Working with people through feedback from their core reports prepares these same activity streams for more processing and discovery. We surface patterns and click streams, create new shortcuts beyond the defaults, apply more models, Machine Learning and AI.
The models generate more recommendations from this additional processing. People review and refine the recommendations made and accept what makes sense within their flow. These refinements inform the next passes of additional processing, and the iterative improvements continue while people see the benefits.
Industry-specific file types and leveraging previous work and workflows from yourself and others to avoid creating new one-offs
Today’s Small, Good Thing involved a CD with MRI images from a radiologist’s office. It’s labeled “For Physicians Only,” and while I’m not a doctor, I’ve played one on IRC. The files didn’t open with the default image viewer, but their properties showed they’re all “DICOM image (application/dicom).” A quick search led me to Aeskulap – DICOM Viewer package which installed quickly & easily from standard apt repos. Aeskulap opened the MRI images and let me cycle through them like flipbook videos which I suppose is as-intended.
Small, Good Thing was (re-)discovering Ubuntu’s native screencasting tool via Ctrl+Alt+Shift+R. This captured :30 of my flipping through Aeskulap images at 3840×1080 resolution. Small, Good Thing was ffmpeg’s ability to crop and resize the screencast webm file and spit out a reasonably-sized gif in real-time.
It took me a minute to get the crop where I wanted, and I grabbed the gif output params from earlier work. All-in it was less time with fewer distractions than it’d take me in Blender. It’s great and greatly-appreciated that people put so much good work into things and make it available.
Artist community marketplaces let us solicit and manage custom contributions at scale. Each artist gets a custom experience. Artist teams see what’s relevant in one place. Brands, sponsors, and co-marketing partners see campaign-level activities, assets and reporting approved by artist teams. Fans get exclusive access and clear pathways to earn more rewards through their participation.
We target outbound assignments and requests for short- and long-form contributions from artist teams, staff and UGC.
We handle inbound responses, review & approval workflows, and delivery by artist, team, and campaign. Approval workflows are as high-touch or hands-off as the diverse roster of artists demand.
It’s a framework for successful collaborations with influencers and agencies and lets us discover and develop more talent of our own.
It’s not news repeat customers yield returns on multiple levels. The last 18 months highlighted a few of these and given smart brand teams good lessons to apply over the next six quarters. Repeat customers adding directly to the top and bottom lines is good in itself. Loyal customers also
Spend more overall
Build MRR base
Solidify next quarters’ projections & planning
Will spread the word, especially when given incentive to do so
We see similar trends and markers in tens of thousands of transactions across retailers and programs also surface within single shops and campaigns. Explicitly showing these returns on loyalty working together and amplifying each other gives teams permission to focus more on ways to treat customers as the very important people they are.
elasticsearch with fscrawler handled these well for my needs.
Low-touch / unattended ingestion
Sensible, configurable defaults
Helpful, accurate dry runs
Non-catastrophic re-runs (i.e. smart enough to minimize overwriting or duplicating existing entries)
Customizable / scriptable input and output handling
File meta data capture
Full-Text indexing of file content
“Clever de-duping” is TBD, and starting with rsync or rclone helps there. elasticsearch runs lean enough and is straightforward enough to configure for my local dev env. Defaults are sensible, and re-building indexes is a matter of
fscrawler job_name --loop 1 --restart
These help balancing up-front config time with GIGO and “we’ll take care of it in post-production.”
Eye-balling ingestion and indexing processes is fine for seeing initial results, tweaking, and discovering more as index searches yield more results. Getting media consolidated and indexed locally was one set of goals met. Locating assets I needed for other work was another win, and Kibana surfaced more follow-up opportunities than log viewing alone.