Payload, Compression & Battery Efficiency — Mobile API Best Practices

Radio state machines keep the cellular radio in a high-power state for seconds after each transfer, so many small requests cost more battery than one larger request. Batch analytics and non-urgent sync, use Brotli/gzip, avoid oversized JSON (IDs instead of nested copies where clients cache), paginate with cursors, send image URLs sized for the device (via an image CDN), and use HTTP/2 or HTTP/3 to multiplex on one connection (HTTP/3/QUIC also survives network changes via connection migration).

Efficiency techniques

Technique	Saves	Notes
Brotli / gzip	60–90% of JSON bytes	Negotiate via Accept-Encoding
Sparse fields	Unused fields	fields= or GraphQL selection
Conditional GET (ETag)	Whole body on 304	Great for config/catalogs
Batching non-urgent calls	Radio wake-ups	Analytics, logs, prefetch on Wi-Fi/charging
Image CDN variants	Megabytes per screen	w=, dpr=, format=avif/webp
HTTP/2 / HTTP/3	Handshakes, head-of-line blocking	QUIC connection migration across networks
Protobuf / binary	Some bytes + parse CPU	Usually less impact than compression + trimming

// Defer non-urgent uploads until unmetered + charging; the OS batches work.
val constraints = Constraints.Builder()
    .setRequiredNetworkType(NetworkType.UNMETERED)
    .setRequiresCharging(true)
    .build()
val upload = OneTimeWorkRequestBuilder<AnalyticsUploadWorker>()
    .setConstraints(constraints)
    .setBackoffCriteria(BackoffPolicy.EXPONENTIAL, 30, TimeUnit.SECONDS)
    .build()
WorkManager.getInstance(context)
    .enqueueUniqueWork("analytics-upload", ExistingWorkPolicy.KEEP, upload)

// Background-capable, discretionary transfer: the system picks a good time.
let config = URLSessionConfiguration.background(withIdentifier: "com.example.sync")
config.isDiscretionary = true            // wait for Wi-Fi / power when possible
config.sessionSendsLaunchEvents = true
let session = URLSession(configuration: config, delegate: self, delegateQueue: nil)
session.uploadTask(with: request, fromFile: batchFileURL).resume()

Common pitfalls

Sending analytics events one request each.
Full-resolution images to small screens.
Polling every few seconds in the foreground.

Key points

Batch and defer non-urgent work
Compress and trim payloads
Let the OS schedule background transfers

Further reading: Android: Persistent work (WorkManager)

Practice questions

Why can many small requests drain more battery than one large request?

Answer

The cellular radio stays in a high-power state for a tail period after each transfer; frequent small transfers keep re-waking it, while batching amortizes the wake-up.

What does HTTP/3 offer mobile clients specifically?

Answer

QUIC's connection migration lets a connection survive IP changes (Wi-Fi ↔ cellular) and avoids TCP head-of-line blocking, improving latency on lossy networks.