The cash flow forecast

Budgets look at months. The forecast looks at days, which is where balances actually run out.

What feeds the projection

The forecast projects your balance forward from three ingredients, and a mode picker decides the mix. The first is what you have told it exactly: scheduled transactions, with their amounts and dates. The second is your monthly budget, for the categories you have budgeted. The third is your history: the regular payments we found, on their dates, and your usual spending in every category without a budget that month. The Combined (Smart) mode uses all three; Scheduled Only and Budget Only isolate the first two. The result is a day-by-day line, with hover tooltips showing the projected balance, inflows, outflows and net flow for each day.

How your history fills the gaps

Regular payments on their dates. Salary, rent, direct debits and subscriptions we found in your current accounts, but that you have not scheduled, are placed on the day they usually land. We learn each one's day of the month and what happens when it falls on a weekend or bank holiday: a salary paid the Friday before, a direct debit collected the Monday after.
Usual spending for the rest. For each category without a budget that month, its typical month over your last six full months (or its average over up to a year, for something you pay only now and then), spread evenly over the month. Spending on cards, subscriptions included, is paid from your current accounts at the end of each month.
Nothing counted twice. A budget sets the total for its category, with the scheduled and regular payments in it counted inside it, not on top; a payment you have scheduled is never counted again from history; and payments that have stopped are left out.
How much history it needs. Usual spending needs two full calendar months in the accounts you forecast; a regular payment shows up once it has been paid three times. When the forecast rests mostly on under six months of history, it says it is a rough guide.
What it tells you. On the web, a note above the chart says when the forecast is built mostly from history, Built from shows the mix, What this forecast assumes lists every regular payment and monthly figure used, and Use my history switches the history part off. The mobile app shows the same projection without these notes for now.

Reading it well

The dip matters more than the endpoint. A month that ends fine can still spend three days below zero in the middle. The forecast exists to show the valley between salary and rent, while there is still time to move something.
Bands admit uncertainty. The best-case and worst-case bands, toggleable in the modes that use your budget, widen as the projection reaches further out, because pretending week six is as knowable as week one would be a lie drawn as a line.
Scope it to the question. The account selector focuses the projection: your everyday accounts for the will-rent-clear question, everything liquid for the bigger picture. Tracking accounts stay out, as everywhere. On the web you can choose only active accounts, and your choice is remembered: an account in it that you later close, archive or delete drops out, and if none is left the forecast starts again from all your accounts.

A forecast is a consequence, not a promise

The projection is downstream of your schedules, your budget and your history; it improves exactly as fast as they do. History spreads everyday spending evenly across the month, so the dip before payday is only as right as payday's date: scheduling your salary and bills pins them to their real days, and a budget replaces an estimate with your plan. Transactions still waiting in the review queue are not in your history yet, so recent months read lower than they were until you review them.

What next

The levers that sharpen the forecast are all a page away: scheduled transactions for the knowable, a budget for the planned, and a worked-through review queue for the history it learns from.