Practical guide: VOD cost optimization
In this tutorial, you learn how to optimize ApsaraVideo VOD costs by configuring transcoding templates, managing storage tiers, and enabling adaptive bitrate streaming. By the end, you will be able to systematically reduce your VoD bill while maintaining playback quality, using a single break-even formula to evaluate every transcoding decision.
Prerequisites
ApsaraVideo VOD service activated
RAM permissions to manage transcoding templates, storage strategies, and CDN configurations
Domain name configured and CDN acceleration enabled for your VOD content
Familiarity with VoD billing models (pay-by-traffic or pay-by-bandwidth)
(Recommended) Access to VoD console at VoD console
VoD pipeline and cost centers
Short videos, live replay, online education, and micro-drama overseas are opening new growth curves for many enterprises. As play minutes and UV climb, the VoD pipeline contains significant cost optimization opportunities through smarter configuration.
The VoD pipeline flows as follows: upload → transcoding → storage → CDN distribution → playback. Three cost centers dominate the bill:
Transcoding: One-time cost per video, billed per minute of transcode.
Storage: Ongoing cost per GB stored, varies by storage tier.
Traffic (CDN): Per-GB cost for each playback delivered.
The transcoding node is the leverage point of the entire pipeline. It occurs only once as a fixed cost, but the output bitrate simultaneously affects downstream storage and traffic costs. Every 1% bitrate reduction cuts both the transcode output file size and per-playback traffic by 1%. Higher playback volume multiplies the leverage. This is why getting the transcode template right first, then reviewing the traffic bill, is the highest-ROI sequence.
Traffic consumption and traffic cost are not linearly related. Actual cost depends on the CDN tiered pricing strategy.
Three cost-reduction tools at a glance
The following table summarizes the cost-reduction capabilities available at each pipeline stage. Each tool is explained in detail in the implementation sections that follow.
Pipeline stage | Available cost-reduction capabilities |
Upload → Transcoding | Narrowband HD 1.0/2.0, H.265 encoding, per-scenario template differentiation |
Transcoding → Storage | Transcoded source file cool-down, tiered lifecycle management |
Storage → CDN | Multi-bitrate ABR, resolution-tier optimization |
CDN → Playback | Player-side adaptive switching |
Narrowband HD 1.0/2.0 reduces output bitrate at equivalent quality through scene-aware encoding. Version 2.0 additionally repairs compression artifacts such as ringing and flickering.
H.265/HEVC reduces bitrate compared to H.264 at the same resolution. Terminal compatibility must be evaluated before adoption.
Storage tiering migrates assets from Standard to lower-cost tiers (IA, Archive, Cold Archive) based on age and access patterns.
Adaptive bitrate streaming (ABR) packages multiple bitrate streams into an HLS stream and switches based on network conditions.
Recommended governance priority order is Storage → Transcoding → Traffic → Playback, from highest to lowest cost proportion.
New business: building optimal configuration from zero
Step 1: Choose transcode templates by business type
Different business scenarios have different quality requirements and terminal distributions. The following table provides scenario-specific recommended configurations:
Business scenario | Recommended encoding | Resolution tiers | Output format | ABR |
Micro-drama (2–5 min/episode, overseas) | H.265 (note: compatibility risk, retain H.264 fallback) | 540P / 720P / 1080P | HLS / MP4 | Recommended |
Online education (teaching/landscape, high text density) | Narrowband HD 1.0 (H.264) | 540P / 720P / 1080P | HLS | Optional |
Film & TV (series/long video 30 min+) | Narrowband HD 1.0 (H.264) | 720P / 1080P / 2K/4K | HLS | Recommended |
Three configuration tips:
Do not exceed 4 clarity tiers. Too many tiers slow down ABR decisions and increase storage costs.
If ABR seamless switching is enabled, set GOP to 5s or 10s. No GOP requirement if ABR is not enabled.
Phase your rollout: start with an H.264 multi-tier baseline, then overlay Narrowband HD, then overlay H.265. Harvest gains incrementally.
Step 2: Configure transcode template groups in the console
Go to the VoD console. In the left-side navigation pane, choose Transcode Template Group.
Click Add Transcode Template Group, and configure the following settings:
Template Name: Name per the convention above for easy management.
Transcoding Type (Transcode Type): Select Narrowband HD 2.0, Narrowband HD 1.0, or Standard Transcode.
Encoding Format: H.265 for popularity-based transcoding, or H.264 as fallback.
Container Format: HLS.
Resolution Tiers: Configure per the resolution tiers in the table above. Click Add Template to add resolution tiers. Each tier can have an independent bitrate cap.
Key Frame Interval (GOP): If ABR is enabled, set to 5s or 10s. For other scenarios, the default is fine.
NoteGOP supports two modes: maximum time interval between key frames in seconds (recommended 2–10 s) and maximum frame count between key frames (recommended 50–300). The default is frame mode with a value of 250.
Set the template group as the default. Newly uploaded media assets will automatically apply it.
You should see the new template group in the transcode template group list.
Expected outcome: With an H.265 + ABR configuration, traffic bandwidth spend drops 24–45%. Transcoding-side spend increases because the H.265 unit price is 5x that of standard H.264. The net benefit depends on CDN reduction and playback minute distribution, with an estimated net saving of 8–25% off the total bill.
Step 3: Player integration
Android / iOS: Integrate ApsaraVideo Player SDK and enable
DEFINITION_AUTOadaptive mode.Web: Use AliPlayer Web SDK.
Expected outcome: ABR adaptation brings a 3–5% bandwidth reduction. Multi-tier clarity increases transcoding count (N tiers = N × transcode cost). Balance this with the "≤ 4 tiers" constraint from Step 1.
Existing business: audit waste, govern precisely
Step 1: Storage cost reduction
Migrate the original source file to IA, Archive, or Cold Archive storage after a video is uploaded and transcoded. This is a stable, near-zero-risk storage cost-reduction action and should be the first step in storage governance.
Why post-transcode source files are the natural target for cool-down
From upload to playback, video naturally splits into two asset types:
Asset type | Who uses it | Access frequency | Recommended storage |
Source file (uploaded original) | Almost exclusively read once during transcoding, editing, or AI processing | Very low (one-time consumption) | IA / Archive / Cold Archive |
Transcode output (HLS/MP4 multi-tier artifacts) | Player, CDN origin pull | High (continuously played) | Standard storage |
Once transcoding succeeds, the source file's mission is essentially complete. Keeping it in Standard storage means paying per GB monthly for data that will rarely be read again. Moving it to IA, Archive, or Cold Archive reduces the per-GB monthly cost by 30%, 70%, or 85% respectively, while transcode outputs remain in Standard storage and playback performance is unaffected.
Configure storage strategy
Go to the VoD console. In the left-side navigation pane, choose Storage Strategy.
Click Create Strategy and select Media Asset Cool-down as the Strategy Type.
Fill in each field according to the table below:
Field
Recommended value
Notes
Policy Name
Source file to IAName per business for multi-strategy management
Strategy Type
Media Asset Cool-down
Media Time Condition
Select the time range for target media assets
Determines which media assets are affected
Storage Type
Select the current storage type of source files
Select "All" to cover uploaded source files
Target Storage Type
IA Storage / Archive Storage / Cold Archive Storage
IA is recommended as a starting point. Add Archive or Cold Archive later. Must ensure target asset storage duration ≥ minimum for the tier (IA: 30 days / Archive: 60 days / Cold Archive: 180 days)
Effective Scope
Source File Only (Scope=SourceFile)
Critical setting: only cool down source files. Keep transcode outputs in Standard storage for playback
Start Time
Immediate / specific date
Recommend scheduling a non-peak time to begin
Priority
Default or custom (higher number = higher priority)
Explicitly control execution order when multiple strategies coexist
Status
Enabled
Effective once created
Click Submit to save. After the strategy is dispatched, the system asynchronously executes cool-down actions at the set processing pace (daily limit of 50,000 items).
Verify the strategy is active: go to the storage strategy list and confirm the strategy status shows Enabled. Check the processed asset count to ensure migration is proceeding.
(Optional) Create overlay strategies to form a tiered cool-down pipeline for long-lifecycle assets. Repeat steps 2–4, adjusting the Media Time Condition and Target Storage Type to create "180 days → Archive" and "365 days → Cold Archive" strategies.
① Maximum 5 strategies per strategy type, 10 total. ② Daily processing limit for all strategies: 50,000 items. ③ When multiple strategies match, the one with the higher priority number takes precedence. ④ Trigger time must be ≥ the minimum storage duration of the target storage tier (IA 30 days / Archive 60 days / Cold Archive 180 days), otherwise validation will reject. ⑤ Single files smaller than 64 KB are not cooled down (billing at 64 KB makes it meaningless).
Step 2: Transcoding cost reduction
Whether you are deciding "should I adopt Narrowband HD?" or "which version to pick between 1.0 and 2.0?", the answer can be computed with the same formula.
Unit price baseline
The following table shows transcoding unit prices and compression capabilities. Prices apply to the China mainland region, for H.264 encoding, billed per minute, based on the 2026-03-20 published price list.
Transcoding type | LD (360P) | SD (480P) | HD (1080P) | 2K | 4K | Multiplier | Typical compression depth | Transcode time |
Standard H.264 | 0.0217 | 0.0326 | 0.0651 | 0.14 | 0.28 | 1x | 20–30% reduction | ~1x |
Narrowband HD 1.0 | 0.0651 | 0.0978 | 0.1953 | — | — | 3x | 30–40% reduction | ~1.5x |
Narrowband HD 2.0 | 0.3255 | 0.4885 | 0.9765 | 2.10 | 4.20 | 15x | 50–60% reduction | ~4–6x |
Unit: CNY/minute (billed per second, settled hourly). Narrowband HD 1.0 does not support 2K or 4K. Fall back to standard encoding for those resolutions.
Quick decision table
The following table provides recommendations for a 1080P HD, 5-minute video with a 90-day lifecycle. Thresholds are calculated based on the parameters 1080P / 5 min / 2.6 Mbps / 0.21 CNY/GB.
Average daily plays per video | Recommended path | Reason |
< 1 play/day | No transcoding / Standard low tier | Long-tail content. Transcoding increment will never be recovered |
1–3 plays/day | Standard H.264, 30% reduction | Lowest cost path. 90–270 cumulative plays in 90 days |
~2 plays/day | Narrowband HD 1.0 | ~180 cumulative plays in 90 days. Exceeds the ~163-play break-even threshold |
~9 plays/day | Narrowband HD 2.0 | ~810 cumulative plays in 90 days. Exceeds the ~814-play break-even threshold |
10 plays/day | Narrowband HD 2.0 + H.265 | 900+ cumulative plays. Exceeds the ~814-play threshold with additional H.265 savings |
Break-even formula
Transcoding is a one-time investment while traffic savings accumulate per playback. As long as there are enough plays, the cumulative traffic savings will exceed the transcoding cost. N_min is the break-even play count.
The formula is the same regardless of which two paths you are comparing:
N_min = ΔP_tc / (ΔR × 0.00732 × P_cdn)N_min: Break-even play count. Cumulative plays exceeding this number make the more expensive transcoding worthwhile.
ΔP_tc: Unit price difference between the two transcoding paths (CNY/minute).
ΔR: Output bitrate difference between the two paths (Mbps).
0.00732: Traffic conversion constant. Calculated as 60 seconds/minute ÷ 8 bit/Byte ÷ 1024 MB/GB. Converts Mbps bitrate × play duration to GB traffic.
P_cdn: Traffic unit price (CNY/GB).
The video duration cancels out during simplification. N_min is independent of video duration and depends only on unit price difference and bitrate difference.
Practical calculation
Case A — Overseas region
Conditions: 1080P source at 2.6 Mbps, 3-minute duration, 786 cumulative plays. Traffic at 0.58 CNY/GB. Narrowband HD 1.0 H.264 HD (overseas region) at 2.62 CNY/min.
Option | Output bitrate | Traffic cost (786 plays) (billed by traffic) | Transcode cost (3 min) | Total cost | vs. no transcoding |
No transcoding, direct distribution | 2.6 Mbps | 26.19 CNY | 0 | 26.19 CNY | Baseline |
Narrowband HD 1.0, 30% reduction | 1.82 Mbps | 18.33 CNY | 7.86 CNY | 26.19 CNY | ±0% |
Formula verification: N_min = 0.3639 / (0.78 × 0.00732 × 0.081) ≈ 786 plays. This is the break-even point for Narrowband HD 1.0 with 30% reduction. Below this number, transcoding is not worthwhile.
The overseas break-even point is higher than domestic because the overseas traffic unit price (0.58 CNY/GB) is relatively lower, while the Narrowband HD unit price in overseas regions (2.62 CNY/min) is significantly higher than domestic.
Note: Upgrading to Narrowband HD 2.0 (unit price approximately 5x that of 1.0) raises the transcode fee to approximately 39.3 CNY. The total cost exceeds the no-transcoding baseline by 120%. Decide whether to adopt 2.0 based on actual experience requirements and cost considerations.
Case B — China mainland (50–100 TB tier, 0.21 CNY/GB)
Conditions: 1080P source at 2.6 Mbps, 5-minute duration. CDN at 0.21 CNY/GB. The table compares the impact of transcoding unit price differences on N_min.
Path | Output bitrate | Unit price (CNY/min) | Transcode cost (5 min) | Per-play CDN saving | N_min |
Narrowband HD 1.0, 30% reduction | 1.82 Mbps | 0.1953 | 0.977 CNY | 0.0060 CNY saved | ~163 plays |
Narrowband HD 2.0, 30% reduction | 1.82 Mbps | 0.9765 | 4.883 CNY | 0.0060 CNY saved | ~814 plays |
Comparing Narrowband HD 1.0 vs. 2.0, version 1.0 needs 163 plays to recover transcoding cost while 2.0 needs 814 plays. The real value of Narrowband HD lies in better quality at the same bitrate. If actual compression exceeds 30% (official: 1.0 achieves 20–40%), N_min will be even lower.
Five common misconceptions
"Narrowband HD 2.0 is always more cost-effective than 1.0" — Videos with fewer than 5 daily plays on 2.0 actually lose money. The transcode fee is 5x higher while traffic saves only 12% more.
"More bitrate reduction always saves more" — Increasing reduction from 30% to 60% saves 33% more on traffic, but transcode price jumps from 3x to 15x, requiring 3.3x more plays to break even.
"Micro-drama scenario" and "general scenario" modes swapped — Micro-drama mode is optimized for landscape and high text density. General mode is for portrait and rich visuals. Choosing the wrong mode results in 5–10% higher bitrate with degraded quality.
"Apply domestic N_min directly to overseas" — Overseas traffic unit prices differ from domestic. Break-even points are completely different and must be calculated per region.
"Storage product cost can be ignored" — A single video is negligible, but multi-tier ABR multiplies by 4–6 tiers. The annual storage cost for long videos can reach 20–30% of transcoding fees.
Troubleshooting
Issue | Possible cause | Solution |
Storage strategy is not triggering migration | Media Time Condition time range does not match any assets, or target asset storage duration is shorter than the minimum for the target tier | Verify the Media Time Condition covers the intended assets. Confirm target asset storage duration meets the tier minimum (IA: 30 days / Archive: 60 days / Cold Archive: 180 days) |
Storage strategy daily limit reached | More than 50,000 assets match the strategy on a given day | Reduce the scope of the strategy or stagger the Media Time Condition across multiple strategies |
H.265 transcode output fails to play on some terminals | Terminal does not support H.265 hardware or software decoding | Check terminal distribution before enabling H.265. Retain an H.264 fallback stream. For more information, see the H.265 compatibility risk note in Appendix B |
ABR stream switching causes stuttering | GOP interval does not match the ABR slice duration | Set GOP to 5s or 10s, matching the HLS slice duration. Ensure all bitrate tiers use the same GOP setting |
Transcoding template group not applied to new uploads | Template group is not set as the default | Go to the VoD console, open the transcode template group list, and set the desired group as the default |
Appendix A: Glossary
Term | Full name / meaning |
Narrowband HD | ApsaraVideo VOD's proprietary transcoding technology. Uses scene-aware encoding to reduce output bitrate at equivalent quality. Version 1.0 is suitable for most business scenarios (30–40% reduction). Version 2.0 overlays image quality repair on top of 1.0, at the cost of higher transcoding fees and longer transcoding time |
ABR (Adaptive Bitrate) | Multi-bitrate adaptive streaming. Packages multiple clarity tiers into a single HLS stream. The player switches in real time based on network conditions — higher quality on good networks, smooth playback on weak networks |
HLS | HTTP Live Streaming. An Apple-defined streaming protocol that slices video into a series of .ts segments and one .m3u8 manifest. ApsaraVideo VOD's default packaging format |
VoD | ApsaraVideo VOD, Alibaba Cloud's video on demand service |
GOP (Group of Pictures) | Key frame interval. Shorter GOP provides more flexible seeking and multi-bitrate switching but slightly reduces compression efficiency. ABR recommends GOP equal to slice duration (for example, 5s) |
Hardware decoding / Software decoding | Two video decoding methods. Hardware decoding uses a GPU or dedicated decoder chip for low CPU usage and low power consumption. Software decoding uses the CPU, which can result in higher CPU load and device heating |
Appendix B: Document usage notes
Regional differences: Case A in this document is based on the overseas AP1 region. Case B is based on the China mainland 50–100 TB traffic tier. Other overseas regions (AP2/AP3/EU/NA/ME) have different unit prices. Substitute your region's prices into the formula.
Prices subject to official adjustment: All unit prices (Narrowband HD, traffic tiered pricing, storage tiered pricing) are subject to the Basic service billing page. Prices quoted are from the 2026-03-20 published price list and may be adjusted or new discount packages introduced.
Delete strategy is irreversible: Any "delete media asset" action cannot be undone. Before going live, ensure sufficient canary release, manual spot checks, and business stakeholder confirmation. It is recommended to validate with "cool to Cold Archive" (soft delete) before considering hard delete.
H.265 compatibility risk: H.265/HEVC has incomplete coverage or only software decoding support on older Android (< 5.0), some domestic browsers, and HarmonyOS NEXT. Assess terminal distribution before enabling and retain H.264 as a fallback stream to avoid widespread playback failures.
Case numbers are illustrative only: The N_min break-even points in this document's tables are calculated based on specific parameters (1080P / 5 min / 2.6 Mbps) and do not represent actual customer business data. Use the formula with your own business variables as the basis for decision-making.
Conclusion
VoD cost governance is not a one-size-fits-all project but a series of refined decisions layered on top of each other. The core logic is simple: every cent of transcoding investment must be earned back through sufficient playback count on traffic savings.
With the formula N_min = ΔP_tc / (ΔR × 0.00732 × P_cdn), you can independently determine the break-even point for any transcoding path without relying on experience-based guessing.
Recommended governance priority: Storage cool-down (immediate impact) → Transcode template differentiation (per-scenario breakdown) → ABR adaptation (client-side harvesting). Start by saving on storage costs, then precisely calculate transcoding paths using the formula, and finally make fine-grained optimizations on the player side.
Unit price data sources: VOD Basic Service Billing (2026-03-20 edition) and Narrowband HD 1.0/2.0.