People who begin recording what they spend usually spend less, without having set any limits. The act of measurement is doing the work on its own.
Most spending is not decided
A large share of purchases are habitual or prompted, made without any comparison against alternatives or against what remains available.
Recording a transaction inserts a moment of attention into that sequence, and the purchases that do not survive scrutiny are precisely those that were never really chosen.
The effect is strongest for small frequent purchases, which are individually unremarkable and collectively significant.
Totals are invisible without a record
Nobody accurately estimates their own category totals, and the error is consistently in one direction. Frequent small amounts are underestimated far more than occasional large ones.
A running total makes the aggregate visible while the month is still in progress, which is when behaviour can still change.
A statement reviewed afterwards reports a result that can no longer be influenced, which explains why reviewing is far less effective than tracking.
Categories determine what is learned
Too few categories hide the useful detail, and a single large bucket for general spending reveals nothing actionable about what is inside it.
Too many produce classification decisions for every transaction, and the effort causes the practice to be abandoned within weeks.
The productive level is usually the point where each category corresponds to a decision that could actually be made differently.
Automation trades accuracy for durability
Applications that import transactions remove the manual effort, which makes tracking sustainable over long periods rather than for a few enthusiastic weeks.
They also remove the moment of attention that produced much of the original effect, since a transaction categorised automatically was never consciously recorded.
Reviewing imported transactions weekly restores part of that attention while keeping the effort low, which is generally a better balance than either extreme.
The effect fades and can be renewed
Tracking loses potency as it becomes routine, because attention that has become automatic no longer interrupts anything.
Periodic intensive tracking, applied for a month at a time when circumstances change, tends to produce more insight than continuous low-attention recording.
Changes that outlast the tracking period are those converted into structure, such as a cancelled commitment or an automated transfer, rather than those relying on continued attention.