updated 24 Aug 2022
Helium is a LoRaWAN network built on a blockchain, to provide incentives to run a 'hotspot' that provides network coverage for IoT devices using LoRaWAN radio technology. It has been growing extremely fast, adding up to 80,000 hotspots per month in 2022, with close to 1 million installed and providing coverage by the second half of 2022 (except for the ones run by scammers; location spoofing is a big problem for the Helium network).
Of course the assumption has been that more and more people will start using this network to send real IoT device data. Sending data over the network is very cheap, and uses Data Credits (DC).
The question we are trying to answer here is, nevermind the hype, how much real data traffic does the Helium network currently see?
If you prefer a video, Coffeezilla has released a good one in Aug 2022, The Billion Dollar Network No One Uses.
Web3 Index does an excellent job showing the trend of DC related to data traffic: it is clear that despite the continuous growth of the Helium network, data traffic is still extremely small, after peaks in May and June 2022 (which was not real traffic but a scam using Helium console free DC as reported on the Helium Discord server). Looking at this graph, it is also possible that the entire surge of Jan to June 2022 was a result of this scam, with data traffic going back to 2021 levels.
Here we want to dig a bit deeper to check data packet numbers vs hotspots: how many hotspots are actually doing real work?
My Python script uses the Helium Blockchain API to retrieve data traffic information. It ranks the most active hotspots by data packet count, as this is the best indicator for real traffic (vs DC value as shown by the Web3 Index).
The main findings are in the table below, and this plot is also revealing: it shows that only about 15,000 hotspots see any significant traffic (at least 100 packets per week), and only 130,000 (14% of the 932K+ proudly 'deployed' hotspots) see any traffic at all. Around 800,000 Helium hotspots do not see any data traffic. At an average cost of USD 500 per hotspot, that is USD 400 million of stranded assets (Data-Only Hotspots are cheaper but only account to less than 0.2% of the total hotspots).
While this analysis is based on 7 days of traffic data (17-24 Aug 2022), the below plot looked very similar in Feb/March/May 2022.
The helium_data_traffic_extraction_1.py script tells us (see extraction_sample_output.txt):
| Feb 2022 | May 2022 | Aug 2022 | |
|---|---|---|---|
| Total number of data packets over 7 days | 82,130,820 | 46,953,101 | 17,869,289 |
| Average data packets per day | 11,732,974 | 6,707,586 | 2,552,756 |
| Total number of active hotspots | 122,258 | 149,072 | 130,265 |
| Active hotspots vs total Helium network | 21.27% | 17.89% | 13.97% |
| Average DC per data packet | 1.63 | 6.38 | 1.47 |
| Average DC value per day | USD 191.46 | USD 428.23 | USD 37.43 |
| Most active single hotspot share of total traffic | 1.38% (161,930 packets/day) | 2.68% (179,727 packets/day) | 1.13% (28,766 packets/day) |
| Details of most active hotspot | Rough Mango Starling: Browan/MerryIoT, its location had not been set, but Taiwan dominated the list of most active hotspots, probably related to stress testing | Mammoth Mint Butterfly: Data-Only Hotspot without location | Active Navy Poodle: SenseCap mining hotspot in Philadelphia, USA (earned $6 in this week, PoC+data) |
| Number of hotspots accounting for 20% of total traffic | 54 | 11 | 45 |
| Traffic distribution over hotspots | Half of the active hotspots saw less than 10 packets over a week | Half of the active hotspots saw less than 10 packets over a week | Half of the active hotspots saw less than 10 packets over a week |
With 'active hotspot' I mean a hotspot that sees real data traffic, not only the Proof-of-Coverage exchanges.
The most significant changes over Feb-May-Aug 2022:
- Total traffic in terms of data packets has dropped from 11M per day (Feb 2022) to around 2.5M per day, and traffic has been stable since mid June 2022 (in terms of DC, as seen on Web3 Index)
- The average packet size has gone back to 1.47 DC per packet (35 bytes) after a peak in May, caused by a scam abusing free Helium console DC. Scammers were sending max length packets, pulling up the average.
- Location-less Data-Only Hotspots which were the top 60 most active hotspots in May 2022, have disappeared (scam above).
- The share of active hotspots is decreasing, showing that the growth in hotspot deployment is still outpacing any growth in data traffic (sensor deployment).
- Data traffic over the Helium network is not growing much, and that does not seem to bother Helium: they have moved on to the next hype, their 5G network.
Based on the above ranking of most active hotspots I retrieved location information of the top 2,000 hotspots in the week 17-24 Aug. Below is the list grouped by country (57), accounting for 11,849,283 packets or 66% of the total traffic.
- The USA clearly dominates the data traffic, registering 50% of the total packets (with around 38% of total installed hotspots). Europe takes about 25% of traffic, led by France and Germany.
- Location-less Data-Only Hotspots which took 69% of total traffic in May 2022 dropped to 0.18% (country 'None').
- The Bahamas are a clear outlier: they have more traffic than China, with some very busy hotspots (7x more traffic per hotspot than the average 5k packets per week). All hotspots on the Bahamas seem to be on Nassau island, around 170, and they see more traffic than the 93k hotspots in China (based on this list of top 2,000 hotspots).
- Overall the numbers have not changed much since May 2022; the USA has seen a significant increase in traffic, especially the city of Philadelphia has 11 of the top 20 most active hotspots. Other countries have seen increases too, but at 50,000-100,000 packets per day that is probably still less than 1,000 sensors deployed.
| country | packet count (week) | share of traffic | hotspot count | avg packets per hotspot (week) |
|---|---|---|---|---|
| United States | 5,945,596 | 50.18% | 744 | 7,991 |
| France | 770,801 | 6.51% | 138 | 5,586 |
| Germany | 577,521 | 4.87% | 136 | 4,246 |
| The Bahamas | 485,541 | 4.10% | 14 | 34,682 |
| China | 372,804 | 3.15% | 69 | 5,403 |
| Italy | 352,039 | 2.97% | 98 | 3,592 |
| Taiwan | 334,299 | 2.82% | 40 | 8,357 |
| Canada | 324,822 | 2.74% | 98 | 3,315 |
| United Kingdom | 324,188 | 2.74% | 93 | 3,486 |
| Netherlands | 315,655 | 2.66% | 61 | 5,175 |
| Turkey | 258,679 | 2.18% | 80 | 3,233 |
| Spain | 255,637 | 2.16% | 50 | 5,113 |
| Bulgaria | 156,392 | 1.32% | 15 | 10,426 |
| Australia | 134,689 | 1.14% | 35 | 3,848 |
| fail | 130,252 | 1.10% | 0 | inf |
| Switzerland | 121,619 | 1.03% | 41 | 2,966 |
| Romania | 114,774 | 0.97% | 32 | 3,587 |
| Poland | 78,636 | 0.66% | 12 | 6,553 |
| Lithuania | 75,380 | 0.64% | 19 | 3,967 |
| Iran | 74,209 | 0.63% | 6 | 12,368 |
| Denmark | 70,997 | 0.60% | 19 | 3,737 |
| Greece | 52,840 | 0.45% | 9 | 5,871 |
| Thailand | 43,457 | 0.37% | 8 | 5,432 |
| Belgium | 40,641 | 0.34% | 15 | 2,709 |
| Colombia | 40,447 | 0.34% | 8 | 5,056 |
| Argentina | 36,312 | 0.31% | 4 | 9,078 |
| South Africa | 30,938 | 0.26% | 14 | 2,210 |
| Sweden | 28,167 | 0.24% | 11 | 2,561 |
| Ireland | 23,717 | 0.20% | 6 | 3,953 |
| Ukraine | 23,358 | 0.20% | 5 | 4,672 |
| Cyprus | 22,183 | 0.19% | 2 | 11,092 |
| None | 21,544 | 0.18% | 5 | 4,309 |
| Portugal | 19,812 | 0.17% | 13 | 1,524 |
| Hungary | 19,218 | 0.16% | 7 | 2,745 |
| Norway | 18,819 | 0.16% | 9 | 2,091 |
| Czechia | 16,953 | 0.14% | 7 | 2,422 |
| Slovenia | 16,691 | 0.14% | 5 | 3,338 |
| South Korea | 16,428 | 0.14% | 3 | 5,476 |
| Mexico | 14,192 | 0.12% | 4 | 3,548 |
| Slovakia | 12,852 | 0.11% | 2 | 6,426 |
| Brazil | 11,691 | 0.10% | 3 | 3,897 |
| Costa Rica | 9,375 | 0.08% | 1 | 9,375 |
| India | 9,016 | 0.08% | 1 | 9,016 |
| Serbia | 7,144 | 0.06% | 4 | 1,786 |
| Zimbabwe | 6,998 | 0.06% | 1 | 6,998 |
| Belarus | 5,359 | 0.05% | 2 | 2,680 |
| Puerto Rico | 4,753 | 0.04% | 3 | 1,584 |
| Austria | 4,180 | 0.04% | 3 | 1,393 |
| Estonia | 3,441 | 0.03% | 2 | 1,720 |
| Venezuela | 2,618 | 0.02% | 1 | 2,618 |
| New Zealand | 2,570 | 0.02% | 1 | 2,570 |
| Malaysia | 1,731 | 0.01% | 1 | 1,731 |
| Martinique | 1,699 | 0.01% | 1 | 1,699 |
| Finland | 1,583 | 0.01% | 1 | 1,583 |
| Albania | 1,482 | 0.01% | 1 | 1,482 |
| Oman | 1,277 | 0.01% | 1 | 1,277 |
| Singapore | 1,267 | 0.01% | 1 | 1,267 |
Based on the above ranking of most active hotspots I retrieved location information of the top 2,000 hotspots in the week 24-31 March. Below is the list grouped by 56 most active countries, accounting for 38,781,194 packets or 82% of the total traffic. It also shows the average packets per hotspot, with a few clear outliers:
- 'None' is the category of Data-Only Hotspots without asserted location, which take almost 70% of the total traffic, with only 85 hotspots. These hotspots see on average 1,885 packets per hour: they are over 50 times more busy than the 591 busiest hotspots in the USA.
- 'fail' is a category of Helium accounts that should be hotspots ('client' in the state_channel summaries) but are listed as 'account' in Explorer. There are 121 of these in our list of 2,000 most active hotspots (in Feb 2022 there were 90 but i had assumed those were glitches).
- Compared to Feb 2022 (see below), USA traffic has dropped from 7M3 to 3M3 packets per week. The average packets per hotspot has not changed much, 5,642 vs 6,534 before. It can be assumed that some of the USA traffic has moved to Data-Only Hotspots, but it seems impossible to know for sure from the data.
- Taiwan has fallen from 1st place to 9th, or around 1% of total traffic, probably because some manufacturers stopped their stress testing. Maybe they are stress testing Data-Only Hotspots now, but as long as they don't assert the hotspots' locations, we don't know.
- Sweden went from 6th place to 4th, tripling its average packets per hotspot from 9,676 to 29,719. 80% of its packets are accounted for by 1 Data-Only Hotspot, one of the few Data-Only Hotspots with an asserted location. Sweden has only 24 hotspots in the top 2,000 compared to 624 before, and packets dropped from 6M to 713K per week.
- Serbia fell from 2nd to 23nd place (from 15.42% to 0.18%), reducing its packets from 10M6 to 71K, with only 32 hotspots in the top 2,000 vs 757 before. Greece is similar. Other European countries remain in the top 20 but have seen their traffic reduced to 10-20% of February levels. It is unlikely that this is due to a reduction in real sensor data traffic, it probably means that Data-Only Hotspots have been deployed in those countries, and the packets that they pick up are not added to this country counts.
While overall traffic has obviously decreased, a large share of real traffic has probably moved to Data-Only Hotspots making our country ranking rather meaningless, as Data-Only Hotspots are not assigned to countries. However, the most active Data-Only Hotspots are seeing unrealistically high volumes of traffic so it is possible that some kind of arbitration is happening.
[for the week of 24-31 May 2022]
| country | packet count | share of traffic | hotspot count | avg packets per hotspot |
|---|---|---|---|---|
| None | 26,919,910 | 69.41% | 85 | 316,705 |
| United States | 3,334,659 | 8.60% | 591 | 5,642 |
| fail | 2,039,701 | 5.26% | 121* | inf |
| Sweden | 713,262 | 1.84% | 24 | 29,719 |
| Germany | 473,790 | 1.22% | 108 | 4,387 |
| France | 462,413 | 1.19% | 50 | 9,248 |
| Spain | 450,853 | 1.16% | 73 | 6,176 |
| China | 443,137 | 1.14% | 74 | 5,988 |
| Taiwan | 402,835 | 1.04% | 46 | 8,757 |
| Canada | 373,616 | 0.96% | 82 | 4,556 |
| Switzerland | 308,456 | 0.80% | 78 | 3,955 |
| Italy | 305,433 | 0.79% | 41 | 7,450 |
| United Kingdom | 272,289 | 0.70% | 70 | 3,890 |
| The Bahamas | 258,637 | 0.67% | 12 | 21,553 |
| Romania | 256,460 | 0.66% | 61 | 4,204 |
| India | 205,142 | 0.53% | 18 | 11,397 |
| Turkey | 197,812 | 0.51% | 65 | 3,043 |
| Netherlands | 162,370 | 0.42% | 60 | 2,706 |
| South Africa | 131,391 | 0.34% | 26 | 5,054 |
| Lithuania | 95,554 | 0.25% | 32 | 2,986 |
| Australia | 92,394 | 0.24% | 21 | 4,400 |
| Portugal | 79,477 | 0.20% | 27 | 2,944 |
| Serbia | 71,120 | 0.18% | 32 | 2,222 |
| Greece | 67,621 | 0.17% | 19 | 3,559 |
| Colombia | 66,793 | 0.17% | 7 | 9,542 |
| Liechtenstein | 49,060 | 0.13% | 11 | 4,460 |
| Poland | 48,638 | 0.13% | 7 | 6,948 |
| Ukraine | 45,690 | 0.12% | 22 | 2,077 |
| Slovenia | 44,850 | 0.12% | 6 | 7,475 |
| Belgium | 44,154 | 0.11% | 14 | 3,154 |
| Bulgaria | 38,062 | 0.10% | 8 | 4,758 |
| Denmark | 35,826 | 0.09% | 11 | 3,257 |
| Thailand | 28,750 | 0.07% | 6 | 4,792 |
| Norway | 27,882 | 0.07% | 13 | 2,145 |
| Oman | 27,136 | 0.07% | 12 | 2,261 |
| Armenia | 26,448 | 0.07% | 13 | 2,034 |
| Hungary | 24,911 | 0.06% | 6 | 4,152 |
| Mexico | 21,362 | 0.06% | 5 | 4,272 |
| Czechia | 20,486 | 0.05% | 10 | 2,049 |
| Latvia | 18,898 | 0.05% | 4 | 4,724 |
| Ireland | 15,522 | 0.04% | 6 | 2,587 |
| Argentina | 10,817 | 0.03% | 2 | 5,408 |
| Cyprus | 9,954 | 0.03% | 3 | 3,318 |
| Puerto Rico | 8,511 | 0.02% | 1 | 8,511 |
| Venezuela | 7,592 | 0.02% | 1 | 7,592 |
| South Korea | 7,251 | 0.02% | 2 | 3,626 |
| New Zealand | 6,474 | 0.02% | 2 | 3,237 |
| Estonia | 6,349 | 0.02% | 3 | 2,116 |
| Hong Kong | 3,479 | 0.01% | 1 | 3,479 |
| Zimbabwe | 3,024 | 0.01% | 1 | 3,024 |
| United Arab Emirates | 2,586 | 0.01% | 1 | 2,586 |
| Austria | 2,437 | 0.01% | 1 | 2,437 |
| Malaysia | 2,422 | 0.01% | 1 | 2,422 |
| Finland | 2,086 | 0.01% | 1 | 2,086 |
| Israel | 1,996 | 0.01% | 1 | 1,996 |
| Malta | 1,752 | 0.00% | 1 | 1,752 |
| Singapore | 1,664 | 0.00% | 1 | 1,664 |
I made a Google Map showing the 2,000 most active hotspots (70% of total data traffic).
And this is the list of most active countries based on the 7,000 most active hotspots in the above week. It also shows the average packets per hotspot, with 2 clear outliers: Taiwan and The Bahamas. ('fail' is a set of 90 hotspots for which the API call for details failed, I did not bother to re-try).
| country | packet count | share of traffic | hotspot count | avg packets per hotspot |
|---|---|---|---|---|
| Taiwan | 10,889,283 | 14.71% | 95 | 114,624 |
| Serbia | 10,676,756 | 14.42% | 757 | 14,104 |
| Greece | 9,380,946 | 12.67% | 577 | 16,258 |
| France | 7,824,936 | 10.57% | 748 | 10,461 |
| United States | 7,390,235 | 9.98% | 1,131 | 6,534 |
| Sweden | 6,037,877 | 8.16% | 624 | 9,676 |
| Germany | 3,379,396 | 4.57% | 239 | 14,140 |
| Spain | 3,235,256 | 4.37% | 285 | 11,352 |
| United Kingdom | 2,736,144 | 3.70% | 357 | 7,664 |
| Turkey | 1,310,688 | 1.77% | 355 | 3,692 |
| Lithuania | 1,186,031 | 1.60% | 183 | 6,481 |
| Italy | 1,115,696 | 1.51% | 141 | 7,913 |
| Netherlands | 907,577 | 1.23% | 224 | 4,052 |
| Switzerland | 849,860 | 1.15% | 146 | 5,821 |
| The Bahamas | 765,755 | 1.03% | 11 | 69,614 |
| fail | 725,098 | 0.98% | 0 | inf |
| China | 721,135 | 0.97% | 70 | 10,302 |
| Latvia | 388,932 | 0.53% | 103 | 3,776 |
| Canada | 338,463 | 0.46% | 71 | 4,767 |
| Georgia | 261,502 | 0.35% | 88 | 2,972 |
| Romania | 260,670 | 0.35% | 58 | 4,494 |
| Norway | 246,787 | 0.33% | 30 | 8,226 |
| Hungary | 238,782 | 0.32% | 92 | 2,595 |
| Portugal | 236,827 | 0.32% | 84 | 2,819 |
| Poland | 216,773 | 0.29% | 23 | 9,425 |
| Thailand | 192,455 | 0.26% | 26 | 7,402 |
| Slovenia | 159,028 | 0.21% | 61 | 2,607 |
| Estonia | 151,687 | 0.20% | 55 | 2,758 |
| Belgium | 120,043 | 0.16% | 26 | 4,617 |
| Australia | 104,269 | 0.14% | 21 | 4,965 |
| Slovakia | 94,683 | 0.13% | 39 | 2,428 |
| Colombia | 85,544 | 0.12% | 19 | 4,502 |
| Lebanon | 73,234 | 0.10% | 30 | 2,441 |
| Denmark | 64,882 | 0.09% | 17 | 3,817 |
| Ukraine | 61,862 | 0.08% | 21 | 2,946 |
| South Africa | 54,308 | 0.07% | 11 | 4,937 |
| Liechtenstein | 36,388 | 0.05% | 4 | 9,097 |
| Ireland | 33,061 | 0.04% | 11 | 3,006 |
| Czechia | 30,802 | 0.04% | 11 | 2,800 |
| Bulgaria | 29,216 | 0.04% | 14 | 2,087 |
| Croatia | 28,097 | 0.04% | 14 | 2,007 |
| Mexico | 21,820 | 0.03% | 2 | 10,910 |
| Luxembourg | 19,663 | 0.03% | 1 | 19,663 |
| Chile | 14,843 | 0.02% | 4 | 3,711 |
| Singapore | 13,846 | 0.02% | 6 | 2,308 |
| Austria | 8,353 | 0.01% | 4 | 2,088 |
| Brazil | 8,302 | 0.01% | 5 | 1,660 |
| Uruguay | 6,615 | 0.01% | 1 | 6,615 |
| South Korea | 5,491 | 0.01% | 1 | 5,491 |
| Puerto Rico | 5,243 | 0.01% | 1 | 5,243 |
| Zimbabwe | 4,879 | 0.01% | 1 | 4,879 |
| Philippines | 4,145 | 0.01% | 2 | 2,072 |
| Cyprus | 1,891 | 0.00% | 1 | 1,891 |
| Finland | 1,494 | 0.00% | 1 | 1,494 |
| India | 1,328 | 0.00% | 1 | 1,328 |
(this is based on the above 7 day period, results may change over time)
- Only around 20% of the published Helium hotspots (574,000+) are involved in data traffic, the other 80% (450,000+) hotspots are not seeing any data traffic.
- At least 10% of data traffic is seen by a dozen hotspots in Taiwan, probably for stress testing devices/gateways. Germany also has some extremely busy hotspots.
- Around 90% of data traffic is seen by around 7,000 hotspots; in other words, 90% of traffic is handled by 1.2% of total installed hotspots.
- When running the analysis for the last 24 hours only, the number of active hotspots can go down to 10%, but the ratio of 90% of traffic seen by 7,000 hotspots seems to hold.
- Europe sees around 60% of the total data traffic, and the USA around only 10%, which is less than each of Serbia, Greece and France.
Note that although I hope this is interesting, data packet traffic is not a reliable metric to evaluate the actual use of the Helium network, as it can easily be inflated by people who have an interest to do so; that would be cheap both in terms of hardware and DC required.
UPDATE 25 March 2022: I ran the script again for the last 3 days (only retrieving the location of top 2337 hotspots) and saw these differences:
- United States has increased from 10% to 18% of total data traffic
- more than 5% of data traffic is seen by hotspots without location, i assume these are testing setups
- the Netherlands and China have seen a large increase in data traffic
The details of data traffic transactions happen in state channels, a 'layer 2' technology to keep the huge number of transactions off the main 'layer 1' blockchain. Only the summaries of these DC transactions become part of the Helium blockchain. Helium allows us to access the blockchain via an API.
The python script helium_data_traffic_extraction_1.py uses this API to analyse the data traffic:
- for a given NUMBER_OF_DAYS we retrieve all the state channels
- sum the data packets and DCs
- keep track of the data packets per hotspot
- rank this list to find the most active hotspots
- save this list as CSV, plot a graph and save as PNG
- list the most active hotspots with their location, until a given FRACTION_THRESHOLD is reached
- rank the countries of the most active hotspots
The only configuration parameters are at the top of the script:
FRACTION_THRESHOLD = 0.2 # cut the hotspot ranking when we reach this fraction 0-1
NUMBER_OF_DAYS = 3 # how many days to go back from now
Increasing FRACTION_THRESHOLD slows down the completion of the script, as it will make an API call for each hotspot to retrieve its location.
I decided to use only the 7,000 most active hotspots, representing 90% of total data traffic. The script helium_hotspots_csv_details_2.py takes the csv output of above extraction script to retrieve hotspots details including locations. With this output csv I created the Google Map above, and I used this script helium_data_traffic_country_3.py to create the country list, resulting in this country_sample_output.txt.
It is very well possible that mistakes were made in the script and in the above interpretation, I am open to corrections and suggestions.
I got interested in Helium when I saw dozens of hotspots appearing in my area, which all turned out to be fake. I am trying to contribute to mapping of real Helium network coverage in Hong Kong, and I remain a big fan of The Things Network.