Collaborative Experiment Tracking Product Launch Strategy

NeroOkwa commented 1 year ago

Context

This captures the product launch activities for the new collaborative experiment tracking feature #1295 on Kedro-Viz. This includes preparing a launch blog post, and presentation, and updating/reviewing the documentation.

Goal

The goal is to increase awareness and adoption of kedro experiment tracking, and Kedro-Viz, especially Kedro-Viz/Kedro users.

Target audience

This is targeted at technical (DS, DE, and MLE) users with a 2 prong approach:

Marketing to kedro users to use kedro-viz
Marketing to non-kedro users to use kedro/kedro-viz

Content

This would include:

What is experiment tracking?
What is collaborative experiment tracking?
Why would you choose to use Kedro experiment tracking vs other tools?

Plan/Checklist

Admin:

[x] Create Github issues to track product launch strategy and sub-activities - blog post, and presentation. Key Activities:
[x] Blog article - #73 - Introducing this new feature and it's value proposition for users. This would address the questions in the content section.
[ ] Presentation - #72 - prepare a showcase presentation for internal & external audiences on collaborative experiment tracking.
[x] Update Kedro-Viz docs #1353 with new feature.

stichbury commented 1 year ago

Thanks @NeroOkwa -- here's a link to the template to create a brief about your blog post.

Please could you take it as a guide and plan the post out as a comment here, so we can iterate and get a great structure ready for you and the team to write it up? cc @antonymilne @rashidakanchwala @tynandebold

NeroOkwa commented 1 year ago

What is the problem this blog post solves?

The problem this blog post aims to solve is directing users looking for a way for their team to find (visualise, compare, and collaborate on their pipeline runs) the best performing runs to this feature in Kedro-Viz.

Who will find this information useful?

DS, DE, MLE

What title would you choose?

Is your Experiment tracking tool multiplayer?
How to collaborate with your team on your next pipeline run.
Experiment tracking: the only way is collaborative.
Finding your best pipeline run, with your team.

What keywords are you trying to hit?

experiment tracking, mltracking, machine learning, pipeline run, data science pipeline

How long will this post be? For example, a short 3 minute read, a 20 minute hands-on tutorial, a 10 minute thought-leadership piece.

10 minutes intro and how-to tutorial

Introduction

Covers the pain points of ingesting, processing, modelling and tracking your data (the data flow) to get the best parameters, especially when working within a team of DS, DE, and MLE. Explains that we will show how to optimise your workflow within Kedro-Viz.
- CTA "What is Kedro-Viz?"

Narrative subsections

Section 1 - What is Kedro Experiment Tracking?
Section 2 - Get started with collaborative experiment tracking - Explain that this is being added in stages, and this is about the first version.
- Collaborative experiment tracking: version 1 - Brief set of features
Section 3 - How to use Kedro's collaborative experiment tracking (tutorial).
Section 4 - Kedro vs MLFlow and Neptune
Section 5 The future of Kedro's collaborative experiment tracking
- Explain future plans
- CTA: Find out more about Kedro

@stichbury @tynandebold @rashidakanchwala @antonymilne

stichbury commented 1 year ago

Thanks for this, it's a nice summary. Can I suggest a minor revision for the narrative, something like this:

Introduction

Covers the pain points of ingesting, processing, modelling and tracking your data (the data flow) to get the best parameters, especially when working within a team of DS, DE, and MLE. Explains that we will show how to optimise your workflow within Kedro-Viz.

CTA "What is Kedro-Viz?" -- I'll write you a short CTA box to include here

What is experiment tracking?

Very brief explanation and pointer to old blog Also introduce collaborative experiment tracking here -- it's still experiment tracking just with collaboration added in. Explain the added value of sharing results (maybe try to get a quote here from a user to say why it's valuable).

Get started with collaborative experiment tracking on Kedro

Explain that this is being added in stages, and this is about the first version.

Kedro experiment tracking: version 1

Brief set of features

How to use Kedro's collaborative experiment tracking

@rashidakanchwala has a good point -- let's get into the tutorial asap

Kedro vs MLFlow and Neptune

Add a short section of comparison here

The future of Kedro's collaborative experiment tracking

Explain future plans

CTA: Find out more about Kedro (this is written already)

stichbury commented 1 year ago

Can I request the following as you write the post:

Please create all screen grabs/videos in the new branding to save updating them
Please follow the Kedro style guide

Also, (edited to add) please can you add a section that explains the reason for using Kedro for experiment tracking rather than MLFlow or Neptune (which has the collaborative feature, I think). I have edited my note above to add where I think this would go.

tynandebold commented 1 year ago

Closing this now, as the last task has it's own ticket here: https://github.com/kedro-org/kedro-devrel/issues/72.

kedro-org / kedro-devrel