Agile, data, and writing: STELLAR retrospective #3
May 2024 progress vs. my STELLAR roadmap goals for writing here on Substack: recap on post sizes, new lessons learned, actions, and plan adjustments for June
This post was originally published on agileteams.substack.com on June 5, 2024. As of July 12, all agile Teams content has moved to the agile section of karensmiley.substack.com. Please join us there!
When I posted at the start of March on the STELLAR Agile Strategy Canvas workshop, I promised to do a retrospective each month on how I’m doing vs. my goals. This post recaps what I learned in my third full month on Substack, and what I’m going to do differently in June.
Thanks for reading agile Teams! Subscribe for FREE to receive new posts (about monthly).
How It Started
My plan as of early March 2024, based on the STELLAR Agile Strategy Canvas:
ST=writing online regularly (quality rating=2)
E=publishing new content (3-5x/week, quality rating=7)
LL=write drafts on my phone to 80-90% completeness, and then move to the desktop site (Substack or LinkedIn) for publishing.
A=set up my Substack presence; build a post log in Google Docs
R=review monthly to see if I hit my target of 12-20 new posts for the month
In retrospective #2, these are the goals I set for May:
6-9 Posts: 2-3 per week for the first 3 weeks, in a mix of lengths: 1 Short, 2-5 Short or Medium, 1 Medium, 2 Long
12-18 Notes or Comments (Very Short)
How It’s Going
How’d I do in May? Here’s a look at each element of the STELLAR framework (other than ST, which isn’t changing).
E - Progress towards End goal
My goal for May was to publish 6-9 posts. I blew that away and published:
20 posts on Substack across 3 publications - 5 Short, 12 Medium, 3 Long
16 posts on LinkedIn - 15 Very Short and 1 Short (includes 3 posts to share links to the Medium and Long posts from Substack)
34 Substack Notes (most brief; I didn’t measure sizes of these)
Archive links are below in the References.1
Between Substack and LinkedIn, I wrote a total of 36 posts in May (over 1 per day). Mostly this was due to big increases vs. April in Very Short and Medium posts:
Why was the May post count so high? Well, at the end of April, I already had a solid draft of one very Long (way too long) article (
article PART 3 on ethics of generative AI for music). During May, I broke the PART 3 draft into 3 sections. One section, on Fairly Trained companies, was finished & published on May 23 after further splitting to create 9 detail pages (one page per company profiled).There are two reasons this high post count was possible in May with a reasonable level of effort.
The very long draft for PART 3 on the Fairly Trained companies was already mostly written (WIP completed in previous months).
My planned vacation for the end of May (expected to limit my available working hours) was disrupted by other events. Those events took my time, but not as much time as the vacation would have.
Despite successfully carving off that 10-post section in May, I still have significant WIP that isn’t yet ready to ship. The other 2 draft sections of PART 3 are waiting, and they are going to be BIG too!
In terms of my quality level target, I estimate that I returned to level 6 in May (up from 5 in April). I did exceed my End goal for posts/month, but other aspects of my writing aren’t quite where I’d want them to be to claim a higher quality level.
LL - Lessons and Limits
Here’s what I did with the top 3 new lessons I learned during April and the associated actions.
April Lesson 1: Connecting better with readers
(1) Use Notes more (I made progress during April; 20 Notes vs. 14 in March)
Result: in May I wrote 34 Notes (including restacks).
Impact:
As of June 1, I got a grand total of 38 reactions (most notes had none 😔) and some kind comments. (THANK YOU folks for the feedback!)
I enjoyed some nice discussions in Notes with a few new people I’m now subscribed to, such as
😊It was a lovely surprise to see
include our publication in a LinkedIn post naming “15 AI newsletters to read”. I’m now subscribed to his newsletter, too 😊
(2) Add audio voiceover to posts
Result: I manually recorded voiceovers for 4 posts, totaling 32.5 minutes of talk time.2
Impact: Unknown. It’s not obvious how to tell from my detailed post stats if anyone actually played the audio. I did not have any feedback or comments from anyone indicating that they listened to it. 🤷♀️
It’s worth noting that I added the 4 voiceovers to already-published posts, i.e. after the newsletter emails went out. So none of my email subscribers would have been notified that the voiceover was available, allowing them to choose whether to read or listen. That was ok for just experimenting with the recording & audio-embedding process, but it’s not very helpful for accessibility or flexibility.
April Lesson 2: Becoming a better writer
I gave myself two actions in April to begin measuring readability of my writing. Both actions are now completed (see below).
I’m currently using the Readability tool built into Microsoft Word. I’m not sure it’s trustworthy, though, especially on short posts.
Sentence counting is off. Example: On one short post, I manually counted 5 sentences. Word says it has 2. Wrong sentence counts may be skewing the metrics.
Word sometimes says that a post has fewer sentences than paragraphs. How does that make any sense?
Example: my month 1 retro supposedly has 80 paragraphs and 64 sentences.
I did notice that if I don’t end my bullets with periods, the ‘sentence’ counts are way lower. Even after doing this, the counts still don’t always make sense.
Sometimes Word won’t calculate a Flesch Reading Ease score for a post.
Word scored some very short LinkedIn posts as requiring 19+ years of formal education to understand. Looking at the posts, that seems absurd. One was a congratulatory note to former colleagues who had published an AI paper. Word scored it as requiring over 25 years of education to understand!
I calculated ratios for characters/word, words/sentence, and sentences/paragraph from the character, word, sentence, and paragraph counts Word displayed. Oddly, Word’s calculated ratios often do not match.3
Both Flesch formulas use syllable counts4, which Word doesn’t show in the statistics dialog, and sentence & word counts. Not having syllable counts impedes manually validating the Flesch scores.
Maybe the Word readability tool simply doesn’t work well on writing that has bullets, tables, quotations, images, pull quotes, hyperlinks, citations, and such?
To date, I’ve written and logged 65 posts and over 54,000 words. Keeping in mind these concerns about data quality, and ignoring the short LinkedIn posts, here are 3 things the readability data is currently telling me about my writing.
Posts in Agile Analytics and Beyond (AAaB) and in 6 ‘P’s in AI Pods (6P) are becoming shorter and more readable. Paying attention is working.
AAaB posts are more readable than the more technical 6P posts. This makes sense. However, the metrics are a bit contradictory. In both publications, in May:
The average Grade Level went down (good direction).
The average Reading Ease score went down slightly (wrong direction).
agileTeams (aT) posts — basically, these retrospectives — became less readable in May, but are still below 10th grade level.
A - Actions
My self-assigned actions from the end of April, and their statuses, were on pursuing audio voiceover, and working on improving my writing.
Audio Voiceover Experiments
(A5) Install an audio recording tool, set up a better-quality microphone, and verify audio quality (Status: Done).
(A6) Look into ethical AI-based text-to-speech tools that allow me to train a model on my own voice and use it to create my audio voiceovers (Status: Deferred).
(A5) was already done when the retro was published. For (A6), I still need to do some more diligence on Speechify and ElevenLabs, and compare the audio quality. My initial experiments with the 4 manual voiceover recordings weren’t bad. I didn’t need as many takes as I feared. So I’m going to keep the AI voice cloning tool on the back burner for a bit longer, and wait for feedback on whether people find the audio voiceovers useful (or not). I don’t want to invest scarce effort and funds into adding audio with voice cloning if my subscribers don’t value having audio!
Improving My Writing
I had self-assigned 2 new actions for May towards improving my writing.
(A7) Enable readability statistics, realtime checking for conciseness, and additional refinement checks in Word on my primary laptop (Status: Done).
(A8) Add readability metrics and average sentence length to my post log for past and upcoming articles (Status: Done, will be ongoing).
(A7) was already done when the retro was published, and (A8) is now done.
I’m now closer to my target of 50-50 mix of short and long articles. Overall, 60% of my 65 March-May posts are Very Short to Short (under 800 words).
New self-assigned Actions are below.
R - Review (Retrospective)
I committed to do a monthly review at the end of May to see where I stand on my goals and actions - status: this article is it 😊
Bottom line:
I made good progress on my Actions.
I reduced my WIP by finishing and posting some draft content I’d already written.
Thanks to using my WIP and exceeding my planned level of effort, I exceeded my posting targets for May.
I think readability is improving, but I want to look for a better measuring tool.
I’ll reassess my targets and plan in my next monthly retro.
New Targets for June 2024 (month 4)
Since my Memorial Day holiday week in May was disrupted by events, I’m expecting to take some time off in June to make up for it. But I really want to get the next section of PART 3, and its company profiles, out the door. My June goals are:
15-20 Posts in a mix of lengths:
1 Medium agileTeams retro post (this is it)
1-2 Long 6P posts and another 10 Medium pages on ethics of genAI for music (WIP sections of Part 3, plus detailed company profiles)
3-7 Short or Medium posts in Agile Analytics and Beyond - topics unplanned; on whatever I get inspired to write about
24-32 Notes or Comments (size not measured, assumed to be Very Short): 6-8/week for 4 weeks (assumption: I can do these on mobile even when traveling)
My effort budget for May posts will be 6-8 hours/day, or 30-40 hours/week, for 3 weeks of June. This is still well below the target of 50-60 total hours/week identified in a recent post5 about a successful Substack author and their effort level.
New Actions for June 2024
April-May Lesson 1: Connecting better with readers - Notes and Voiceovers
(A9) Change my publication process when adding voiceovers: During June, when I want to add voiceovers, I’ll finish the post, record and add the audio, and then send out the post to my subscriber list.
Carving out quiet time for recording will probably mean a delay of a day or more before the post is published, though. So I may not do audio on time-sensitive posts on AI or ethics. (If I had my own voice clone ready to use, this delay could be shorter.)
April-May Lesson 2: Becoming a better writer: readability
Monitoring length and readability of my posts seems to be of value. But using Word for calculating readability (1) has a lot of shortcomings, and (2) is kind of a tedious manual process. I’d like something much more automated and more suitable. I’m assigning myself a followup action:
(A10) Investigate alternatives to Microsoft Word for calculating readability statistics on my posts.
Look for a forthcoming post about readability metrics. In the meantime, I’ll keep using Word and tracking my data.
May Lesson 3: Improving UX for my readers
I use lots of charts and graphs and tables in my posts. But the Substack editor doesn’t really support them well. PNG images are an option, but content in PNG images can’t be explored or sorted or searched easily. I’ve given myself an action to:
(A11) Find a better way to embed interactive data visualizations into my posts.
As this May post explains, I investigated Datawrapper.de and mostly like what I see. I am going to start using it on posts in June.
“It does not matter how slowly you go, as long as you do not stop.” (Confucius)
What’s Next
That’s my writing retrospective for May. I commit to doing another one after the end of June. I’d love to hear your thoughts on any of this!
Thanks for reading agile Teams! Subscribe for FREE to receive new posts (about monthly).
References
End Notes
Archives analyzed in this retrospective
Substack:
LinkedIn:
the agileTeams page
Posts with audio voiceover added in May
Example: The math is off in pre-publication Readability Statistics for this post (no footnotes).
111 sentences / 100 paragraphs = 1.1 sentences per paragraph, not 2.1.
2177 words / 111 sentences = 19.6 words per sentence, not 15.2.
11860 characters / 2177 words = 5.4 characters per word, not 4.4.
How Flesch Reading Ease and Flesch-Kincaid Grade Level are calculated (readable.com).