Crane Engine^™

Budgets set by evidence, not gut feel.

Most accounts are paced by habit: last month plus a bit, spread evenly, nudged when something looks off. The Crane Engine^™ replaces that with a model. It reads live performance across Google, Microsoft, Meta and LinkedIn, works out where each pound earns the most right now, and splits the budget so the whole account lands on plan. The same logic runs for every client, and a senior practitioner reads it, sanity-checks it and signs it off before anything changes.

It backs the next pound, not the average.

This is the difference that makes the whole thing work. A campaign returning thirty seven to one on average is not the same as one that will return thirty seven to one on the next pound you add to it. Once a campaign has captured most of the demand available to it, each extra pound buys less. The engine reads where each pound stops working and equalises the return at the margin across every campaign, which is the mathematically optimal way to split a fixed budget, known as the equimarginal principle. In plain terms: it stops pouring money where it has quietly stopped converting, and moves it to where it still does.

Diminishing-returns curves to find each campaign's ceiling, confidence-weighted shrinkage so thin data does not mislead, equimarginal allocation to split the whole budget, and damping so budgets ease rather than lurch. Established techniques, applied to your account and read by a human.

Eight signals behind every budget.

No single number decides a budget. The engine weighs these together, then a human reads the result against everything the data cannot see.

1. 01

   Marginal return

   It backs the next pound, not the historical average, so budget follows where money still converts rather than where it once did.

2. 02

   Demand ceilings

   It reads impression share to know when a channel is full. It will not overfund a campaign that cannot spend more without losing to rank rather than budget.

3. 03

   Recency, weighted

   The last seven days count most, steadied by the last fourteen and thirty, so the model reacts to a genuine recent change without letting one noisy week swing the plan. Because conversions take days to come in, the freshest days are treated as still maturing rather than read as a real dip.

4. 04

   Confidence

   Campaigns with real conversion history are trusted to move decisively. New or thin ones are held near the account benchmark and given room to prove themselves rather than judged on noise.

5. 05

   Seasonal demand

   It leans budget toward the days that historically convert, using last year's demand curve for your market, so spend meets school holidays and seasonal peaks instead of running flat through them.

6. 06

   True incrementality

   Platforms flatter their own numbers. The engine can discount reported returns to the real causal effect, so budget is never chased toward results a channel is taking credit for but did not cause.

7. 07

   Time and pace

   It counts the time of day and the days left in the month, and treats today as only the spend still ahead of it, so the account lands exactly on budget rather than sprinting early or coasting to the finish.

8. 08

   The whole account

   It judges budget across the portfolio, not campaign by campaign, protecting the coverage and shared learning that smart bidding needs to keep working.

Re-solved every week, on fresh data.

The engine runs on live platform data, refreshed through the day. Budgets are set at the start of each month and trued up every Monday, so the account is always pacing to the current picture, not last month's. Where a campaign uses a target based bid strategy, the model also keeps the target honest as budget scales, so efficiency holds instead of drifting.

Live data from Google, Microsoft, Meta, LinkedIn and GA4. Set monthly, trued every Monday, signed off by the person who does the work.

How the engine works.