Lat −23.5505Lon −46.6333Z 3
Physical-world intelligence

The grid of the real world.

HYPR turns the physical world into a layer of intelligence: places, movement, spending, attention and perception.

Talk to our teamExplore the platform
Hover over the mapScroll to explore ↓
We work with Latin America's 500 largest companies, including
  • Heineken
  • L'Oréal
  • Unilever
  • Itaú
02 · The problem

Online, brands know everything. At the shelf, almost nothing.

Advertise online and you know who saw it, who clicked and for how long. Advertise to drive in-store sales and you know what you spent, not what happened next.

Digital · event logLive
Anonymous ID · 7f3a…c21
07:10:04Impression in a mobile gamein-app
07:10:128.3 s of attention on the adattention
15:02:17Searched “ketchup”e-commerce
15:03:02Read 4 reviewse-commerce
15:04:48Added to carte-commerce
15:05:10Left without buyingsession
Events logged0
Physical retail · same personIn-store
Same day, after 8 p.m.
20:24Walked past the storeno record
20:27Walked into the supermarketno record
20:31Recalled the ad at the shelfno record
20:36Receipt includes the advertised SKUno record
Media effect · exposed vs. control?
Closing that gap is what HYPR does.
03 · The thesis

Intelligence has become a commodity.

GPT, Claude, Gemini. Your competitors license the same models you do.

We built the part that doesn't ship with the model: the grid of the real world.

Where people go, what they spend, what they pay attention to and what they think.

The better AI gets, the more valuable the grid becomes.

Without it, the model infers. With it, the model knows.

The grid answers what no AI can: whether someone bought your product in store after seeing your ad.

01 / 04HYPR manifesto
04 · Connecting the dots

A coordinate alone says nothing about what's actually there.

A directory tells you there's a bakery at that address. A good commercial broker knows whether people pass by, who they are, what they buy and whether the location is on the rise. We built that broker for millions of places, continuously updated.

−23.5725° · −46.6959°SHOPPING MALL
What's around itPlaces Graph
How people get aroundGeocodify · isochrone
The area's commercial profileIBGE · 2022 Census
What sells, and how muchGroundflow · sell-out
05 · The platform

Four IQs. One platform.

Individually, each one solves a known problem. Together, they connect place, spending, perception and attention in a single layer of intelligence.

The location layer: what surrounds every point, how people move between places and the commercial profile of each neighborhood.
Location intelligence · powered by Places Graph
Inside
O2OOOH AmplifierGeocodifyOverlapsPlaces DiscoveryDownloaded Apps
Places Graph
Supermarket
República · São Paulo
1,420places in radius
12categories
500 mradius
Hover over a placeReal places · Overture Maps · illustrative affinities
The transaction layer. Built on electronic tax receipts, the most objective record of a sale, it ties media spend to what later rang up at the register.
Transaction intelligence · powered by Groundflow
Inside
RMNDSignalsPatternsPlanSplitLiftEquityExplorerDemandshiftReal EstateCashify
HYPRLocation data23.5410° S
46.6415° W
+10.2%Lift on the advertised SKU
+4.9%Category-wide
Soft drinks case · 822 stores · 71,602 receipts
GroundflowPurchase data
SUPERMARKET · STORE 0214
CENTRO · SÃO PAULO/SP
CNPJ 00.000.000/0214-00
ELECTRONIC CONSUMER INVOICE (NFC-e) · AUXILIARY DOCUMENT
DESCRIPTIONQTYBRL
001 FRENCH BREAD KG0.4127.40
002 WHOLE MILK 1L210.98
003 COLA SODA 2L221.98
▲ Advertised SKU
004 GUARANA SODA 2L18.99
● Same category
005 BANANAS KG1.1208.95
006 LIQUID SOAP 3L132.90
TOTAL ITEMS8
AMOUNT DUE BRL91.20
DEBIT CARD91.20
CONSUMER NOT IDENTIFIED
NFC-e NO. 071602 · SERIES 001 · 09/25/2026 8:32 PM
3526 0900 0000 0021 4650 6500 1000 0716 0210 0071 6021
Look up the access key at nfce.fazenda.sp.gov.br
askIQ turns opinion into data. It measures perception, intent, preference and other declared signals to understand the why behind the results.
Research intelligence · natural-language insights
Inside
Brand LiftSurveyGeo Lift SurveyCustom ResearchBrand Query
9:41
news.
LatestBusinessSportsCulture
BusinessIn-store spending rises for the third month in a row12 min ago · 3 min read
Advertisement
Quick survey1 of 2
Which of these ketchup brands have you seen in ads over the last few days?
CitiesLong weekend expected to boost street retail
Brand Lift
Rubra · exposed vs. control
ControlExposed
Ad recall+19 pp
22%
41%
Brand awareness+13 pp
48%
61%
Consideration+9 pp
31%
40%
Purchase intent+6 pp
24%
30%
n = 2,418 responses95% confidence
Illustrative example · fictional brands
An ad served isn't an ad seen. High-impact formats validated by measured attention, so you decide not just where to show up, but how.
Format intelligence · powered by Max Attention
Inside
Tap to Game · betaTap to CarouselTap to MapTap to RevealTap to ChooseClick to CalendarFree Form
Max Attention
Advertisement
Mercado Livre · 320×480
Tap the ad to interact
06 · The edge

Four IQs, operating as one single system.

Isolated data produces isolated answers.

At HYPR, the four IQs run on the same data and build a single view.

  1. Where
  2. How
  3. How much
  4. Why
  1. 01Location determines where media runs.
  2. 02Media is measured at the register.
  3. 03Sales reveal what to ask next.
  4. 04Research recalibrates location.
07 · Activation

From pocket to shelf.

One day, one person, six moments. The choice made in the aisle at 8:30 p.m. has been building since morning.

Person on the couch playing a mobile game, with a ketchup ad on screen
07:10Discovery
First screen of the day
Still on the couch, between levels of a game. The ketchup appears for the first time.
High-impact format with measured attention.
Bus shelter with a ketchup poster
08:05Recall
Bus shelter
On the way to work, the same message at street scale.
OOH Amplifier: people who pass the shelter are retargeted on mobile.
Hand holding a phone showing ketchup reviews in an online store
15:00Consideration
Researching online
Comparing prices, reading reviews. The brand is no longer a stranger.
Measures brand recall: exposed vs. control.
Person with headphones listening to music on a phone, with a ketchup audio ad
18:00Intent
Music after a workout
Leaving the gym, headphones on, a few blocks from the store. Audio gets the message through without competing for visual attention.
Audio targeted to people near a point of sale.
Ketchup ad on a connected TV in a living room, during a series
19:00Preference
Streaming on CTV
Back home, an episode before dinner. The brand is back, this time on the living-room screen.
The same audience on the big screen, one step closer to purchase.
Man in front of the sauce aisle, with ketchup bottles in focus
20:30Decision
At the shelf
After the episode, a quick trip to the store. At the shelf, they reach for the brand that's been with them all day.
The receipt confirms whether the sale happened.
We deliver acrossTwitchSpotifyGloboplayDisney+Prime VideoHBO MaxActivision
08 · Measurement

Did the campaign drive in-store sales? And by how much?

We split matched stores into two groups: one is exposed to the campaign, the other isn't. Using receipt data, we compare sales of the advertised SKU and its category across both. The difference is the incremental effect of media, on the same principle as a clinical trial.

300M+
In-store transactions
~2M
Points of sale mapped
Unicamp
Statistical significance model

Delivery audited by independent third parties.

  • DoubleVerifyIASBrand safety and inventory quality
  • AdelaideAttention metrics
  1. 01Map
  2. 02Split
  3. 03Activate
  4. 04Measure

Matched points of sale, identified through receipt data.

Half form the exposed group. The other half, the control.

The campaign runs only near the exposed stores.

The difference in sales between the groups is the lift.

Exposed24Control24Matched stores
Campaign launch
Lift on the advertised SKU+10.2%+4.9% category-wide
Soft drinks case · 822 stores
AVG. SALES PER STORE · ILLUSTRATIVE CURVETIME →
09 · Case studies

The grid in action.

Each case covers the challenge, what we activated and what changed. Reach isn't a result. Measured impact is.

BEVERAGES · SP + RS · GROUNDFLOW LIFT

When actual purchase data drives media.

Challenge
One of the leading soft drink brands wanted to gain share in a highly competitive category, led by a player with nearly half of all sales.
Activation
The first wave, across 822 stores in São Paulo, delivered a +10.2% SKU lift. In the second, we recalibrated the approach with actual purchase data and scaled to 3,651 stores across SP and RS.
+19.2%
Advertised SKU lift in SP, second wave
+8%
Category lift in SP
+7.7%
SKU lift in RS, a stronghold of regional brands

How we measuredGroundflow Lift · exposed vs. control · 3,651 stores · Apr–Jun 2025

soft drink · ice
+19.2%
Advertised SKU lift in SP, second wave
+8%
Category lift in SP
+7.7%
SKU lift in RS, a stronghold of regional brands
BABY CARE · SP ·

What the survey indicated, actual purchases confirmed.

Challenge
Three months of media for a heritage baby care brand in São Paulo. The question was whether the campaign would shift perception alone, or purchases too.
Activation
We measured both dimensions. A survey compared 1,364 exposed respondents with a control group of 1,246, while purchase data tracked the brand's share across more than 500 retail banners.
+21.3%
Lift in consideration: 57% exposed vs. 47% control
+3.1pp
Share gain in SP for the quarter, from 31.2% to 34.3%
−10pp
“None of the above” answers: fewer undecided shoppers

How we measuredBrand lift · exposed vs. control · receipt-based shampoo share · SP · Apr–Jun vs. Jan–Mar 2026

care · baby
+21.3%
Lift in consideration: 57% exposed vs. 47% control
+3.1pp
Share gain in SP for the quarter, from 31.2% to 34.3%
−10pp
“None of the above” answers: fewer undecided shoppers
FOOD SERVICE · WORLD CUP 2026 · · CLICK TO CALENDAR

The ad became a date on the calendar.

Challenge
During the 2026 World Cup, a global fast-food chain launched sandwiches inspired by the national teams, each available on a specific day. Brand recall wasn't enough: people had to remember the day.
Activation
The audience was built from real-world behavior: frequent visitors to fast-food chains, bars and restaurants interested in soccer, the World Cup and food delivery. One tap on the ad saved the sandwich to their calendar, with a weekly reminder. Brazil's was the most saved.
36K+
Calendar reminders saved
94%
Acceptance rate among users who tapped

How we measuredO2O · Click to Calendar · confirmed calendar events · World Cup 2026

stadium · world cup
36K+
Calendar reminders saved
94%
Acceptance rate among users who tapped
FINANCIAL SERVICES · · OOH + O2O

Competitors owned the street. The brand owned the recall.

Challenge
A major bank needed to build brand recall on streets where competitors dominated the digital screens.
Activation
We mapped the competitors' DOOH screens and delivered the bank's message to the phones of people who passed them. The same recall survey was then run with the exposed group and a control group.
+11.7pp
Brand recall uplift vs. control
40.9%
Recall among exposed respondents
29.2%
Recall in the control group

How we measuredOOH + O2O · brand lift · exposed vs. control

financial district · city
+11.7pp
Brand recall uplift vs. control
40.9%
Recall among exposed respondents
29.2%
Recall in the control group
BEAUTY · HAIR · GROUNDFLOW LIFT

The category didn't move. The SKU did.

Challenge
Hair color follows a steady repurchase cycle year-round, with multi-brand beauty retailers outselling supermarkets. In 2025, the category was flat in São Paulo.
Activation
We ran media around 1,558 stores selected using receipt data and compared them with matched stores that had no campaign. With the category flat across both groups, the SKU's growth could only have come from share gains.
+6.3%
Lift on the advertised SKU
−0.4%
Category lift, same stores and period
28days
Average repurchase interval in the category

How we measuredGroundflow Lift · exposed vs. control · 1,558 stores · SP · Jul 2025

hair · color
+6.3%
Lift on the advertised SKU
−0.4%
Category lift, same stores and period
28days
Average repurchase interval in the category
BEAUTY · LAUNCH · CARBON NEUTRAL

A luxury launch with zero carbon debt.

Challenge
Digital media has a carbon footprint too: servers, networks and screens consume energy with every ad served. For a luxury fragrance launch, that impact was factored into the plan.
Activation
We calculated the greenhouse gas emissions of the campaign's entire digital media plan and offset the total with carbon credits from Amazon rainforest projects.
14.59t
Greenhouse gas emissions from the media plan
100%
Emissions offset

How we measuredDigital media plan emissions inventory · carbon credit offsets

fragrance · bottle
14.59t
Greenhouse gas emissions from the media plan
100%
Emissions offset
In the press2025–2026
Valor EconômicoJUL 2026HYPR and Unicamp build an AI platform on 250 million receiptsRead in Portuguese ↗VejaHYPR feeds R$ 2 trillion in purchases into AI to forecast spendingRead in Portuguese ↗ExameDEC 2025The startup showing brands where they are losing sales in retailRead in Portuguese ↗ExameMAR 2026Coordinates and consumption: Veloe lifts purchase intentRead in Portuguese ↗
10 · Applications

One grid, many applications.

The grid that guides a campaign also forecasts demand, prices a neighborhood by the square meter and backs an investment thesis. The question never changes: what actually happens in that place?

01 · Where HYPR started
Media

Campaigns run in the right catchment areas, in formats that hold attention, and are measured by what rings up at the register: exposed stores vs. control.

Where to activate
Which format to use
Whether it drove in-store sales
Whether the brand stayed top of mind
~2M
points of sale mapped
300M+
in-store transactions
+19.2%
SKU lift · soft drinks
02 · New vertical · Demandshift
Demand forecasting

Next quarter has already rung up at the register. With 1.11 billion purchase line items, from receipts to property deeds, we forecast demand by category, brand and region, and every forecast is blind-tested.

Which categories heat up or cool down441 categories
Which brands gain share of basketitem level
Where autos and housing are headedregistrations · deeds
92.7%
accuracy · consumer, 3 mo
95.4%
accuracy · autos, 4 mo
97.1%
accuracy · housing, 4 mo
03 · Property and site selection
Real estate

The price per square meter comes from the deed, not the listing. Layer property deeds, foot traffic, spending and demographic context on top, and you can price an asset, pick a chain's next location or decide where to launch a development.

What a square meter is worth, district by districtDeeds
Who passes by, and when
How much the neighborhood spends
Who lives nearbyIBGE · 2022 Census
R$ 5,565
per m² · apartments in SP
193K
deeds in 12 months
2.5M
deeds since 2006
04 · Land, property and consumer demand
Investment funds

A real-asset thesis needs observed prices, liquidity and demand, not market averages. We read what actually transacts to show where to invest, what to pay and what to expect.

Land: price per hectare and land-use suitabilityHYPR Agro
Risk: water, access and resilienceResilience Score
Property: price per m² and liquidity by regionDeeds
Demand: consumption by category and regionDemandshift
6,506
rural properties assessed
R$ 369B
in land listings
15
land-use categories
05 · Context for AI
AI models

A general-purpose model doesn't know what sold in a store this week. Connected to the grid, it does: context and training data on the physical economy, the part that isn't online.

Real-time queriesAPI
Training data licensingDataset
Works with any modelAnthropic · OpenAI · Google
1.11B
purchase line items
441
consumer categories
96
SP districts priced
11 · Why we're different

The model is interchangeable. The data isn't.

01
Observed transactions, not inferred ones
The industry typically relies on statistical models, correlation or self-reported surveys. We compare actual purchases between exposed and control groups.
02
Statistical rigor
With Unicamp's Department of Statistics, we developed the model that sets how many receipts a point of sale needs to be statistically significant.
03
Model-agnostic
The same data works with Anthropic, OpenAI or Google models without changing a line of code. We supply the context, not the reasoning.
04
Scale through infrastructure
Like a highway, the heavy cost is in building it. Each additional transaction costs less to process, so we can analyze more.

HYPR was founded in 2020 on a clear asymmetry: digital understood what happened online, while the physical world remained fragmented. We built the infrastructure to connect the two.

12 · Contact

Everyone claims to be AI-powered. Our AI is HYPR-powered.

From pocket to street, from point of sale to investment thesis.

Get in touchIt takes less than a minute.
Get in touch
It takes less than a minute.

By submitting, you agree to be contacted by HYPR as described in our Privacy Policy.

▪ © HYPR 2026The grid of the real world−23.4966 · −46.8480HYPR Ventures Tecnologia da Informação Ltda. · CNPJ 37.598.529/0001-69 · HYPR Ventures US LLC