Using AI to Turn Website Traffic into Investable Signals

Manus just launched an integration with Similarweb. If you're a software investor, this is a big deal.

Back when I worked at SoftBank, I used Similarweb constantly when evaluating growth equity names. Website traffic. Engagement metrics. Geographic breakdowns.

I still run this exact playbook today on my public equities book. It's one of my go-to techniques for software names.

This is the alternative data that helps you front-run earnings surprises before the Street catches on.

The problem? Similarweb is painfully manual.

Pulling data for one company is tedious. Doing it across your entire watchlist? Forget it.

And even when you do pull the data, most investors don't know what to do with it.

Is a 5% bounce rate increase bad? What does declining visit duration actually mean for LTV? How do you separate seasonal noise from real cohort deterioration?

This Manus integration finally lets you scale it.

Here's what I reverse engineered into 1 singular prompt:

1. Pull granular engagement metrics (pages per visit, visit duration, bounce rate, repeat visitors)
2. Slice the data by geo and device to find where quality is improving or deteriorating
3. Use engagement trends as leading indicators for LTV and monetization... before it shows up in the financials

Pod shops and growth equity funds have been doing this manually for years. AI lets you run it in minutes.

I'm going to show you how I used this to assess Airbnb's traffic quality as a proxy for cohort health and LTV trajectory.

The Prompt

I have the prompt below. All you have to do is this:

1. Go to Manus​
2. Fill in the appropriate variables in my prompt below
3. Run the prompt
4. enjoy :)

Use Prompt Here: Link

The Result

Full output here: Link​

PDF Similarweb report here: Due_Diligence_Report_Airbnb_Engagement_Quality_(Q4_2025).pdf​

The headline: Airbnb's engagement quality is stable with positive momentum.

Here's what AI surfaced from the Similarweb data:

The bullish case:

• Visit frequency improved every month of Q4 (1.400 → 1.408 → 1.413). Users are coming back more. That's a direct LTV signal.
• 64.7% direct traffic on desktop. Nearly two-thirds of users are typing "airbnb.com" into their browser. No paid acquisition needed. This is brand moat in action.
• 25.2% cross-device users. These are high-consideration buyers who research on desktop and book on mobile. Higher intent. Higher LTV.

The bear case:

• International engagement is a problem. India has a 66% bounce rate. Canada is at 52%. France at 48%. If management is pitching international expansion as a growth lever, these numbers say otherwise.
• The U.S. is 78.5% of traffic. Concentration risk. The core market is performing exceptionally (29.5% bounce, 15.5 pages/visit). But replicating this abroad? The data says they haven't cracked it.

What I would have missed without AI:

The geographic dispersion analysis. Manually pulling bounce rates by country across a full quarter would take hours. Manus did it in minutes and immediately flagged the international engagement gap as an LTV risk.

This is exactly the type of signal that shows up in engagement data 1-2 quarters before it shows up in reported revenue mix.

Why this matters beyond Airbnb:

This Manus + Similarweb integration unlocks a whole toolkit of techniques I've been running manually for years. Now you can scale them.

A few ideas to try:

• Intra-quarter demand nowcasting: Track visit trends against reported revenue to build your own beat/miss model. When traffic diverges from consensus, you have a data-driven basis to size conviction.
• Channel mix as CAC proxy: Rising paid traffic share with flat engagement? That's a red flag for deteriorating unit economics. Rising direct and organic? Brand strength that justifies a higher multiple.
• Competitive share tracking: Define a peer set, track traffic share by geo and device, and test whether management's "share gain" narrative is real.
• Product launch validation: Did that flagship feature actually move the needle? Track traffic to specific pages post-launch. Fast fade-outs = disappointing adoption before the company admits it.

The same techniques pod shops and growth equity funds have been running manually... now available to anyone willing to write the right prompt (which I already did for you above haha)

Try it on your own watchlist and let me know what you find.

Personal:

I'm planning out course and content calendar for 1Q-2Q ...

Any requests for investment styles or asset classes?

Up on the docket:

• Fixed Income
• More niche investment styles like merger arb
• Private Equity
• Investment Banking
• How to hack job interviews with AI
• Sales and fundraising using AI

Curious what you do for your investing style .... I assume most folks who read my newsletters are equity guys so I've been equity biased but I'll do other investing styles if there is interest!