Made with Humane Club
Handling Local news in personalization models can be challenging.
CopiedMedium will de-amplify Generative AI generated content from its platform and will actively evolve its personalization system to spot and de-amplify AI written content.
CopiedDriving factors: Differentiation
Differentiation typically comes from building authenticity on your core product, i.e., editorial product. This involves owning more of the value chain:
- Are you a trusted brand or voice for a well-defined topic?
- Are you able to set the agenda for a well-defined audience segment?
- How strong is your recall among that well-defined audience segment?
- Have you stitched the most active members of that audience segment into a tight knit community?
- Is the community co-building with you or via you?
- How are you ensuring that you are not only building but also retaining 1:1 relationship within that community?
- What’s your information architecture and content strategy?
- How deep are you serving that vertical?
- Is your website (yes! a website even after GenerativeAI) representative of who you are?
Amplification factors
While the above gets your driving factors in place, the next aspects define how much you can amplify. This includes:
- Distribution
- Ability to sell advertisements, subscriptions, or events
- Ability to market and develop audience
CopiedThere are various fundamental aspects of an editorial product business. Most of these businesses have been edged out of the amplification factors:
Distribution: We’ve lost distribution to SEO, Social, platforms like Substack, etc. The only owned media channels that exist now are: website, mobile app, and newsletters.
Selling advertisement: Companies find it more efficient to buy advertisements from ad-networks on search and social than directly via publications.
CopiedAdvertisements in print and TV give much more revenue than digital.
| Print and TV | Digital | |
|---|---|---|
| How is it sold? | Advertisements in print and TV are sold directly by the news company. Given that, 100% of the revenue belongs to them. | Online, most quantity of advertisements is sold indirectly by ad networks. Given that, the publisher loses out 30-40% of ad revenue to the ad network. |
| How is reach measured? | In print and TV, it is assumed that the entire circulation (MAU) is reading. | Online, one can precisely calculate the DAU, which for most news websites turns out to be only 3-8% of MAU. |
| Brand Premium | There’s is a prestige to advertise with big news brands. | It isn’t feasible for brands to select that they want to advertise with a specific publisher. |
| Bid up or down | Sometimes companies will bid up to block out competitors from advertising on specific print and TV channels. | Ad engines force publishers to bid down, i.e., whoever has the lowest price will get the ad. |
| Implication on content | Premium content is valued. | All content is measured against page view. Hence, a detailed investigation is measured against a cat video. |
CopiedHow to think about AI for news?
When Steve Jobs returned back to Apple in 1997, he famously cut the product line by 70% down to 2 desktop devices and 2 portable devices!
Rule of thumb: If you can’t explain it and choose for yourself, then surely your customers wont!
- Don’t put dollars behind scaling to scale until you’ve completed value proposition discovery.
- Don’t put dollars behind marketing until your communication is so simple that everyone understands.
- Often we overcomplicate stuff that no one really understands.
CopiedAvoid phrasing decisions as binaries. Reality is nuanced. Hence, it is better to phrase decisions as “We want to do X while maintaining or growing or reducing Y.”
CopiedIt is always advisable to have one major source of revenue and then diversify by having multiple minor streams of revenue that de-risk you from that major source of revenue. For example, YouTube’s revenue split is 1:3 between subscriptions and advertisements.
Why it matters:
- It protects against risk of ruin.
- It protects against ups and downs. For example, revenue from advertisements is seasonal.
How
There are different forms of revenue models:
- Digital Advertisements, where revenue is Sessions Per User x Users x Pages Per Session x Ad Impressions Per Pageview.
- Subscription, where revenue is Average Revenue Per User (ARPU) x Active Subscribers (i.e., Existing Subscriber Base + New Conversions + Renewal)
- Affiliate and deals
- Events
- Sell playbooks, courses, trainings, etc. using micro-transactions
- E-commerce: For example, launch branded products
- Offer services like Job Boards
Downside
Some of these revenue models can be in opposition to each other and make commoditized businesses complicated. This in turn mandates the need for propensity models.
Copied
Copied
Allocate 5% to 10% of the content on which Artificial Intelligence algorithms and multi-arm bandit can test out different SEO hacks and observe causality, i.e., is it resulting in an improvement in the SEO score of those pages.
Why it matters: Currently, SEO is a cat and mouse game with Google.
- Once someone tests a new rule once and it works, no one returns back to it and validates if that rule is still valid. Over a period of time, you’ll end up with a dead set of rules that may or may not be valid.
CopiedModel performance metrics
- accuracy
- precision
- recall
Copied
Sometimes the model can overfit the training data and hence underperforms with real data. To reduce complexity, we use regularization.
- Add a penalty factor on complexity i.e. number of features and values of weights.
- Lasso: Force coefficients to zero if not relevant. Ends up performing feature selection.
- Ridge: Reduces coefficients of irrelevant features but does not drop them to zero.
CopiedDecision trees are versatile tools that can be employed for both Regression and Classification tasks.
When developing a decision tree, it’s crucial to ask a series of targeted questions in order to zero in on the most accurate label. The objective is to construct an efficient tree that minimizes the number of splits necessary to divide the data. This process continues until no further splits can be made.
It’s important to find the right balance, as shallow trees tend to under-fit the data, while deep trees can lead to overfitting, with each example ending up as its own leaf node.
Copied