100% synthetic persona — no real PII
How personalized can a website get — and where does each fact come from?
One visitor, Maya Chen. Climb the tiers to see how much a site can know — from an anonymous landing page, to a Google sign-in, to a purchased data append. Every fact is tagged with its provenance: collected by us, declared, granted via Google, or bought. Then flip tasteful → creepy to see the difference a surface policy makes.
0 · none
Anonymous landing
1 · cookie
Returning visitor
2 · email
Identified — gave email
3 · google
Signed in with Google
4 · google
+ Purchased append
5 · google
+ Existing customer
6 · google
Full identity graph
We've seen this browser before (a cookie, or a device fingerprint that survives clearing it).
18of 52 facts knowable at this tier
0stated (say)
11used to steer (allude)
7known but withheld (hold)
2bought from third parties
Build a career in AI — nights and weekends.
Live classes that fit around a full-time job.
Here's the night cohort, with the financing options up front.allude (steer)
Classes run after work, from home — no commute.allude (steer)
→ See the next night cohort🛟 What the visitor never sees: We *know* far more (see the ledger) — schedule pressure, comparison shopping, a new baby, income. A surface policy keeps it out of the copy: we let it quietly shape what we feature, but we don't say it.
Where the data comes from
Six provenance classes. The count is how many facts each one yields at this tier. See the full paid/free source list on the enrichment catalog.
Collected ourselves (passive) 12
The HTTP request + a few lines of JavaScript, the instant the page loads. No login, no form, often no cookie. You give it to every site you visit.
💵 $0 (optionally a few ¢ for a reverse-IP or fingerprint lookup)
Our own database 4
Behaviour we logged on previous visits (cookies / device match) and records in our CRM. We already own it — no third party involved.
💵 $0 (we collected it)
Declared by the visitor 0
A field they filled in — a newsletter box, a lead form, an account signup. The cleanest data there is: they chose to tell us.
💵 $0
Granted via Google sign-in 0
"Sign in with Google" hands us OAuth scopes. Profile is one click; but the consent screen can also grant calendar, Gmail metadata, YouTube, contacts and location history — an enormous jump for one tap.
💵 $0 (the price is the permission)
Purchased (data broker append) 2
Match a name+email+address against a data broker (Acxiom, Experian, Epsilon, Oracle Data Cloud) and append what they've compiled: income, life events, demographics, propensities. Billions of attributes, sold per record.
💵 ~$0.05–0.25 per record matched
Identity graph (cross-device + offline) 0
An identity-resolution vendor (LiveRamp, Tapad) ties this browser to your other devices, your household, and offline data sold by retailers — loyalty cards, credit-card panels, smart-TV viewing. The complete picture.
💵 $$ platform subscription
Unlocked by reaching “Returning visitor”
- Our own database Visit history: 4th visit in 6 days ●●●●●
- Our own database What you looked at: 'Night cohort' ×3, pricing ×2 ●●●●●
- Our own database How you read: Read 90% of the financing FAQ, 2m11s ●●●●●
- Our own database Unfinished actions: Started the application, didn't submit ●●●●●
- Purchased (data broker append) Ad retargeting pool: Meta audience 'career-switch-warm' · shown 7 ads ●●●●●
- Purchased (data broker append) Comparison shopping: Visited 2 competitor bootcamps this week ●●●●●
The provenance ledger — every fact, where it's from, what we did with it
said stated literally · steer shapes selection, not stated · withheld known, held back · shown printed (creepy) · locked needs a higher tier
| Data point | Value (synthetic) | Where from | Cost | Creepy | Policy | Here |
|---|---|---|---|---|---|---|
| Approximate location Resolve the visitor's IP address to a city. |
Austin, Texas (metro) | Collected ourselves (passive) MaxMind GeoIP2 / IP2Location |
$0 (free GeoIP DB) | ●●●●● | allude | steer |
| Neighborhood Same IP lookup, at neighborhood resolution. |
~2 mi of ZIP 78722 (Mueller) | Collected ourselves (passive) IP geolocation (ZIP-level) |
$0 | ●●●●● | hold | withheld |
| Connection Map the IP to its owning network and connection type. |
Spectrum cable · residential | Collected ourselves (passive) IP → ISP / ASN lookup |
$0 | ●●●●● | allude | steer |
| Company (reverse-IP) Match the IP to a company's network to de-anonymize B2B visits. |
(residential — no company match) | Collected ourselves (passive) Clearbit Reveal / KickFire |
~$0.01 / lookup | ●●●●● | allude | steer |
| Device & OS The browser announces device, OS and version on every request. |
Apple iPhone 13 · iOS 17.4 · Safari | Collected ourselves (passive) User-Agent + Client Hints |
$0 | ●●●●● | allude | steer |
| Device economics Infer age and price bracket from the device model. |
≈2-yr-old, non-Pro model | Collected ourselves (passive) Model → release-date + price tier |
$0 | ●●●●● | hold | withheld |
| Screen & theme A few JS properties read on load. |
390×844 · dark mode · battery-saver on | Collected ourselves (passive) JS: screen, prefers-color-scheme |
$0 | ●●●●● | allude | steer |
| Battery level JS reads the device battery level and charging state. |
18% and dropping (not charging) | Collected ourselves (passive) Battery Status API |
$0 | ●●●●● | hold | withheld |
| Local time The browser's clock and timezone, read in JS. |
11:47 PM, Tuesday | Collected ourselves (passive) JS Date + IANA timezone |
$0 | ●●●●● | allude | steer |
| Languages The ranked language list your browser sends. |
en-US, then zh-CN | Collected ourselves (passive) Accept-Language header |
$0 | ●●●●● | hold | withheld |
| Where you came from The link that sent you carries the campaign + creative. |
Instagram ad · campaign 'career_switch_q3' | Collected ourselves (passive) Referrer + UTM parameters |
$0 | ●●●●● | allude | steer |
| Device fingerprint Hash your canvas, fonts and GPU into an ID that survives clearing cookies. |
fp_9f3a… (canvas+fonts+GPU hash) | Collected ourselves (passive) FingerprintJS |
~$0.005 / match (pro tier) | ●●●●● | hold | withheld |
| Visit history Tie this session to prior ones we logged. |
4th visit in 6 days | Our own database First-party cookie / fingerprint |
$0 | ●●●●● | allude | steer |
| What you looked at Every page, in order, time-stamped. |
'Night cohort' ×3, pricing ×2 | Our own database Site analytics |
$0 | ●●●●● | allude | steer |
| How you read JS records scroll position and time on each block. |
Read 90% of the financing FAQ, 2m11s | Our own database Scroll-depth + dwell tracking |
$0 | ●●●●● | allude | steer |
| Unfinished actions Partial form state captured field-by-field, before submit. |
Started the application, didn't submit | Our own database Form analytics |
$0 | ●●●●● | allude | steer |
| Ad retargeting pool A 3rd-party pixel adds you to ad audiences across the web. |
Meta audience 'career-switch-warm' · shown 7 ads | Purchased (data broker append) Meta Pixel / Google Ads tag |
ad spend | ●●●●● | hold | withheld |
| Comparison shopping Cross-site browsing bought from a data-management platform. |
Visited 2 competitor bootcamps this week | Purchased (data broker append) DMP / cookie-sync (Lotame, Oracle BlueKai) |
$$ subscription | ●●●●● | hold | withheld |
| Name They typed it into a field. |
Maya | Declared by the visitor Newsletter / lead form |
$0 | ●●●●● | say | locked |
| Email They typed it. |
maya.chen@gmail.com | Declared by the visitor Form field |
$0 | ●●●●● | say | locked |
| Stated goal They wrote it in their own words. |
"switch into AI without going broke" | Declared by the visitor Free-text form field |
$0 | ●●●●● | say | locked |
| Linked photo & accounts Hash the email and look it up across services. |
Gravatar photo + 9 sites tied to this email | Collected ourselves (passive) Gravatar / hash lookup |
$0 | ●●●●● | hold | locked |
| Breach exposure Check the email against breach corpora. |
Appears in 3 known breaches | Collected ourselves (passive) Have I Been Pwned |
$0 | ●●●●● | hold | locked |
| Verified name & photo One click grants name, photo, locale. |
Maya Chen + verified profile photo | Granted via Google sign-in Google OAuth · profile scope |
$0 (consented) | ●●●●● | say | locked |
| Verified email + recovery Verified address, and that a recovery phone exists. |
maya.chen@gmail.com (verified) · recovery phone on file | Granted via Google sign-in Google OAuth · email scope |
$0 (consented) | ●●●●● | allude | locked |
| Account maturity Account age and activity hints. |
Google account since 2009 · 'power user' | Granted via Google sign-in Profile metadata |
$0 | ●●●●● | allude | locked |
| Your calendar The consent screen can include calendar read — most people don't notice. |
'OB checkup Thu 2pm', 'daycare tour Sat' | Granted via Google sign-in Google OAuth · calendar.readonly |
$0 (one extra checkbox) | ●●●●● | hold | locked |
| Inbox metadata Senders + subjects reveal purchases without reading bodies. |
Receipts from Pampers, BuyBuyBaby, a fertility clinic | Granted via Google sign-in Google OAuth · gmail.metadata |
$0 | ●●●●● | hold | locked |
| Watch history Watch + search history as interest signals. |
Recently: 'newborn sleep', 'career switch at 34' | Granted via Google sign-in Google OAuth · YouTube Data API |
$0 | ●●●●● | hold | locked |
| Location history Months of timestamped places. |
Home: Mueller · work: downtown · 2 hospital visits this month | Granted via Google sign-in Google OAuth · Maps Timeline |
$0 | ●●●●● | hold | locked |
| Contacts graph Your whole address book, with labels. |
1,840 contacts · partner 'Jordan', an OB-GYN, your mom | Granted via Google sign-in Google OAuth · People API |
$0 | ●●●●● | hold | locked |
| Household income Append modeled income to a name+address match. |
Modeled $115–135K band | Purchased (data broker append) Experian / Acxiom income model |
~$0.05–0.25 / record | ●●●●● | hold | locked |
| Home & residence Public deeds + broker compilation. |
Homeowner · ~$540K home · 4 yrs in residence | Purchased (data broker append) Property records / Acxiom |
bundled in append | ●●●●● | allude | locked |
| Net worth band Modeled from assets, home, credit. |
$250–500K | Purchased (data broker append) Experian wealth model |
bundled | ●●●●● | hold | locked |
| Life event: separation Brokers sell change-of-status triggers as they happen. |
Trigger: 'recently separated' | Purchased (data broker append) Epsilon / Experian life-event triggers |
premium trigger | ●●●●● | hold | locked |
| Life event: new baby New-parent is one of the most-traded triggers. |
Trigger: 'new parent', infant 0–6mo | Purchased (data broker append) Epsilon life-event triggers |
premium trigger | ●●●●● | hold | locked |
| Vehicle Registration + service data compiled and sold. |
Drives a 2021 Subaru Outback | Purchased (data broker append) Oracle Data Cloud / Polk auto |
bundled | ●●●●● | allude | locked |
| Education & occupation Compiled demographics. |
BS · occupation: software/IT | Purchased (data broker append) Acxiom InfoBase |
bundled | ●●●●● | allude | locked |
| Political profile Voter rolls + modeling, sold for targeting. |
Leans Democrat · high turnout · past donor | Purchased (data broker append) L2 / Aristotle voter file |
voter-file license | ●●●●● | hold | locked |
| Health ad audiences Inferred condition audiences sold for ad targeting. |
'Expectant/new parent', 'seasonal allergy sufferer' | Purchased (data broker append) Health-adjacent ad audiences |
audience license | ●●●●● | hold | locked |
| Ethnic affinity Name + geography modeled into an 'affinity'. |
Modeled: Chinese | Purchased (data broker append) Acxiom ethnic-affinity model |
bundled | ●●●●● | hold | locked |
| In-market signals Predicted near-term purchases. |
Minivan, baby gear, term life insurance | Purchased (data broker append) Bombora / Oracle propensity |
subscription | ●●●●● | hold | locked |
| Customer history We already have an account record. |
Took 'Intro to Python' with us, 2023 | Our own database Our CRM |
$0 (we own it) | ●●●●● | say | locked |
| Spend & support Lifetime value and support load. |
LTV $349 · 1 purchase · 3 support tickets | Our own database CRM / billing |
$0 | ●●●●● | allude | locked |
| Risk segment A score we computed on our own data. |
Churn-risk 0.31 · 'price-sensitive' | Our own database Internal model |
$0 | ●●●●● | hold | locked |
| Payment on file Stored from the last purchase. |
Visa •••• 4242, exp 11/26 | Our own database Billing system |
$0 | ●●●●● | hold | locked |
| Advocacy Your own feedback to us. |
NPS 9 · left a public 5★ review | Our own database Survey + reviews |
$0 | ●●●●● | say | locked |
| Cross-device Deterministic + probabilistic linking of all your devices. |
This iPhone + a work MacBook + a home iPad = one you | Identity graph (cross-device + offline) LiveRamp / Tapad |
$$ subscription | ●●●●● | hold | locked |
| Household Devices + addresses clustered into a household. |
2 adults (Maya, Jordan) + 1 infant | Identity graph (cross-device + offline) LiveRamp household graph |
$$ | ●●●●● | hold | locked |
| Offline purchases Loyalty-card baskets sold and matched to your identity. |
Target loyalty: diapers + formula weekly; Whole Foods 3×/wk | Identity graph (cross-device + offline) Retail loyalty data resold to brokers |
$$ | ●●●●● | hold | locked |
| Card-spend panel Anonymized-then-rematched card data sold to marketers. |
Baby-gear spike; $0 restaurants since April | Identity graph (cross-device + offline) Credit-card transaction panels |
$$ | ●●●●● | hold | locked |
| Smart-TV viewing Your TV reports what's on screen, frame-matched. |
Heavy HGTV + late-night cartoons | Identity graph (cross-device + offline) Smart-TV ACR (Samba, Vizio Inscape) |
$$ | ●●●●● | hold | locked |
Text only, for now — the same signals would drive the layout next: reorder sections, swap hero imagery, change the offer. The provenance ledger is the point — nothing reaches the page without a receipt for where it came from.