one famous exception, Flappy Bird, is the lottery, not a strategy.
| App | Year | Intrinsic social mechanic | What carried distribution |
|---|---|---|---|
| Draw Something | 2012 | 2-player asynchronous Pictionary | In-game invitations to friends |
| Words With Friends | 2009 | Asynchronous multiplayer Scrabble | Long-running games kept friends pulled in |
| Pokemon Go | 2016 | Location + friends | Real-world social proof on streets |
| Wordle | 2022 | Shareable result grid (the boxes) | Daily ritual that generated daily shares |
| Flappy Bird | 2014 | None, single-player frustration | Cultural moment, screenshot sharing |
What we see: In 4 of 5 cases, the social mechanic was intrinsic to the experience. Flappy Bird is the exception that proves the rule, it went viral through cultural happenstance, not design.
Why doesn’t this happen the way it used to?
Distribution conditions changed materially between 2012 and 2026. Every lever that helped Draw Something is harder now.
| Distribution variable | 2012 | 2026 |
|---|---|---|
| App Store total apps | ~700,000 | ~1.8 million |
| Google Play total apps | ~600,000 | ~2.5 million |
| iOS attribution | Deterministic IDFA | Probabilistic SKAdNetwork + ATT |
| Top-of-chart dynamics | Low-budget viral hits possible | Paid-UA dominated |
| Network arbitrage | WhatsApp / SMS / Facebook reach available | Closed gardens, low organic reach |
What we see: The App Store has roughly tripled in size. Attribution is harder. Paid UA dominates the charts. App store size data confirms what every UA team feels, distribution is structurally tougher.
What did I actually learn from this?
There are two kinds of viral apps. Confusing them is the most common mistake I see UA teams make.
Type A: apps with intrinsic social mechanics
Draw Something. Words With Friends. Pokemon Go. The mechanic IS the distribution. You can design for this. It’s repeatable. Most of the playbook still transfers to 2026, even though the conditions are tougher.
Wilson made a point on the podcast that doesn’t usually make it into the retellings. By the time Draw Something launched, OMGPOP had spent three years building gaming portals and Facebook apps. They had engineering, data infrastructure, and an iteration cadence ready for spike load.
The team controlled the keyword database. They ran bi-weekly iterations on infrastructure. When 50 million installs landed in 50 days, they didn’t go down.
Most “viral attempt” launches fail because the team isn’t ready when the mechanic actually fires. A viral spike against an unprepared backend looks like a failed launch.
Wilson called out a specific tactical lever: invitation-flow optimization. They used network logins to amplify invites and reduce friction. The mechanic was the distribution, but the team still spent product engineering making the mechanic faster.
Type B: apps that hit a cultural moment
Flappy Bird. Sometimes Wordle’s boxes-shareability falls into this bucket too. Single-player or near-solo experiences that go viral through cultural happenstance, not designed mechanics.
You cannot engineer Type B. It’s the lottery. If you build a strategy on top of it, you’ve built a strategy on luck.
The trap most marketers fall into is trying to engineer Type B. You can only design for Type A. And in 2026, even Type A is harder.
What happens after the cultural moment ends
Pokemon Go had a tremendous 12-month run, then the press moved on. By any “viral cultural moment” metric, the story was over. Yet Pokemon Go still generates a billion dollars in revenue years later.
Type A apps have a longevity tail that Type B apps don’t. Words With Friends launched in 2009. It still generates meaningful organic installs in 2026. Flappy Bird, by contrast, was removed by its creator and the clones never reached the original’s heights.
| App | Original viral peak | 5+ years later |
|---|---|---|
| Draw Something | Feb 2012 | Acquired Mar 2012, sunset by Zynga ~2013 |
| Words With Friends | 2009-2010 | Still in top word games charts in 2026 |
| Pokemon Go | Summer 2016 | Still a major revenue title in 2026 |
| Wordle | Early 2022 | NYT acquired, daily ritual for millions |
| Flappy Bird | Early 2014 | Removed by creator; clones never matched original |
What we see: Type A apps retain value years after the press moves on. Flappy Bird is the cleanest counter-example, without an intrinsic mechanic to compound on, the cultural moment was the entire story.
Three mistakes UA teams make in this area
From running creative for app launches across our portfolio, the same three mistakes show up across teams:
- Trying to make a single-player app go viral. If your app has no intrinsic social mechanic, no amount of “share this” buttons will change that. You’re not in the viral business, you’re in the paid-UA business. See the 3C principle on creative concentration.
- Treating virality as a marketing add-on, not a product feature. Asking marketing to “make this go viral” when the product wasn’t designed for it is asking the wrong department. The viral mechanic gets built into the product. Marketing amplifies what the product earns.
- Confusing “shareability” with intrinsic social mechanics. A “share your score” button is shareability. A 2-player asynchronous game where you literally cannot play without inviting a friend is an intrinsic social mechanic. The first is opt-in. The second is the experience itself.
What I saw at Words With Friends, and why “no UA budget” is the wrong story
Here’s where most retellings of this whole era get it wrong. At Zynga, watching Words With Friends operate, I saw the game running large paid UA campaigns, multiple millions per month. Words With Friends was not a “no UA budget” story. Neither, ultimately, was Draw Something post-acquisition.
What the virality did was raise LTVs to a level that justified that paid UA spend. Every player who installed brought one or two friends in within their first week.
Each acquired user was structurally worth more than they would be in a non-social game. That LTV uplift is what made the paid UA economics work: not virality replacing UA, but virality multiplying its return.
This is the part most “this app went viral with no UA budget” stories miss. The early spike was organic. The compounding business was paid UA scaled against virality-supported LTVs. Virality isn’t a substitute for paid UA, it’s the multiplier that makes paid UA profitable at scale.
The mechanic, asynchronous turn-based Scrabble against people you knew, forced re-engagement on a daily cadence. Every notification was a friend’s move, not a generic push. That re-engagement loop is what generated the LTVs that justified the marketing spend. The product did the work that made the marketing work.
That’s the bar most apps trying to “go viral” are missing. The product has to make the math work. See: concept testing before production for how to evaluate this before you’ve shipped.
Draw Something didn’t have the same long-term trajectory because the mechanic (bursty 2-player games) had less re-engagement gravity than Words With Friends’ multi-day rolling sessions. Mechanic durability matters as much as mechanic strength.
What does this actually mean for a 2026 launch?
Going viral isn’t a strategy. It’s an outcome. The right question in 2026 isn’t “how do we go viral”, it’s “what social mechanic does our app have, and how do we lean into it from day one.” Here’s the realistic playbook:
- Audit your app for intrinsic social mechanics. Be honest. If there’s none, accept it. You’re not in the viral business, you’re in the paid-UA business. See: how to scale mobile UA from $1K to $100K/day for what that looks like.
- If you have one, make it the spine of the experience, not an afterthought. The mechanic should be load-bearing. If your app works fine without it, you don’t have one, you have a feature.
- Don’t try to engineer “viral moments” via paid amplification. Paying influencers to make your app trend is just paid UA with a vanity metric attached. See: how to find and brief UGC creators for the version of paid amplification that actually works.
- Build the engineering foundation before you need it. Wilson’s three-year head start at OMGPOP wasn’t optional. A spike that hits an unprepared backend looks like a failed launch, not a successful one. See: how to use data to inform creative decisions.
- Optimize the invitation flow specifically. If your mechanic is intrinsic, the invitation/onboarding step is where most growth dies. Wilson’s team spent product engineering on reducing friction in invites. That’s the bar.
- The realistic 2026 launch. Paid UA + a clean intrinsic social mechanic + patience for compounding. Expect organics to take 6 months, not 6 weeks. See: creative velocity for mobile gaming for the cadence that makes this work.
- Plan for the longevity tail. Type A apps retain value for years after the press moves on. Type B apps flame out. If you’re investing for 5+ years, you have to be in Type A territory. See: the best framework for A/B testing ad creatives if you want to test which angle leans into your mechanic best.
- Stop asking “how do we go viral.” Start asking “what social mechanic do we have, and are we making it the most prominent thing about the experience.” That’s the only version of this question with a useful answer in 2026.
Key insight: The teams that try to recreate Draw Something’s launch in 2026 are usually trying to recreate the result. The teams that succeed are recreating the conditions, intrinsic social mechanic, deep engineering foundation, infrastructure ready when traction hits.
Frequently asked questions
Can a mobile app still go viral without paid acquisition in 2026?
Possible but rare, and only for apps with a strong intrinsic social mechanic. Most apps that “went viral” in 2009-2014 had multiplayer or social-by-design mechanics carrying the distribution.
In 2026, the App Store is roughly three times larger, attribution is harder under SKAdNetwork, and paid UA dominates the charts. An intrinsic mechanic is necessary but no longer sufficient.
What made Draw Something hit 50 million installs in 50 days?
An intrinsic 2-player social mechanic (the game required inviting a friend to play) combined with a 3-year engineering foundation that prevented downtime under spike load.
Wilson Kriegel, COO of OMGPOP at launch, attributes the success to the mechanic plus the team’s readiness for scale. They had also optimized invitation flow specifically. “Going viral” was the outcome of design choices, not luck.
Why don’t apps go viral the way they used to?
Three structural changes. The App Store grew from roughly 700,000 apps in 2012 to roughly 1.8 million in 2026.
iOS attribution shifted from deterministic IDFA to probabilistic SKAdNetwork. Distribution arbitrage on social networks (WhatsApp, SMS, Facebook) closed as platforms tightened reach. Even apps with the right mechanic face tougher distribution today.
Is Flappy Bird a useful case study for going viral?
Not as a playbook. Flappy Bird is the exception that proves the rule: it went viral through cultural happenstance, not designed mechanics.
You cannot engineer Flappy Bird’s success. Apps that copy the playbook build a strategy on luck. Flappy Bird also had no longevity tail; once removed, clones never matched the original.
What is the difference between viral and shareable?
Shareability is opt-in: a “share this” button users tap. An intrinsic social mechanic is the experience itself: a 2-player game you literally cannot play without inviting a friend. The first asks users to do extra work. The second is the work being done as a side effect of normal use.
How do you engineer virality for a mobile app?
You cannot engineer virality directly. You can design intrinsic social mechanics that increase the probability of viral spread when paired with paid UA and patience.
The 2026 playbook: audit for an intrinsic mechanic, make it the spine of the experience, avoid paid “viral moments,” build engineering foundation before you need it, optimize the invitation flow, and plan for the longevity tail.
Was Draw Something’s success replicable, or just luck?
Partly replicable. The intrinsic mechanic and engineering foundation are designable conditions. The cultural moment of February 2012, open distribution, deterministic attribution, Facebook arbitrage, is not replicable in 2026. Teams trying to recreate the result will fail. Teams recreating the conditions can give themselves a real chance.
Why did Draw Something not last when Words With Friends did?
Mechanic durability. Draw Something’s gameplay was bursty, quick rounds of drawing and guessing. Once the novelty faded, the re-engagement loop weakened. Words With Friends has multi-day rolling sessions and persistent relationships, which generate compounding re-engagement over years. Mechanic strength gets you to the spike. Mechanic durability keeps you alive afterward.
Related reading
