A system that holds your revenue plan — and steers you back to it
It learns on your company's own data. It plans and forecasts revenue, costs and operating profit, catches deviations before they become fact, prepares corrective actions and measures what they achieved.
The system reads your data and changes nothing at the source. Nothing has to be moved, reconfigured or switched off: whatever your company runs on today keeps running.
Six steps that close the circle and start again
This is not a sequence of reports but a turn of a wheel: every step leaves a trace in the system, and the next turn starts with the result of the previous one already factored in. Click a step to see how it looks inside the product.
After the sixth step the loop returns to the first: the plan is recalculated using what actually worked
A plan that doesn't freeze in December
The system builds a plan for revenue, costs and operating profit across the horizon and recalculates it whenever actuals move. The plan lives next to the forecast, not as a separate file on the CFO's desktop.
- A plan for every unit and every period, not just for the company as a whole
- Costs and revenue in one model instead of two separate reports
- Recalculated as actuals change, with no manual rebuild
Profit is lagging further behind than revenue — costs are outgrowing the plan
A range, not a single number for the board pack
The model gives a pessimistic, base and optimistic scenario for every unit and period. And it owns its forecasts: accuracy is checked after the fact rather than quietly forgotten.
- A P10–P90 interval instead of one point on a chart
- Manual drivers for what the model cannot know by itself
- Backtesting: how well previous forecasts actually held
Actual · P50 forecast · upside to P90. Past accuracy: 91% of outcomes landed inside the interval
Deviations show up in money and in time
Actuals are continuously checked against the norm and the plan. The queue is sorted not by percentage but by the amount at risk and by how much time is left to act.
- The norm is calculated per unit, not from a company-wide average
- Money at risk in roubles, not a share in percent
- The deadline after which intervening is pointless
23 open deviations worth 47.3 m ₽ in total
Not an alert — a finished decision with a price tag
For every deviation the agent prepares options with an expected impact and the limits they may be applied within. You can see why this particular action was proposed and which signals the conclusion rests on.
- The cause and the signals it is built on
- Expected impact in roubles, not "improved performance"
- Guard rails: the price floor and the budget ceiling
Alternative: a 7% price cut yields +9.1 m ₽ of revenue and −3.4 m ₽ of margin
The decision reaches a person and a system
An approved action becomes a task with an owner and a deadline, or it executes automatically — if autopilot is switched on for that class of decision and the limits are set.
- A task with a named owner and a deadline
- Autopilot within the limits you set yourself
- A log of who launched what, and when
Every action is recorded: what was proposed, who approved it, when it went live
Actual result against what was promised
Once an action is executed the system compares the outcome with what it promised and records the gap. That gap is what it learns from — the next estimate will be sharper.
- Expected versus actual, reconciled for every decision
- For marketing — incremental lift measured against a holdout, not budget spent
- A decision history builds from day one and is there when the season closes
The weekend fare didn't work — the model has adjusted its estimate for similar cases
Not just sales. Three axes that meet in profit
Managing revenue in isolation from costs is pointless: a discount lifts occupancy and eats margin, advertising brings bookings and burns budget. The system treats all three axes as one problem.
Revenue and sales
Pick-up pace, occupancy, price, channels and segments — for every unit and period, against its own norm rather than a company average.
- A forecast as a range, not a point
- The norm is per route, per date, per vessel
- Deviation shown in money, not in percent
Costs
Operating and marketing costs sitting next to the revenue they produced. Standards, actuals and variance in one model instead of separate reports.
- Unit cost and break-even point
- Marketing budget as a controllable lever
- Cost of the sales channel, not just the rate
Operating profit
The headline metric of the loop. Every proposed action is judged by its contribution to profit, not by lifting one number at another's expense.
- Each decision's contribution is counted in roubles
- Scenarios compared before acting
- Actual impact reconciled with the expected one
Four layers of knowledge about your business
The full history is loaded in; sources connect as they become ready. The longer the system runs, the sharper its norm and the fewer manual corrections it needs.
Your own data
Sales and bookings, unit registries, tariffs, cost standards, operations and maintenance. All the history available, not just the latest period.
Marketing and attribution
Budgets, channels, campaigns and the path from impression to payment. The system knows which effort turned into what, and how long it took.
Market and competitors
Competitor prices and occupancy for the same dates and routes, exchange quotes, freight rates, cargo flows.
External environment
Event calendars, weather, seasonality, exchange rates, flight connections and regulation — everything that moves demand from outside.
Division of labour inside the loop
Not one general-purpose assistant but a set of agents, each with its own specialism and its own remit. Each keeps a log: what it saw, what it proposed, what it launched and how it ended.
Planning
Builds the revenue, cost and profit plan across the horizon and keeps it alive rather than frozen in December.
Forecast
A P10–P90 range for every unit and period, with backtesting of its own forecast accuracy.
Deviation control
Continuously checks actuals against norm and plan, and finds where and by how much they diverge.
Diagnosis
Explains the cause: price, demand, channel, competitor, cost or an external factor — with the signals and a confidence level.
Corrective actions
Prepares options with an expected impact in roubles and the constraints they may be applied within.
Pricing
Calculates the price lever per unit and date, with guard rails and contractual constraints.
Marketing
Allocates budget against a specific task, picks channels and audiences, and measures lift with holdout groups.
Execution and impact
Launches the approved decision, tracks delivery and reconciles the actual result with the expected one.
You decide how far the system acts on its own
Autonomy is switched on gradually and area by area. You can start with observation on a single route and keep autopilot for decisions where the cost of a mistake is clearly small.
Observation
The system calculates, forecasts and surfaces deviations. Every decision is made by a human.
Recommendation
Proposes an action with its rationale, expected impact and limits of application.
Approval
Prepares the whole decision — price, campaign, task. All that's left for a human is to press launch.
Autopilot within limits
Acts by itself inside the limits you set and reports after the fact. You define the boundaries.
Level 1 is on by default. Moving up is a separate decision you make for each class of task.
A shortfall is cured with demand, not only with a discount
Conventional revenue management has one instrument: price. So every problem ends in a discount — it delivers volume, eats margin and teaches the market to wait for the sale.
- A unit that is behind gets targeted promotion with spend capped at its own shortfall
- Response forecast: how many bookings a rouble buys on this route at this distance from cut-off
- Holdout groups: impact is measured as lift, not as budget spent
- Over time the agent plans campaigns itself, as part of the corrective actions
Incremental return per rouble of budget, measured against a holdout. Brand search harvests demand that would have arrived anyway — only this kind of measurement reveals it.
Your business specifics are configured, not forced to fit
One core engine, your own domain model. Units, inventory, cut-off dates, pricing rules, cost structure and constraints are all described for your company.
Sources
Booking, cargo booking, PMS, dispatch, accounting systems and ad platforms are connected. Read access only — no writes, no migration.
History
All available history of sales, costs and events is loaded. It is what the norm is built on and what the models learn from.
Business model
Your units, inventory, seasons, channels, contractual constraints and cost structure are described.
Rules and limits
Price and budget caps, agent autonomy levels, roles and areas of responsibility are set.
One mechanism, different insights
In all of these businesses inventory is finite and vanishes at a hard date. But the unit of control, the decision horizon and the main lever differ — pick yours and we'll speak your language.
Decisions take hours, not weeks
Weather, city events and time of day move demand faster than a report can be assembled.
See it for operatorsYou sell vessel time, not slots
Income is counted per vessel-day. Ballast legs, waiting and the choice between enquiries cost more than the rate itself.
See it for fleetsThe block is paid for before sales start
Allotment depth is chosen months before demand becomes clear, while exchange rates and flight schedules change everything in days.
See it for tour operatorsA cabin exists only until departure day
A long sales cycle, agents across markets and currencies, and a pricing mistake that only surfaces months later.
See it for cruise linesRoom inventory resets every night
Rate and occupancy pull in opposite directions, and the same tariff yields different profit through different channels.
See it for hotelsIf inventory is finite and there's a cut-off
Rentals, events, clinics, parking, warehousing — the mechanism is the same. Tell us about your business and we'll say honestly whether it fits.
Let's talk