Steem’s fair-weather friends - December update on user retention numbers

You may find the title provocative. You may even think it sounds a little negative. But in fact there’s an important point to be made. You’ll have to read through to the end to find out what it is.

User retention statistics are notoriously difficult to measure. This is particularly true for an environment like the steem blockchain where:

• There are numerous different ways to participate. For example we have posting and commenting, reading articles or viewing videos, as well as curating (upvoting and downvoting);
• Some of the above actions can be automated and so may be performed by bots. These automated actions may continue long after an account has stopped actively using the platform. Curation in particular can fall into this category;
• Not all of the actions can be reliably measured through the information sources available. Currently viewer and reader numbers are not easily harvested and there are doubts over the accuracy of the measures being shown.

I undertook my last study of user numbers and user retention just over one month ago. Given the above complications I decided upon the following definition for an active account:

Active account = Account which has posted or commented within the last thirty days

The definition is subjective and the measure, thirty days, is an arbitrary figure based simply on my experience of the blockchain over the last six months. I think it feels about right. One or two weeks could be a holiday but if you haven’t posted or commented in the last month you’re falling into the category of inactive.

---

November analysis

In my last analysis I used the above definition to look at the percentage of accounts that are still active (i.e. those accounts that had posted within the last thirty days). I then charted the results grouped by account joining date to see if older accounts were stickier.

I also looked at the percentage of accounts that were still posting thirty days after their joining date. This was intended as a measure of how quickly people buy-in to or turn away from the platform.

These were the results:

(Note that only those accounts that have posted or commented were included).

Overall, not a particularly positive picture. User retention was drifting along at around 10%; a worryingly low level. This was the case irrespective of joining date over 2017.

For the second metric, around 40%-60% of users stopping posting within 30 days of their join date: broadly somewhere around half of all users.

---

December analysis

So how has the picture changed over the last month? We have a backdrop of rapidly rising steem and SBD prices and greater press coverage of cryptocurrencies. Have these factors translated through a more attractive proposition and increased user retention?

Here is the same analysis one month further on:

Immediately we can see a gradual incline in the percentage of accounts still posting (the blue line) pointing to improved user retention for accounts that signed up to the blockchain more recently.

But if we look more closely, we can also see that there is in fact a rise in user retention irrespective of joining date; from our previous figure of around 10% to something closer to 20%.

I have to admit, this threw me a little at first. So I drew up a comparison to look at the figures more closely:

There's a clear impact. The user retention has moved up closer to the 20% level across all joining dates. What could this mean? Well, the only thing I could think of is that we are seeing users returning to action. That is to say, users who had previously stopped posting have put their fingers to the keys once again. And in significant numbers.

To investigate this phenomenon I looked at all accounts on the blockchain that had posted or commented and broke down their activity by month. Here are the results:

• In green we have the month in which the account first posted.
• In red we have accounts which stopped posting in the month (i.e. users who posted in the previous month but did not post in the month in question).
• In blue we have accounts which restarted posting in the month (i.e. users who did not post in the previous month but did post in the month in question).

Immediately we can see a huge number of accounts (11.5k) returned to posting in December after being inactive in November. Given the rise in user retention across all joining dates, some of these users will have been inactive for a significant period of time.

---

As a final comparison I looked at the net movement in poster numbers from the above chart (shown below in gold) and overlaid the end of month steem price. As might be expected, there's a clear impact from the currency price on poster numbers. When the steem price rises we see more posters. Previously this took the form of a rush of new accounts. Now we're also seeing a surge in poster numbers from returning users.

This brings us to the important point to be registered. To some greater or lesser extent, the value of the steem blockchain and steem currency will be influenced by user numbers. Whilst user retention numbers in the past have been fairly poor, there are clearly large numbers of inactive users still interested in this blockchain. Whether they exist as readers and viewers, as passive curators, or whether they have returned from completely off the grid is not clear. However there remains untapped potential which can be enticed back into a more active role under the right conditions. This is another avenue of growth that is worth investigating.

---

Questions

That's all for today.

If you have any questions or spot any errors please do not hesitate to leave a comment!
Thanks for reading!

---

Methodology and Tools for Analysis

Tools

Raw data was obtained through sql queries of steemsql using Valentina Studio.
Data was analysed in LibreOffice and illustrated in Numbers (spreadsheet tools).
Data was obtained for various timescales, including a particular focus on 2017 and the comparison between studies in November and December.

SQL query
I used the following SQL query:

SELECT
    CONVERT(date, Accounts.created) AS [AccountStart], 
    CONVERT(date, Accounts.last_post) AS [LastCommentOrPost],
    datediff (day, CONVERT(date, Accounts.created), CONVERT(date, Accounts.last_post)) AS [stayDuration],
    Count(Accounts.name) AS [Accounts]

FROM
    Accounts (NOLOCK)

WHERE
    Accounts.post_count > 0 

GROUP BY
   CONVERT(date, Accounts.created),
   CONVERT(date, Accounts.last_post),
   datediff (day, CONVERT(date, Accounts.created), CONVERT(date, Accounts.last_post))

---

Posted on Utopian.io - Rewarding Open Source Contributors

---