Hi Briana,
Thanks for putting this together. I love seeing the results each year.
To give better feedback and think about the respondent's UX I went ahead
and answered
the whole thing in two passes: one my own and one with a coding agent, to
get more accurate/proven
answers on some aspects (versions, instance and DAG counts, etc).
Sharing a few observations:
- Consider letting every question take an optional free-text note
and an explicit "skip" / "I don't know", even for the multiple-choice ones.
A few of my own
answers were misleading as a bare choice and only became accurate once I
could add a
sentence. For instance, none of the upgrade-frequency options matched how
we actually decide
("we upgrade when it's worth it"), and for "how important is Airflow" I
wanted to understand
whether it meant my team or the whole company. An additional note captures
nuance that
might otherwise be lost (Self contained explanation of the reasoning [1]
<https://alexhans.github.io/posts/add-a-skip-and-a-note.html>)
And a few smaller ones:
- It would help to define the scope of "your" consistently (business /
organization / team /
company). At a large company those differ a lot, and a few of the
company-wide counts were
hard for me to answer meaningfully. Even a one-line note on how to
interpret them would do it.
- For "longest running task", I wasn't sure if it meant total wall-clock or
active compute,
or whether deferred / sensor wait counts. For me that was the difference
between a few hours
and well over a day.
- "How often do you deploy new DAGs" I read as brand-new DAGs, though it
could also mean
redeploys / changes, which are quite different rates.
- I wasn't sure how to map the DAG-distribution options onto a
managed-service (S3) setup.
- For "additional tools to write DAGs", I wasn't certain what counts as a
"framework" (e.g.
dbt-to-DAG code/dbt astronomer cosmos).
- The "challenges delivering value" question felt closer to my personal
pain points than to
Airflow itself so I hesitated. Airflow is definitely not my biggest pain
point and more of a joy
to work with :)
- The dev-list participation scale could use a "sometimes participate"
option between
"follow" and "actively participate".
- For "do you contribute", intent, frequency, and number of PRs felt a bit
tangled together.
- On the feature checklist, I wasn't sure what an "Other" would be.
- A short "Useful links" section at the end (docs, AIPs, Slack, GitHub,
mailing list) could
help anyone who hits a term they don't know, and point newcomers toward
how to get involved,
since the survey gets them thinking about it.
On the agent pass: a chunk of the technical questions are tedious to answer
by hand, such as
the Airflow version, number of schedulers, active DAG counts, the maximum
tasks in a single
DAG, and which features are actually in use. An agent with read-only access
could answer
those with proof, and it corrected a couple of my own guesses (e.g.
scheduler count).
Thanks again for driving this.
Best,
Alex
[1] - https://alexhans.github.io/posts/add-a-skip-and-a-note.html
On Tue, Aug 18, 2026 at 11:48 PM Niko Oliveira <[email protected]> wrote:
> Thanks for putting this together, Briana!
>
> I left a few comments on the doc :)
>
> Cheers,
> Niko
>
> On Tue, Aug 18, 2026 at 6:00 AM Michael Mabin <[email protected]> wrote:
>
> > Hi Briana,
> >
> > Thanks for putting this together. A few answer-choice suggestions:
> >
> > - Add “About the same” to the question about expected use in external
> or
> > revenue-generating solutions.
> > - Make numeric ranges non-overlapping: production instances
> > (“11–20”/“20+”), maximum DAG tasks (“1,001–2,500”/“2,500+”), and
> people
> > working with data or Airflow (“101–500”/“500+”).
> > - Add “Multi-cloud/hybrid” and “Not applicable” to the primary
> > cloud-provider choices.
> > - Add Go and Java to the non-Python language choices, and correct
> > “C+/C++” to “C/C++.”
> > - Consider neutral/uncertain choices for the career-impact question;
> the
> > current answers require a strongly positive or negative position.
> >
> > I’m also supportive of Astronomer sponsoring the survey and
> certifications,
> > provided the survey remains vendor-neutral and the raw results are
> > published as proposed.
> >
> > Best,
> > Michael
> >
> > On Mon, 17 Aug 2026 11:52:30 -0700, Briana Steel via dev
> > [email protected] wrote:
> >
> > Hey All,
> >
> > It’s that time of year again! The draft of the 2026 Airflow Survey is
> ready
> > for review.
> >
> > Here is the draft:
> >
> >
> https://docs.google.com/document/d/1NHn-tLoPYSPkPkNHAD1Bixjgvre6zUHBRAEZoaBEBg0/edit?usp=sharing
> >
> > Before leaving feedback, please note:
> >
> > -
> >
> > Please only leave comments. Do not resolve comments or make changes
> > directly to the draft.
> > -
> >
> > The most valuable feedback is on the answers, not the questions. Our
> > goal
> > is to keep answers aligned with the current state of the project.
> > -
> >
> > This survey is quite long and we are asking many community members to
> > complete it. If we add a question, we must remove another, so please
> > consider this before suggesting changes.
> > -
> >
> > Changing or removing questions makes it harder to compare data year
> over
> > year which can affect the overall analysis.
> > -
> >
> > Please get your comments in by Friday, August 21st, so we can launch
> the
> > survey by September 28th in time for Airflow Summit.
> >
> > Here are last year’s results for reference:
> > https://airflow.apache.org/blog/airflow-survey-2025/
> >
> > Astronomer is offering to sponsor this year’s survey, which includes
> > building the infographic and providing 2 free certification courses as an
> > incentive for community members to complete it. All raw results will be
> > posted as well. If folks agree, please let us know here. Or if that
> > requires further discussion, I can open up another thread.
> >
> > Best,
> > Briana Steel
> >
>
