Pint AI Product Updates: Week 10
A sneak peek to the front-end
Hello hello!
Exciting news — Pint AI now has an actual interface and we've finally started testing with partner data!
Bringing Pint to Life
In our last update, I sent a snapshot of our roadmap. Here it is again:
- Onboarding + Account linking flow ✅
- Settings (organization, workspace, team) ✅
- Elements page (core elements) — Release to design partners 🔄
- Elements page (custom elements) — Release to design partners
- Top Performing Boards — Release to design partners
- Comparison Boards — Release to design partners
- All boards + Overview page
- Reports page (saved reports + sharing) — Release to design partners
We will be building, iterating and releasing a set/group of features at a time, which will then be handed over to our design partners for testing. Since the first draft of our onboarding and settings is done, here's a quick peek of the login/sign up, explained by our front-end developer, Sahil:
You can see Sahil smile as he screws up confirming his password (it happens to me on a daily basis) 😅. There are a whole set of things to work on to make it clean like the copy of everything, the positioning of some elements, the scroller on the left, and a bunch of other things.
But hey, that's what we promised to show you—the good, the bad and the ugly!
The Testing Begins
The GIF totally matches what we know will happen; testing with partners' data will really put our ML algos to the test and we're going to get our ass kicked in more ways than one 🥲 But the earlier we figure out what's working and what isn't, the better.
A big chunk of our core elements are being readied to test with partner data from Monday; we have connected with one of their ad accounts, and will be running our algos on the ad data that we receive. We'll be looking to do a couple of things:
- Identify core elements—people, objects, messaging, themes and hooks
- Find relative accuracy
What I mean by the second point, is that accuracy has to be measured in different ways for different elements. For example, with people and objects, it's relatively straightforward — if a person and an object are present in all creatives, with what accuracy can we identify their presence, their positioning and their timing/duration.
On the other hand, with elements like themes, every team has their own version of what a theme means. This makes it hard to measure objectively for accuracy. What we'll be doing is to see if our classification of a theme makes sense for the team. It's not really accuracy as a %, as much as it is — would you classify it in this manner?
And that's what I mean by relative accuracy. But we've begun this process, and we should get some answers soon!
If you have any comments, questions or ideas for us at any stage — even the super simple login process — we'll be more than happy to take them up. You can just reply to this email (it comes straight to my inbox) or write to Swagam as well at swagam.dasgupta@pint-ai.com.
Have a great end of the week!
Sourya Reddy
Co-founder, Pint AI
