About APKHerd
APKHerd is a small testing lab and publication focused on one narrow problem set: how files actually move between phones and computers, how Android storage fills up and gets cleaned safely, and how app cloning and backup tools behave once you push them past the marketing screenshots. We do not cover general tech news, we do not review phones, and we do not publish download links for modified software. Every guide on this site grows out of a test we ran, a failure we reproduced, or a log we read.
The site exists because file transfer and storage advice online is unusually bad. Most articles about LocalSend, Quick Share, app cloners or cache cleaners are rewritten from other articles, and the numbers in them — transfer speeds, space recovered, battery impact — are invented. Ours are measured, and where a measurement depends on conditions we cannot control, we say so instead of printing a confident figure.
How We Test Transfer Tools
Every transfer tool we write about is run through the same controlled file sets so results stay comparable across reviews and across time:
- A photo set — 500 JPEG and HEIC images totalling roughly 2.3 GB. Many small files expose per-file overhead, which is where most tools quietly lose their speed.
- A video set — three large MP4 files between 1.5 GB and 4 GB. Large sequential files reveal a tool’s real throughput ceiling.
- A mixed archive set — a single 5 GB ZIP plus a folder tree of about 8,000 small documents, APKs and audio files. This is the workload that breaks progress bars and resume logic.
Each set is transferred in both directions between devices, three times per configuration, and we record the median time, the negotiated link rate where the tool or OS exposes it, and whether the receiving copy verifies against checksums of the originals. A transfer that arrives corrupted counts as a failure even if it was fast.
Radio conditions are part of the test, not background noise. We test on a congested 2.4 GHz network, a clean 5 GHz network, Wi-Fi Direct peer-to-peer connections with no router involved, and Bluetooth for the small-file baseline. Router model, band, channel width and distance are noted for every published figure, because a Wi-Fi Direct result from the same room tells you almost nothing about performance through two walls.
Our device matrix is deliberately unglamorous: a current Pixel, a mid-range Galaxy, a budget Redmi-class phone with 4 GB of RAM, an older Android 11 device, a Windows 11 laptop, a Mac with Apple silicon, and a Linux desktop. Tools that only work well on flagship hardware get marked down for it — most people moving files between devices are not doing it with flagship hardware.
How We Audit Cloners, Cleaners and Backup Tools
App cloners and storage cleaners ask for the most sensitive permissions Android offers, so our audit process is stricter than our transfer testing. Before writing about one, we record the full permission list it requests, watch which of those permissions it exercises during a scripted workload, and inspect its network traffic for unexpected endpoints. A cleaner that uploads your file index to an ad network’s server gets described as doing exactly that, however well it cleans.
For cleaning tools we use a standard “dirty phone” state: a device seeded with a known quantity of cache, leftover folders from uninstalled apps, duplicate downloads and thumbnail data. We measure what the tool claims to free, what it actually frees, and — more important — what it deletes that it should not. Tools that clear app data while claiming to clear cache fail our review outright.
Backup utilities are tested by destruction. We back up a configured test phone, wipe it, restore, and diff the result against a checklist of apps, accounts, message history, call logs, app data and settings. A backup tool that cannot restore its own backup scores zero on the only metric that matters, and we publish restore failures with the same prominence as restore successes.
Editorial Standards
- Hands-on or labelled. If we have not run a tool ourselves, the article says so and explains what the assessment is based on.
- No paid placements. Developers cannot pay for coverage, scores or links. If a developer provides a licence for testing, that is disclosed in the article.
- Corrections stay visible. When a guide goes stale — an app changes its interface, an update alters behaviour — we revise the article, note the revision date, and keep the outdated claim out rather than silently rewriting history.
- Downloads go to the source. Where we point readers to software, the link goes to the developer’s own site or official store listing, never to a mirror or a repackaged file.
Questions about a test, a disagreement with a result, or a correction we should make are exactly what our contact page is for. Reproducible counter-evidence is the fastest way to change anything published here.