At the height of the bull market in January 2018 there were 130k authors active on Steem. Fast forward one year to the bear market conditions of January 2019 and this number had fallen to only 40k.

There is a strong correlation between the price of Steem and the number of active authors. This could be due to a general rise and fall in interest for all things crypto when the markets wax and wane but a more likely cause is the rise in dollar rewards available as the Steem price increases.

When considering the path to mass adoption, retaining existing users is as important as attracting new ones. An analysis of the relationship between earnings levels and retention seems like a good place to start.

---

1. Retention by earnings level: January 2019 to February 2019

The basic questions I am investigating here are:

• Is there a relationship between retention levels and earnings levels?
• At what level of earnings do we see a difference in retention?

This first study considers all accounts that posted or commented in the first two weeks of January 2019. The accounts are aggregated into bands of earnings and the number of accounts within each band is determined. Retention rates are then calculated based on whether each account made a post or comment in the first two weeks of February 2019, i.e. after an interval of one month.

For the purposes of all this analysis, earnings are defined as author rewards excluding those rewards purchased from vote bots. No allowance is made for curation or beneficiary rewards since these earnings sources are not directly related to posting. In any case these elements are more of a concern for larger SP holders and this study is focused on smaller accounts. The earnings are total rewards over the two-week period and are quoted in Steem (not STU). Note that the earnings levels are based on the first two weeks of January and take no account of earnings in February or earnings over the intermediate period.

The overall retention rate using this definition was 53%, i.e. 53% of authors posting/commenting in the first two weeks of January also posted in the first two weeks of February, whilst 47% did not. How does this break down by earnings level?

On the chart:

• The blue bars show the number of accounts at each earnings level.
• The orange diamonds show the percentage of the January accounts that also posted in the first two weeks of February, i.e. the retention level.

So there is good news and bad news here:

• Good news: Retention is actually quite sticky for authors earning one Steem or more. We have retention of over 70% at 1 Steem, increasing towards 90%-100% by the time we reach 10 Steem earned over the two weeks.
• Bad news: Retention for the first earnings band, i.e. less than 1 Steem, is 40%. And 70% of authors / commenters fall into this first band.

The overall retention level of 53% is thus heavily influenced by this first earnings bracket. So let's look at the low earnings results at a more granular level.

Here is the same study but with smaller buckets and lower earnings levels:

• Smaller buckets: 0.1 Steem rather than 1 Steem.
• Lower earnings levels: from 0 Steem - 5 Steem rather than 0 Steem - 50 Steem.

A similar pattern to the first chart and, again, there is good news and bad news:

• Good news: Retention levels improve at very low earnings levels - we can see an impact from as low as 0.1 Steem.
• Bad news: Retention for the first earnings band, i.e. less than 0.1 Steem, is close to 30%. And 53% of authors / commenters (across all earnings) fall into this first band.

---

2. Retention by earnings level: Looking at the full month of January

At this point I wanted to do a quick check on whether the 14 day period was having an effect. Is it too short? For this next study I lengthened the earnings period to the full month of January and the retention period to the full month of March, retaining a gap of one month between the two periods.

The two charts below compare the retention rates from the 14 day period (in orange) to the full month study (in red):

Earnings: 0 Steem - 50 Steem, 1 Steem buckets

Low earnings: 0 Steem - 5 Steem, 0.1 Steem buckets

The patterns are really very similar, so that's all good. Fourteen days looks like a reasonable length of time for the study.

---

3. Retention over the following months

Finally a look at how the retention levels progress over a longer period. Here we have:

• Orange diamonds - retention from January 1-14 to February 1-14, i.e. 1 month gap (as in section 1).
• Red diamonds - retention from January 1-14 to March 1-14, i.e. 2 month gap.
• Purple diamonds - retention from January 1-14 to April 1-14, i.e. 3 month gap.

I have not re-evaluated the earnings distribution for each month, so the classification of each account into an earnings bracket is maintained from January. It is likely that the accounts that remain pass into different earnings bands for each month. But I think that the initial distribution is a suitable proxy for authors at different earnings levels and sufficient for the conclusions drawn.

Earnings: 0 Steem - 50 Steem, 1 Steem buckets

Low earnings: 0 Steem - 5 Steem, 0.1 Steem buckets

In both charts retention levels fall by around 5% for each additional month. Retention is pretty sticky in subsequent months in comparison to the first month of the study.

I have checked that studies starting in February and March (rather than January) show a similar pattern of lower retention in the first month (so it is not just a case of March / April retention being stronger). This may suggest that the population splits into:

• A central core of accounts that post regularly, with long-term retention rates that are actually quite high.
• A second group of accounts, perhaps new accounts or returning accounts or occasional posters, for which retention rates are low, and many of which stop posting after the first month.

Further study would be required to confirm this.

---

4. Conclusions

The charts show a clear pattern:

• Retention increases with earnings level.
• Retention increases markedly between zero earnings and low earnings levels.
• Retention rates are surprisingly high (given my initial low expectations) starting from only a few Steem of earnings (e.g. 80% at 2 Steem).

Based on these studies it could be theorised that accounts are more likely to stop posting if they only earn small amounts of Steem and that they are much more likely to stop posting if they earn no Steem at all. This does sound pretty logical.

However, correlation does not imply causation. In fact it is possible that the cause and effect are completely reversed: New accounts arrive on Steem but quickly decide it is not for them. They leave the blockchain after making only a few posts or comments and earn no Steem in the process. Or to put it more succinctly: Accounts earn no Steem because they leave, rather than they leave after earning no Steem.

It is also possible that accounts do not leave at all but simply stop posting and commenting. Users could remain as consumers of content only or they could be busy playing NextColony or MagicDice.

Further study would be required to provide evidence for each of these possible conclusions.

---

5. Thoughts:

I'll admit that I have not concretely demonstrated a causal link between earnings and retention. More work is required. But it wouldn't surprise me if this turned out to be the case. Steem/Steemit adverts have traditionally highlighted the rewards as a main selling point of the blockchain. Users may become disillusioned if they do not obtain the earnings they were anticipating.

So is the answer to channel a greater share of rewards towards promising new accounts and low earners? It might help to improve retention rates. But it feels like a slow and tortuous path to mass adoption, rather than an organic solution.

The reward pool paid out approximately 1.3m Steem (or 450k STU) in author rewards in January 2019. This amount is significantly reduced once purchased votes are removed. The top 10% of authors gained 90% of the rewards; a pattern which is unlikely to change given the nature of stake-based reward voting. This leaves a very small pot with which to retain smaller accounts and keep new users happy.

It seems unlikely that the reward pool mechanism will drive us to mass adoption.

Instead I think that the answer is to look at other social networks and content creation sites, the facebooks, instagrams, and youtubes of this world, and to understand their paths to mass adoption without the use of crypto rewards. There is an opportunity with the arrival of Communities to shift the emphasis away from the promise of rewards towards a rounder offering, treating Steem/SMTs as one element rather than the main focus.

Early thoughts in any case. I have a few more studies planned that relate to mass adoption. Assuming I can get them to work, I'll add more thoughts as I go.

Thanks for reading!

---

Repository:

https://github.com/steemit/steem

This analysis is of data from the Steem blockchain which is an open source project.

---

Tools and scripts:

I used the block.ops analysis system to produce this study. Block.ops is an open-source analysis tool designed for heavy-duty analyses of the Steem blockchain data.

You can find the repository for block.ops here:
https://github.com/miniature-tiger/block.ops

The analysis used all the Steem blocks from the period analysed.

The study can be recreated by (once I upload the new analyses to github!):

• Loading the data for the relevant time period into block.ops.
• Using the earningsdistribution command from the command line, for example:
  $ node blockOps earningsdistribution "2019-04-01" "2019-04-15"

---

Thanks for reading!