Models close on paper almost every time. The problem is that not everything gets put on the paper.
Eight recurring errors
- Only the subscription in costs. Accounts, proxies and model calls together usually exceed it.
- Your time counted at zero. Twenty hours a month is a salary even when you pay it to yourself.
- Conclusions from the best day. Count a full month, including weekends, holidays and downtime from restrictions.
- Warm-up ignored. Month one delivers less volume at full cost, by design.
- Attrition not budgeted. Accounts run out; replacement is a recurring line, not an accident.
- Order value used instead of revenue per lead. If one in five pays, divide by five.
- Linear extrapolation. Ten accounts will not produce ten times as much if the niche holds venues for two.
- Best-case inputs everywhere. Top conversion, maximum allowance, highest order value — each plausible alone, together a scenario that never happens.
Checking your own model
Take last month's actuals and compare them with what the model predicted a month ago. The gap tells you which of the above applies to you. Duller than building a new model, and more useful.
The margin rule
A model that closes exactly does not close: any one of the above will push it under. A one-and-a-half times buffer on costs is not pessimism, it is the normal correction for reality being messier than a spreadsheet. Run three scenarios — conservative, expected, good — decide on the conservative one and plan on the expected. The correct order of calculation is in the break-even piece.