Crowdsim
Simulated audiences for practising media decisions

A world that reacts.

Your campaign tool decides what happened. Crowdsim says what people would say about it: the comments under the ad, why most people scrolled past, what the client thinks. Informed by this week's public conversation, with the sources to check, and labelled simulated every time.

Request early access See a reaction

An API, not a dashboard: it plugs into the tool you already use. Practice, not prediction.

A crowd seen from above, each person a dot Every dot is a person. Every person has an opinion.

A rushed first ad, and what came back.

A student ran one month of a default campaign: everyone in Canada, a stock product photo, "premium dog food for happy, healthy dogs". The campaign simulator produced the inputs. Crowdsim produced everything below them.

The decision · month 1
Q&P Quill & PawSponsored

Premium dog food for happy, healthy dogs. Shop now!

Dry dog food being poured from the bag into a full bowl
quillandpaw.caQuill & Paw Shop now
Spend$608
Clicks540
Orders1.9

These numbers come from the caller's own engine. Crowdsim never changes a number.

What people said Simulated Personas had read the pulse of 10 Oct 2026.
Notices when an ad looks AI-madeBag and a perfect bowl with no actual dog in it… is this AI? "Premium dog food for happy, healthy dogs" could be literally any brand.
Two large dogs, buys the big bag on sale"Premium" usually means pricey. How much and what size bags? Not clicking through just to find a price.
Checks where things are made since the tariffsIs this made in Canada or just sold here? Nothing in the ad says.
Why the 67,015 who didn't click scrolled past
  • "Happy, healthy dogs" is what every dog food says, so nothing here told me why this one's different.
  • No price, no deal, no idea if it's even Canadian, so I kept scrolling.
  • I have cats, and it's just a bag and a bowl anyway.
From the founder, end of month 1The headline was just our name, over a bag and a bowl that could be anyone's. We didn't say the fat is 9%, or that we print the real percentages, or that it's milled outside Paris, Ontario. Next month I'd like the ad to say one true number and show one real dog.

The world learns from the world, and reacts to you.

Three parts, kept apart on purpose.

Performance graphs on a laptop screen

Your engine decides the numbers

Orders, rank, cost per order, frequency: your simulator or your real results. Crowdsim takes them as given and never invents or moves one, so marking stays fair and a plan scores the same twice.

A row of newspapers on a shelf

The world reads the pulse

Every week a research job writes what the public conversation in your industry looks like: dated, weighted, sourced. A recall or a tariff lands in the personas' heads before it lands in your campaign.

A person filming a crowd on their phone

The crowd reacts

Personas written as situations, never demographics, react to your actual wording and choices: comments, scroll-past reasons, what each searcher did, a note from the client. Labelled simulated, every time.

The rule

Nothing a user sends is ever training material. Worlds learn only from the pulse, from what customers say in public, and from real results a caller chooses to report. What students and trainees type gets a reaction, never a lesson.

What you integrate

An API, not a dashboard.

There is nothing to log into. Your simulator, your training tool or your own app sends one request with the decision and the numbers, and shows the reactions however it likes. Keys, usage, worlds and the pulse are endpoints too.

POST https://api.crowdsim.ca/v1/react
Authorization: Bearer ae_…

{
  "world":    "quill-and-paw",
  "channel":  "social",
  "decision": { "headline": "Know exactly what's in her bowl", … },
  "outcome":  { "orders": 14, "clicks": 561, "frequency": 1.9, … }
}

→ { "simulated": true, "reactions": { "comments": [ … ], … } }
The Pulse

What the personas read this week.

Three of nine signals from the pet food world, 10 October 2026. Each dated to its own evidence, weighted for how much it should shape a reaction, and sourced.

Price · weight 5CA · to Aug 2026

Owners are cautious about price but still buying premium food

Pet Valu's same-store sales were flat; customers wait for promotions but keep spending on premium food. "How much?" comes early.

Origin · weight 4CA · to Sep 2026

Buy Canadian is back since August, now argued on price

Counter-tariffs took effect 8 September; grocers brought back origin labels. "Is this Canadian?" is a live question.

Store voices · weight 4CA · to Aug 2026

The questions owners ask before they buy

Puppy or adult? Gluten? Live probiotics? Asked how high the protein is, the rival answered "a guaranteed minimum". Personas ask the same of any ad that doesn't answer.

Why a simulator

A flight simulator for media planners and buyers.

Pilots don't learn crosswinds on a plane full of passengers. The simulator's weather is made up but probable, and an instructor can freeze the storm so every trainee flies the same one, or feed it live weather and let the session go wherever it goes. Everything in a Crowdsim persona's life works the same way: whether they notice an ad, whether a recall puts them on alert this week, whether the dog has a bad month. The only question is whether the dice are fixed.

Snapshot · for classrooms

Every student flies the same storm.

The instructor pins the week, the seed and the personas' lives. The same plan gives the same numbers for everyone, a marker can replay any of it, and the conditions can be set by hand: a recall in the news, a persona starting in a bad month.

Live · for agencies and brands

Anything can happen.

Nothing is pinned. The weekly pulse moves the levers underneath the crowd, lives roll on, and the same plan run a week apart gets different results because the world changed. Practice on a messy public, with every number still traceable to a published rule.

One engine, any industry.

A world is a brief, a set of personas and a watch list. Bring your own client, or start from ours.

Students working on laptops in a classroom

Professors

Attach it to the campaign simulators you already run. Students hear from customers instead of a rubric, and the pulse becomes the social-listening exercise before they write a word. Student work never leaves their browser.

Two people sketching a plan on a whiteboard

Agencies

Juniors make their mistakes here, on a public that talks back, before they make them on client money. Run it live, or freeze a week into a case study for the next cohort.

A woman checking her phone beside her poodle

Brands

Run a headline pair or a creative concept past a crowd that has read this week's news about your category, and see which questions your page doesn't answer.

Early access, Winter 2027.

Crowdsim runs in a Canadian college classroom today and is opening its next worlds in January: new industries, new clients, and the first cohorts outside the course it was built in. If you teach, train or plan media and want a world of your own, leave an address and we'll write when yours is ready.

One email when your world is ready. Nothing else.