Pinned Tweet
Companies invest a lot in analytics - but are these investments valuable?
@IsraeliAyelet and I studied ~1,500 online retailers and found that using a descriptive dashboard increased their weekly revenues by 4%-10%.
>>
#MarTech #BigData #Analytics #ecommerce #DataScience
ALT SynthDiD estimate of ATT of adopting analytics dashboard by ecommerce retailers
Rock ballads in Hebrew have amazing lyrics.
Here's "put a candle in the window" by Stella Maris. I particularly like "put a candle in the window, and we will dance the dance of souls".
youtu.be/52cCWQbLHWQ?si=pESu…
‼️News publishers that block LLM access receive even less traffic compared to not blocking‼️
In a new working paper with @HangchengZhao
we use daily traffic data , human web-browsing data, historical robots.txt policies, job postings, and historical webpage content to document early shifts in online news production and consumption.
We ask:
❓How do LLMs impact news publishers? Does consumer traffic decline? Do publishers hire fewer employees or generate more news slop?
❓How do publishers respond to LLMs? Does blocking LLM traffic have a positive or negative impact?
Our findings:
1️⃣ Traffic decline wasn’t immediate. Publisher traffic remained broadly stable through mid-2023; a Synthetic Difference-in-Differences analysis shows a decline in traffic that started after August 2024 (−13%) compared to top retail websites.
2️⃣ Blocking GenAI crawlers may backfire. ~80% of the top publishers block LLMs using robots.txt.
Staggered DiD analysis of the effect of blocking shows that traffic declined after blocking in both SimilarWeb traffic data (−23%) and Comscore data (−14%).
3️⃣ No “newsroom job replacement” yet: Data from job postings show that editorial/content roles do not exhibit a post-GenAI collapse; replacement or reduction of these jobs in the near term appears limited.
4️⃣ Publishers aren’t scaling text volume; instead, they’re “richening” pages. The number of interactive elements rises (+68%) and ads/targeting tech increases (+50%) , while article volume declines relative to top retail websites used as controls.
Key Takeaway: The impact of LLMs is nuanced and sometimes surprising. For example, blocking LLM access using robots.txt may reduce both bot access and real audience demand.
Links to working paper:
SSRN: papers.ssrn.com/sol3/papers.…
arXiv: arxiv.org/abs/2512.24968
#LLMs #media #news #AI #ArtificialInteligence #DigitalMarketing #digitalmedia
4️⃣ Publishers aren’t scaling text volume; instead, they’re “richening” pages. The number of interactive elements rises (+68%) and ads/targeting tech increases (+50%) 🎯📈 , while article volume declines relative to top retail websites used as controls.
Rest in peace Itay.
יום 760
איתי חזר הביתה
Itay is back home
@hagitchen @bringhomenow
@POTUS @SEPeaceMissions @SebGorka @SecRubio
The NY times seems to have some real issues reporting the news.
"After two soldiers were attacked and killed by Hamas" is what really happened...
New paper alert: Most firms believe better and more precise data = better strategy.
My new paper with @heller_yuval shows this is not always true: sometimes biased algorithms make firms more profitable.
>>
Takeaway:
- Bias in estimates isn’t always a bug — sometimes it’s a feature.
- In oligopoly settings, “naive” analytics can sometimes beat sophisticated econometrics or unbiased A/B tests.
- That complicates both firm strategy and antitrust policy.
Full paper here (open access):
“Naive analytics: The strategic advantage of algorithmic heuristics”
Games and Economic Behavior (2025)
sciencedirect.com/science/ar…