To run a fair mobile network speed test, keep the phone, location, test tool and comparison conditions consistent, turn off Wi-Fi, confirm the data SIM, and repeat the measurement. Record download, upload and latency together with failed attempts. One impressive 5G screenshot cannot tell you which network will work best throughout the day.
This guide shows how to compare Jio, Airtel, Vi or BSNL on your own phone, interpret ping, jitter and packet loss, and distinguish a weak connection from a misleading test. The aim is a useful personal comparison—not a nationwide operator ranking or a laboratory certification.
Sources checked on 18 November 2026. App features and labels can change. All numerical examples in this article are illustrative; Gizmo Times has not measured the operator performance represented by those examples.
A fair mobile speed test in seven steps
| Step | What to do |
|---|---|
| 1. State the question | Decide whether you are comparing SIMs, investigating an evening slowdown, or checking one troublesome room |
| 2. Prepare the connection | Confirm plan validity and allowance, turn Wi-Fi off, and select the intended data SIM |
| 3. Keep conditions consistent | Use the same handset, position, tool and comparable time window |
| 4. Record the endpoint | Note the test server or destination and whether it changes |
| 5. Run a small repeated set | Let each test finish; record errors and interruptions rather than discarding them |
| 6. Read more than download speed | Compare uploads, latency and any reported jitter, loss or loaded latency |
| 7. Check a real task | Try the document upload, video call or other activity that led you to investigate |
First decide what you want to compare
A test of the same SIM in two rooms asks a different question from a test of two SIMs in one room. Changing the handset, location, network mode and operator together can reveal that the overall setups differ, but it cannot identify which change caused the result.
For example, write “compare mobile data on SIM A and SIM B at my desk between 7 and 8 p.m.” before starting. If your question is whether an imported phone performs differently, keep the SIM and place fixed instead. Use a separate test series for each question.
An ordinary speed test measures an end-to-end path between your device and the test destination. The result is affected by more than the radio signal: the handset, network conditions, routing and destination all contribute. Treat it as evidence about that tested path, not a direct measurement of a mobile tower’s maximum capacity.
Prepare the phone and choose your test tool
Confirm that you are measuring mobile data
Turn Wi-Fi off for the mobile comparison and verify that the intended SIM supplies data. Opening a speed-test app on a phone connected to home broadband measures the Wi-Fi/broadband path instead of establishing the mobile operator’s performance.
On a dual-SIM iPhone, Apple’s Dual SIM guidance explains that one number at a time supplies cellular data. It also describes data-switching behaviour during calls. On Android, check the selected data SIM in the phone’s SIM or mobile-network settings; names vary by manufacturer.
Avoid taking a call during the measured test. If the data connection switches or the call interrupts the run, note it and repeat under the intended conditions. Two active SIM names in the status area do not establish which one carried a particular transfer.
Check the plan before blaming the network
Confirm that the SIM has active service and enough usable data. A speed restriction after exhausting an allowance answers a different question from a test made before the allowance runs out. Likewise, check any plan conditions relevant to the network mode you intend to use.
Record the observed technology, such as 4G or 5G, alongside the result. If you want to assess normal everyday use, keep your normal supported mode and note transitions. If you deliberately change a supported setting to compare 4G with 5G, make that a separate, clearly labelled series. The icon alone does not measure throughput or call reliability.
Use one primary tool for the comparison
Choose a reputable tool through its official website or app-store links. These options answer somewhat different questions:
| Tool | Useful starting point | What to remember |
|---|---|---|
| TRAI MySpeed | Mobile-network tests and the metrics offered in the current app | Follow the official download links and identify the selected test mode |
| TRAI MySpeed web tester | A browser-based download, upload and latency check | Keep the browser and test method consistent |
| Speedtest by Ookla | Repeated measurements with the endpoint recorded | Record the server; preserve the same settings in your comparison |
| Cloudflare Speed Test | Connection quality and consistency against Cloudflare’s network | Its destination and methodology can differ from another tool’s |
| FAST.com | A download check with additional upload and latency information | “Show more info” exposes additional measurements; the test uses Netflix servers |
TRAI’s MySpeed description lists download, upload, latency, jitter and packet loss among the app’s measurements. The web tester description identifies download, upload and latency. Do not assume every mode or device shows exactly the same fields.
Use a second tool to investigate an unexplained difference, while keeping its results in a separate column. Do not average unlike tools into one “official speed.” Their servers, traffic patterns and summary calculations may differ.
Keep the handset and background activity consistent
Use the phone you expect to use after buying or porting the SIM. Keep its position and orientation reasonably consistent, and avoid comparing one device resting near an open window with another deep inside a room. Note the model and any meaningful settings difference.
Pause unrelated large transfers such as an app download or photo backup for a baseline comparison. Keep the test in the foreground and let it finish. If the phone becomes unusually hot or a download continues in the background, record that condition and repeat after returning to a comparable state.
Record whether a VPN, proxy or managed network profile is part of the connection. If it is active, the measured route includes it. Compare configurations separately where appropriate and permitted, rather than changing that route silently between operators. There is no need to disable general security protections to perform an ordinary speed test.
Run a repeatable comparison
Choose a place and time window you actually care about
Start at a fixed spot: your desk, a shop counter or another important location. Record indoors or outdoors. A test beside a window is useful, but it does not substitute for the room where you normally work.
Compare networks within a similar time window, then repeat during another relevant period. A morning result from SIM A and a busy evening result from SIM B are not a controlled operator comparison. If you investigate an evening slowdown, keep the SIM and location fixed while repeating across those periods.
Record the server and investigate a second endpoint separately
If the tool permits manual server selection, record the server used and try to keep it the same for the first operator comparison. This reduces one source of variation; it does not make every network’s route to that server identical.
Then, if a result seems unusual, compare another suitable endpoint as a separate check. A server hosted within one operator’s network and a destination reached across other networks can produce different results. Neither one alone establishes performance to every website or app.
If a tool chooses its endpoint automatically and does not expose a manual choice, record whatever destination information it shows. Do not invent an endpoint or imply that every test ran to the same server. Cloudflare’s methodology description explains its use of network routing to reach its test infrastructure.
Repeat without turning a personal test into a data marathon
A practical starting routine is three completed runs at each important place and time window, with a brief pause between runs. Repeat on another ordinary day if the decision matters. This is an editorial suggestion, not a statistical guarantee or a TRAI-prescribed sample size.
When comparing two SIMs sequentially, alternate the order between sessions: A then B in one session, B then A in the next. That can reduce the effect of always testing one network earlier, although conditions still cannot be perfectly controlled.
Do not run separate tools simultaneously. They compete for the same connection and change the conditions you are trying to measure. Keep incomplete attempts in your log, noting whether the cause was a connection failure, an app error or an interruption you introduced.
Check how much data the tests use
Speed tests transfer real data in both directions. The amount depends on throughput, duration and methodology, so there is no universal data cost per test.
For scale, a hypothetical 100 Mbps transfer sustained for 20 seconds carries about 250 MB of payload: 100 × 20 ÷ 8 = 250, using decimal units. That is a calculation for one transfer, not a claim that a particular app always consumes 250 MB. Uploads, other test stages and overhead can add to the total.
Check your allowance and the app or phone’s usage counters before repeating many runs. If data is limited, use fewer measurements and focus on the real task you need to complete.
Read the results without getting misled
Download and upload speed answer different questions
Download throughput describes how quickly the test receives data; upload throughput describes how quickly it sends it. Both are usually displayed in Mbps, meaning megabits per second. A transfer application may instead display MB/s, or megabytes per second. Dividing Mbps by eight gives a rough unit conversion before overhead: 80 Mbps corresponds to 10 MB/s.
A high download number does not establish a fast document upload or a stable outgoing video stream. Nor does it guarantee that one app will deliver at the same rate: the app’s server and the path to it can be different.
Ping or latency: check what the tool measures
Latency is a delay measurement, commonly shown in milliseconds. In many consumer tests, ping represents a request’s round trip to the test destination. Read the tool’s definition: measurement techniques are not necessarily identical across tools.
Lower delay to one test server does not promise the same delay to a game server in another country. Compare the same metric and destination where possible. If a video meeting or game is the actual concern, check that application’s own available diagnostics as well.
Loaded latency can explain a connection that feels slow while busy
An unloaded or idle measurement is made without the tool’s large transfer in progress. A loaded measurement checks responsiveness while traffic is being sent or received. Netflix explains this distinction in its FAST.com latency guide.
For example, a fictional connection showing 25 ms idle latency and 240 ms during upload deserves a closer look if calls become unresponsive while photos upload. That pattern is a clue about performance under load, not proof of which piece of equipment is responsible. Repeat it and compare an actual call with and without the competing transfer.
Jitter and packet loss add context
Jitter describes variation in delay. Packet loss describes test traffic that does not arrive as expected. Both can help explain problems with real-time communication even when the download number looks strong. Cloudflare’s connection-quality explanation discusses why throughput alone is incomplete.
Compare the same tool’s definitions, units and conditions. A missing packet-loss field means the measurement is unavailable, not that loss was zero. A displayed zero-loss result describes that measurement; it is not a guarantee for every future call or every internet destination.
Avoid one universal “good ping” or “good jitter” cutoff for all applications. Interpret the measured values together with your task, destination and repeated experience.
Use the median, and report failures beside it
For five completed measurements, sort the values and take the third one as the median. For an even number, average the two middle values. Keep comparable sessions together rather than combining different rooms and network modes into one reassuring number.
Consider seven fictional attempts: five complete at 8, 12, 14, 18 and 120 Mbps, and two fail. The median completed download speed is 14 Mbps. The arithmetic mean of completed results is 34.4 Mbps, while the maximum is 120 Mbps. Reporting only the peak would hide how different most completed runs were.

Report “median 14 Mbps across five completed runs; two of seven attempts failed” rather than presenting the successful subset as all the evidence. Do not silently replace every failed test with zero, or discard it without explanation. State what failed and how you calculated the summary.
If one tool publishes percentiles using its own convention, read its definitions. Your personal median of five results is a separate calculation from a platform’s aggregation across many users, handsets and locations.
Keep a comparison log
| Field | Example of what to record |
|---|---|
| Session | Date, time window and exact test spot |
| Connection | Operator, active plan state, selected SIM and observed 4G/5G mode |
| Device | Phone model, relevant settings and whether conditions changed |
| Test | Tool/version, app or browser, endpoint and test mode where available |
| Results | Download, upload, latency, loaded latency, jitter and loss if provided |
| Attempts | How many completed; errors, failures and interruptions separately |
| Real task | Whether your upload, call or other task completed acceptably |
Save result IDs or screenshots if useful, but remove personal numbers, public IP addresses and precise home-location details before sharing publicly. Review the chosen tool’s privacy and permission information when contributing measurements or granting location access.
What to do when the numbers disagree
A fast test but one slow app
Try another fresh transfer and a second suitable test destination. If the issue stays limited to one app, investigate that service and its route rather than immediately concluding the entire mobile network is slow. A speed test does not measure every content-delivery system.
Good morning results but poor evening performance
Repeat at the same spot over several comparable days. Keep the plan state, handset and endpoint consistent. A repeatable time pattern can support a congestion investigation, but it does not prove congestion without further evidence; other conditions may also change.
Good outdoor results but weak indoor performance
Compare your normal room, a window-side position and a safe outdoor location as separate rows. The difference is useful evidence for choosing a SIM or reporting a fault. Do not replace the indoor result with the outdoor maximum when deciding whether the connection meets your needs.
One phone or test server is much faster
Change one variable at a time. If you switch the handset, keep the SIM, place and method fixed where practical. If you switch the endpoint, keep the handset and connection fixed. Record the comparison as incomplete if that is not possible.
Persistent failures need more than another benchmark
Check the active plan, allowance, selected SIM and whether other users nearby have the same issue. Then give the operator a concise log of locations, times, technology, completed tests and failures. A specific complaint is easier to investigate than a single unexplained screenshot.
Frequently asked questions
Which mobile speed-test app is the most accurate?
There is no single tool that perfectly represents every application and destination. Use a reputable primary tool consistently, understand what it measures and investigate important discrepancies with another tool or a real task.
Should I test over Wi-Fi or mobile data?
For a SIM comparison, turn Wi-Fi off and confirm the intended data SIM. For broadband or hotspot performance, run a separately labelled test of that full setup.
Is a hotspot test the same as testing directly on the phone?
No. A laptop connected through a hotspot adds another device and a local Wi-Fi or USB link. It is useful if that is how you work, but keep it separate from a direct phone test.
Can I compare 4G on one operator with 5G on another?
You can compare the everyday setups available to you, provided you label them clearly. That does not isolate a like-for-like 4G or 5G comparison. Record the mode actually observed during each run.
Does a speed test prove voice-call quality?
No. Check ordinary incoming and outgoing calls separately. An app-based meeting or internet call also uses a different service path from a normal mobile voice call.
Can a speed test tell me which network is best everywhere?
No. Your results apply to the sampled devices, places, times and test destinations. Use them to answer your own question and describe the scope when sharing.
How many tests are enough before porting?
There is no magic personal-test count that guarantees future performance. Cover your important places and busy periods, keep failures, and confirm your real tasks. Several repeatable observations are more useful than many identical tests at a location you rarely use.
The suggested routine is practical consumer guidance. It is not an original operator benchmark, a regulatory compliance test or a guarantee of future service. The cover is illustrative, and the chart uses fictional measurements.
Related mobile network guides
Use your results alongside our guide to choosing a SIM for villages and highways in India. If the connection fails before you can complete a test, follow our mobile data troubleshooting guide for Android and iPhone.