Synthetic personas and synthetic simulations: test ideas on simulated visitors first
Synthetic personas are simulated visitors built from how real people behave. Synthetic simulations put those personas through your page, offer or experiment to see how each might respond. Used well, they tell you which ideas deserve real traffic. Used badly, they are a chatbot with a job title.
Every testing program runs into the same limit: there is only so much traffic. Synthetic simulations let you explore a wider set of ideas before any of it is spent, then send the strongest one into a proper A/B test.
What is a synthetic persona?
A synthetic persona is a model of a visitor, not a description of one. A marketing persona says "Sarah, 34, operations manager, values efficiency." A synthetic persona says how someone like Sarah behaves on a page: how patient she is, how much she reads, and how ready she is to act when she arrives.
That difference matters. A description can only be imagined. A behavioural model can be run. Put the same page in front of it twice and it makes the same kind of decisions, because its behaviour comes from its traits rather than from a writer's guess.
Patience
How long the visitor stays before giving up, and how much friction they tolerate on the way.
Reading depth
Whether they scan headlines and buttons, or read the proof, the pricing and the small print.
Intent
How ready they are to act, and what they arrived expecting, such as the promise in the ad they clicked.
What is a synthetic simulation?
A synthetic simulation is what happens when you run those personas through something real. You pick a page, an existing experiment, a single variant or a mockup. You set a goal, such as a trial start or a demo request. Then each persona visits every candidate and tries to get there.
The output is a trace for each visit: which version the persona preferred, why, where it engaged and where it dropped off. Across a panel of personas, patterns show up quickly. A layout that works for a skimmer may lose a skeptical researcher. A headline that suits returning visitors may confuse someone arriving cold from an ad.
Those disagreements are the most useful result. They tell you exactly what a real test needs to settle.
Synthetic personas vs. asking a chatbot
The quickest way to fake this is to paste a page into a chatbot and ask it to review the page as a busy CMO. You will get a confident paragraph. You will not get a visit.
- A prompt has no behaviour. Swap the job title and the answer changes tone, not decisions. A synthetic persona's choices come from traits that stay fixed across every page it sees.
- A prompt sees the whole page at once. A real visitor, and a good simulation, has to get through the layout. A button two scrolls down is invisible to someone who never scrolls.
- A prompt has no goal. A simulation is always trying to reach something, so it can tell you where the path breaks.
- A prompt can't be compared. Running the same panel against every variant is what turns opinions into a ranking.
Where synthetic simulations help
Synthetic simulations are cheapest when ideas are still cheap to change. The best places to use them:
Shortlisting
Narrow five concepts down to the two worth a live test slot, instead of queuing all five for months.
Catching obvious problems
Find the buried button or the ad-to-page mismatch before real visitors do, so live tests aren't spent discovering them.
Low-traffic sites
When a site can only support a few experiments a quarter, simulations make sure each one is testing a serious contender.
Audience splits
See whether mobile skimmers, researchers and ad clickers want different things, and design the real test around that question.
What synthetic personas can't do
A synthetic simulation is a signal, not a measurement. It cannot give you a conversion rate, a revenue lift or statistical significance. It does not know your brand's history with a customer, a competitor's promotion this week, or the thing your audience cares about that nobody wrote down.
That is why the simulation is the step before the experiment, not a replacement for it. Real visitors still decide what ships. Anyone selling synthetic personas as a way to skip testing is selling a guess with better formatting.
How RunPivot uses synthetic simulations
RunPivot Simulations runs synthetic personas and live A/B tests in the same workspace, so a promising idea doesn't have to be rebriefed into another tool.
- Pick what to explore: a live page, an experiment, a variant, any URL or an unbuilt mockup.
- Set the goal: the action a visit should end in.
- Choose the personas: match the panel to your real traffic, such as ad clickers, mobile skimmers, skeptical researchers and returning visitors.
- Read the traces: preferences, reasons and drop-off points for every persona on every candidate.
- Test the winner: send the strongest idea into a RunPivot A/B test and measure what real visitors do.
For the launch story and the thinking behind it, read Introducing Simulations.
Try it on a page you already have
The fastest way to judge synthetic simulations is to run one on a page where you already know the answer. Pick a past test, simulate both versions, and see whether the personas point the same way your visitors did.