Most UA teams tighten their TikTok audiences when CPI creeps up. Custom lookalikes, stacked interests, narrow behavior layers. The instinct is right for Meta. On TikTok, it is backwards: every step toward tighter targeting raises CPI, not lowers it.
That is the headline finding buried in TikTok’s 2026 Performance Advertisers Report, drawn from 14,000 app install campaigns across 38 markets. Broad targeting delivers a median $1.58 CPI. Interest targeting climbs to $1.74. Custom-audience targeting reaches $1.96. Tighter costs more, every step of the way.
The reason is structural. On TikTok, creative is the targeting. The algorithm pairs each ad with the micro-audience most likely to install from it. When you narrow the audience manually, you constrain the signal the algorithm needs to do that matching. You fight your own optimization.
This is the most important thing in the report for anyone spending on TikTok in 2026. The rest of this article unpacks it alongside the headline numbers most teams talked about when the report dropped: CPI down 11% year-over-year to $1.72, Spark Ads 27% cheaper than standard in-feed, Value-Based Optimization now a third of all TikTok app install spend.
Page Contents
What is the average CPI on TikTok by app vertical in 2026?
| App Vertical | Median CPI (Global) | Median CPI (US Only) | YoY Change | Avg CVR (Click-to-Install) | Share of Total Spend |
|---|---|---|---|---|---|
| Casual Gaming | $1.09 | $1.48 | –14% | 28.3% | 22% |
| Midcore/Hardcore Gaming | $1.67 | $2.31 | –9% | 18.7% | 11% |
| Shopping / E-commerce | $1.54 | $2.08 | –12% | 22.1% | 18% |
| Food Delivery | $2.41 | $3.12 | –7% | 14.6% | 6% |
| Fintech / Neobanking | $3.84 | $5.21 | +3% | 9.2% | 8% |
| Health & Fitness | $2.18 | $2.87 | –11% | 16.4% | 7% |
| Entertainment / Streaming | $1.93 | $2.54 | –13% | 19.8% | 9% |
| Social / Dating | $2.62 | $3.47 | –6% | 12.9% | 5% |
| Utilities / Productivity | $1.38 | $1.79 | –16% | 24.5% | 4% |
| Education / EdTech | $1.82 | $2.43 | –10% | 17.1% | 4% |
What we see: Gaming pays the least, fintech the most, and every vertical except fintech has seen CPI decline year-over-year.
How do Spark Ads compare to standard in-feed ads for app installs on TikTok?
| Metric | Spark Ads (Median) | Standard In-Feed Ads (Median) | Difference | Statistical Confidence |
|---|---|---|---|---|
| CPI (Global) | $1.41 | $1.93 | –27% | 99% |
| CTR | 1.24% | 0.87% | +43% | 99% |
| CVR (Click-to-Install) | 23.6% | 17.8% | +33% | 98% |
| Day 1 Retention | 31.2% | 26.7% | +17% | 95% |
| Day 7 Retention | 14.8% | 11.3% | +31% | 96% |
| ROAS Day 7 (IAP apps) | 8.2% | 6.1% | +34% | 94% |
| ROAS Day 30 (IAP apps) | 19.7% | 14.9% | +32% | 93% |
| Avg Watch Time | 6.8s | 4.1s | +66% | 99% |
| 6-Second View Rate | 71.3% | 52.6% | +36% | 99% |
| Creative Fatigue (Days to 30% CPI increase) | 12.4 days | 8.1 days | +53% | 97% |
What we see: Spark Ads beat standard in-feed on every metric. Lower CPI, higher retention, longer fatigue window, 66% more watch time.
What are the best creative specifications for TikTok app install ads in 2026?
| Creative Element | Best Practice (Per TikTok Data) | Median CPI When Applied | Median CPI When Not Applied | Lift |
|---|---|---|---|---|
| Hook in first 1.3 seconds | Movement + text overlay + face | $1.49 | $2.61 | +43% install rate |
| Ad length | 15-21 seconds | $1.52 | $1.94 (longer) / $1.88 (shorter) | –18% CPI vs longer |
| Sound/Music | Trending sound + voiceover | $1.44 | $1.97 (no sound design) | –27% CPI |
| Text overlays | 3-5 text cards, 2-3 words each | $1.56 | $2.03 | –23% CPI |
| UGC-style vs polished | UGC with subtle branding | $1.38 | $1.81 | –24% CPI |
| CTA placement | End card + pinned comment CTA | $1.47 | $1.78 | –17% CPI |
| Aspect ratio | 9:16 full-screen native | $1.51 | $2.19 (non-native) | –31% CPI |
| Number of scene cuts | 5-8 cuts per 15s ad | $1.43 | $1.89 (fewer cuts) | –24% CPI |
| Creator/face on camera | Present in 60%+ of frames | $1.39 | $1.74 | –20% CPI |
| Color saturation | High saturation + warm tones | $1.55 | $1.71 | –9% CPI |
What we see: Every creative variable that wins on TikTok is anti-polished. UGC, face on camera, trending sound, hook in the first 1.3 seconds, minimal production value.
Need help scaling your mobile app growth? Talk to RocketShip HQ about how we apply these strategies for apps spending $50K+/month on UA.
How effective is each TikTok audience targeting method for app installs?
| Targeting Method | Median CPI | Median D7 ROAS (IAP) | Avg Audience Scale (US) | Recommended Use Case | Efficiency Rating |
|---|---|---|---|---|---|
| Broad (no targeting) | $1.58 | 7.4% | 150M+ | Scaling phase, high-volume creatives | ★★★★☆ |
| Interest-based | $1.74 | 8.1% | 30-80M | Initial testing, vertical alignment | ★★★☆☆ |
| Behavior-based | $1.82 | 9.3% | 15-40M | High-intent signals, re-engagement | ★★★★☆ |
| Custom Audience (1P data) | $1.96 | 11.7% | 500K-5M | Retargeting, lookalike seeds | ★★★★★ |
| Lookalike (1% value-based) | $1.63 | 10.8% | 5-15M | Scaling quality users | ★★★★★ |
| Lookalike (5% broad) | $1.51 | 7.9% | 20-50M | Volume at moderate quality | ★★★★☆ |
| Smart Targeting (auto) | $1.55 | 8.6% | Dynamic | Low-resource teams, broad testing | ★★★★☆ |
| Hashtag Interaction | $1.91 | 7.2% | 10-30M | Cultural moment targeting | ★★☆☆☆ |
| Creator Audience (Spark) | $1.44 | 9.8% | 2-20M | Creator-led campaigns | ★★★★★ |
| VBO Optimized (auto target) | $1.69 | 12.4% | Dynamic | Mature pixel, quality focus | ★★★★★ |
What we see: Tighter manual targeting costs more per install, period. Broad wins on CPI. Value-based 1% lookalikes win on CPI and ROAS together. This is the table the title is built on.
What are TikTok attribution windows and how do they affect reported performance?
| Attribution Window | Reported CPI | Reported D7 ROAS | Install Volume Captured | Recommended For | Key Consideration |
|---|---|---|---|---|---|
| 1-day click | $1.93 | 9.8% | 62% | Conservative measurement | Misses view-through impact |
| 7-day click | $1.72 | 8.4% | 78% | Balanced default | Industry standard comparison |
| 28-day click | $1.54 | 7.1% | 89% | Long-funnel apps (fintech) | Potential over-attribution |
| 1-day click + 1-day view | $1.41 | 7.6% | 91% | Full-funnel understanding | View-throughs inflate volume |
| 7-day click + 1-day view | $1.38 | 7.2% | 94% | TikTok recommended default | Most common setting |
| SAN (Self-Attributing Network) | $1.72 | 8.4% | 78% | TikTok Events Manager | Compare with MMP data |
| MMP Last-Touch | $2.14 | 10.1% | 58% | Cross-network comparison | Under-counts TikTok assist |
| Incrementality (Ghost Bids) | $2.87 | 13.6% | 41% | True causal impact | Gold standard, complex setup |
What we see: Self-attributed windows over-count TikTok’s impact. Incrementality testing reveals the true causal CPI is roughly 2x the most-generous self-attributed number.
What we learn
On TikTok, creative is the targeting.
You do not reach users by drawing a box around them in the audience builder. You reach them through the creative you put in the feed. Each ad in your portfolio speaks to a specific micro-audience, and the algorithm does the actual matching.
Tighter manual targeting fights that system. When you narrow the audience, you constrain the signal the algorithm needs to match creatives to the right users. The result is higher CPI and lower creative coverage, not cheaper installs.
This applies to gaming and subscription apps alike. Broad outperforms narrow on TikTok across verticals, not because targeting does not matter, but because on TikTok the targeting lives in the creative, not in the audience builder.
Most teams spending on TikTok make three mistakes at this point.
First, they repurpose organic content directly, assuming anything that went viral on TikTok will work as a performance ad. Organic content optimizes for engagement. Direct response needs its own structure: hook, value proposition, reason to act, CTA. UGC-style is the format. DR frameworks are the content. Using Spark Ads to turn a DR-built organic post into a paid ad is the right move. Using Spark Ads to promote an entertainment post that never had a conversion job to do is not.
Second, they do not invest in creative diversity. One hero ad scaled across an audience is not a portfolio. A portfolio is 8 to 12 distinct hooks, each speaking to a different micro-audience inside the same app. The problem-aware user, the intent-driven user, the lapsed churned user: same product, different doors in.
Third, they assume tighter targeting will reduce CPI. The data says the opposite. Broad is the scaling default on TikTok once you have a creative portfolio to feed it.
Campaign structures that make sense
Campaign structure on TikTok is secondary to two things: what you optimize for, and how much creative diversity you feed the algorithm. Get those two right and most structural choices will work. Get them wrong and no structure will save you.
- Optimize for the most downstream metric your app supports. For IAP gaming, that is Day 7 or Day 30 ROAS. For subscription apps, trial start or first-payment events. The algorithm optimizes toward whatever signal you give it. Cheap installs do not matter. Quality users do.
- Start broad, with creative diversity. Do not narrow the targeting to “help” the algorithm. Feed it 8 to 12 creative hooks covering different micro-audiences and let it find the segments each one fits.
- Migrate to VBO once you cross 500 post-install value events per week at $50K+ monthly TikTok spend. VBO’s 12.4% D7 ROAS versus 7.4% on broad is the biggest quality lift in the report.
- Refresh weekly. Spark Ads fatigue at 12.4 days on average, standard in-feed at 8.1 days. Any refresh cadence longer than 7 days leaves performance on the table.
- Use Spark Ads for 40 to 60 percent of your mix, but feed them DR-built content, not repurposed organic. The Spark format inherits credibility from the source post. It does not rescue content that was never built to convert.
Frequently Asked Questions
Why does broad targeting outperform tighter targeting on TikTok?
Because on TikTok, creative is the targeting. The algorithm matches each creative in your ad group to the micro-audience most likely to respond. When you narrow the audience manually, you constrain the signal the algorithm needs to do that matching. Broad targeting plus a diverse creative portfolio gives the algorithm maximum room to find the right users for each hook. Tighter targeting fights it and raises CPI.
Does this apply to subscription apps, or only to gaming?
Both. The mechanism is the same regardless of vertical: creative carries the targeting signal, so broadening the audience + diversifying the creative portfolio outperforms narrow audiences on any app type. The difference is in what downstream metric you optimize for (trial start and first payment for subscription, D7/D30 ROAS for IAP gaming).
What is a good CPI on TikTok for app install campaigns in 2026?
The global median CPI on TikTok for app installs in 2026 is $1.72, down 11% from 2025. “Good” varies dramatically by vertical: casual gaming averages $1.09, fintech apps see $3.84. In the US specifically, expect to pay 30-40% more than global medians across all verticals.
Are TikTok Spark Ads better than regular ads for app installs?
Yes, decisively. Spark Ads deliver 27% lower CPI ($1.41 vs $1.93), 33% higher click-to-install conversion rates, and 31% better Day 7 retention than standard in-feed ads across TikTok’s 2026 benchmark dataset. They also resist creative fatigue 53% longer (12.4 days vs 8.1 days). The caveat: Spark Ads work because they inherit the credibility of the source post, so the source post needs to be built for direct response, not just engagement.
How long should TikTok ads be for app installs?
TikTok’s 2026 data shows 15 to 21 second ads deliver the lowest median CPI at $1.52, outperforming both shorter (under 10s at $1.88) and longer (over 30s at $1.94) formats. The critical window is the first 1.3 seconds, where movement, text overlay, and a face on camera combine to create the strongest hooks. See the ideal TikTok ad length guide for more.
What attribution window should I use for TikTok app install campaigns?
TikTok recommends 7-day click plus 1-day view as the default, which captures 94% of installs at a reported CPI of $1.38. Incrementality testing shows the true causal CPI is closer to $2.87, meaning TikTok’s self-attributed numbers over-count by roughly 52%. Use the 7-day click window ($1.72 CPI, 78% capture rate) for balanced cross-network comparisons, and run quarterly incrementality tests to calibrate.
How many ad creatives do I need for TikTok app install campaigns?
TikTok’s 2026 report indicates ad groups with 5 to 7 active creatives achieve 22% lower CPI than those with fewer than 3. Creative fatigue sets in after 8 to 12 days on average, so you need a production pipeline capable of refreshing 30-50% of your creative portfolio weekly at scale. See how many creatives you need per TikTok ad group for deeper guidance.
How does TikTok CPI compare to Meta and Google for app installs in 2026?
TikTok’s global median CPI of $1.72 positions it 15-20% below Meta’s app install CPI (approximately $2.05 to $2.15 per Adjust’s Q1 2026 benchmarks) and roughly on par with Google App Campaigns ($1.60 to $1.85 depending on format). TikTok’s Day 7 retention tends to run 10-15% lower than Meta for most verticals, so blended cost-per-retained-user metrics often equalize across platforms.
What is TikTok’s Value-Based Optimization and should I use it?
VBO is TikTok’s campaign type that optimizes toward post-install revenue events rather than install volume. It now accounts for 34% of all app install spend on TikTok, up from 18% in 2024. Use it if you have at least 500 value events per week and $50K+ monthly spend on TikTok. VBO campaigns show 12.4% D7 ROAS versus 7.4% for broad targeting, a 68% improvement in early revenue quality.
Looking to scale your mobile app growth with performance creative that delivers results? Talk to RocketShip HQ to learn how our frameworks can work for your app.
Not ready yet? Get strategies and tips from the leading edge of mobile growth in a generative AI world: subscribe to our newsletter.
Related Reading
