Twin Lite · free, runs in your browser
Play a digital twin of your firm with ten numbers.
Twin Lite is a free simulator of a service firm. Set your team, the founder's review time, your mandates, fees and costs: it runs your next 52 weeks 300 times and returns your likely bottleneck, a revenue range (P10 to P90), lead time, cancelled mandates and your cash floor. Nothing you type leaves your browser.
Your likely bottleneck is founder review. In 8 runs out of 10, revenue lands between $2.1M and $2.4M. In the worst 10% of runs, cash never falls below $250K.
Cumulative billed revenue over the next 52 weeks
- Mandate in9 a month
- Research10 consultants, 3 wk a file
- Founder review7 h a weekbottleneck
- Client82% accept
- Billed$28K a file
- Cashpaid in 45 days
Which firm is this? The defaults describe a search and advisory firm of ten consultants. In another service firm, read mandate as engagement or project, consultant as anyone who produces billable work, founder review as the partner's sign-off and a billed outcome as your placement, closed deal or delivered report.
Generic model, not your twin. It is not calibrated on your data and it does not replay your past. A real twin is built from your event logs and must reproduce your last 12 months within ±10% before it advises anything.
Which three experiments teach the most?
Each one is a button under the chart. It starts from your firm today (the dashed line), changes the levers below and prints before → after at once. The figures here are for the default firm.
- Sell more. New mandates go from 9 to 14 a month and nothing else changes. Revenue does not rise, it falls by about 40%, the margin goes from 14% to -46% and, in the worst runs, the bank account goes below zero: more files queue behind the founder, clients wait past their patience and cancel. Selling into a bottleneck makes it worse.
- Automate the wrong step. The AI agent is switched on while the founder still reviews everything. The work gets done faster and billed files barely move (the margin goes from 14% to 13%), because production was never the constraint. Delegate review first (34%), then add the agent (40%): that is the "Delegate review + AI agent" button.
- Take yourself out. The founder steps out. Revenue drops to zero because nothing leaves review. Then tick "Delegate review": the firm keeps running even with fewer new mandates. The first year is what a buyer fears and discounts; the second is what you can show them. More on that in business succession.
How does Twin Lite compute?
Each run steps through 52 weeks. New mandates arrive at random around your monthly average. A mandate starts research only if a consultant slot is free, otherwise it waits and may go elsewhere. Consultants share their week across active files, and above four files each they lose 4% of efficiency per extra file, the cost of switching. A finished search queues for review; review capacity is the founder's hours plus any delegated hours. After review, 82% of files become a billed outcome three weeks later (in a search firm, a placement; elsewhere, an accepted deliverable), and you get paid after your payment delay. A file older than your patience threshold has a 12% chance a week of being cancelled.
Three hundred runs with different random draws give the band on the chart. The margin is revenue minus your monthly costs, before tax. It is the same logic as a full twin, with your firm replaced by ten numbers, so treat it as a way to think, not a forecast. If you have played Littlefield or another classroom game, the mechanics will feel familiar; we compare twelve of them in our business simulation game ranking.
What does a real twin add?
Your own flow, read from your CRM, projects and invoices instead of averages. Client types calibrated on your past proposals and played by agents, the part an AI twin of a company adds. Roles your firm really has. A replay certificate on your last 12 months. Scenarios Twin Lite cannot run: pricing, a new country, a merger, a crisis menu. And a log that checks each decision against reality at 3, 6 and 12 months.
Run another experiment Book a scoping call
A question about your own case? Email us.