Caspase Demo · Underwater

A demonstration on sample data

Your good clients are paying for your bad ones.

A fixed monthly price meets a variable amount of work, and the difference is whatever the client turns out to be like. Some of them cost more than they pay. The monthly numbers never show it, because the healthy ones cover it and only the total gets reported.

Run it on

Sample data, invented for this page Time entries CSV Agreements CSV

Name the worst one.

22%

The whole book, on one axis.

Every client by what the agreement actually earns after the work is costed. The dashed line is the margin the agreements were priced at. Anything left of the solid line is being paid for by somebody else on this list.

Client by client.

ClientSizeCharged /moPrice each Cost eachHours eachMargin A yearWhat to do

Run it on your own book.

Two things. A time export with a client name and hours on each row, and the monthly value of each agreement. Both already exist; they just live in different systems, which is the whole reason nobody has put them together.

Neither leaves this page. No upload, no server. The reading happens in your browser and disappears when you close the tab.

client, hours, and ideally a date, a role and a cost rate

no account, no signup

How it decides.

There is no model here. It is a division that nobody performs because the numerator and the denominator live in different systems.

  • 1
    Cost the work at what the work costs. Every logged hour carries the loaded cost of whoever delivered it. An hour of engineer time and an hour of helpdesk time are not the same money, and averaging them hides exactly the clients this is looking for.
  • 2
    Add what you carry for them. Licences and tooling are real cost of delivery and they scale with the client, so they belong here rather than in overhead.
  • 3
    Divide by what the agreement charges. That is the gross margin of one client, which is a number almost no small firm has, because the time is in one system and the agreement is in another.
  • 4
    Add up two different things, and never confuse them. Gap to target is what the below-target agreements would have earned at the margin the book is priced at, less what they actually earned. A client at 30% against a 50% target contributes to it while making money. Lost outright is the narrower figure: only the clients whose delivered cost is more than what they pay. The headline is the first one, because it is the size of the pricing question. The second is in the strip beside it, because it is the size of the bleeding.
  • 5
    Give both levers, and say which one is available. For a client a little under, the answer is a price. For one far under, the price that would fix it is one nobody accepts, and the honest answer is the other lever: how many hours a month the agreement actually pays for, against how many are being delivered. A recommendation that cannot be carried out is not a recommendation.

Where this does not work.

The honest caveats, and the first one is the big one.

Time entries have to be honest

Where people log eight hours in round numbers on a Friday, the cost side is soft and so is everything built on it. This is the reason the same idea fails in agencies, and it is worth checking before trusting a number.

Gross, not net

This costs delivery only. Sales, admin and the owner's own time are not in it, so a client at the target line is not a client making money overall.

A bad quarter is not a bad client

One migration or one flood makes a good client look terrible for a year. Twelve months smooths some of that. It does not remove it.

Some are worth it anyway

A reference client, a foot in a sector, a favour owed. The number tells you what that costs you. Deciding it is worth paying is still yours.

This took a day.

It is a demonstration, not a product. The version that matters reads the time system and the agreements directly, runs on the morning of a renewal, and knows the difference between a client who had one bad month and a client who has been like this for two years.

If you want to know what this would say about your book, send me the question.

[email protected]

Caspase LLC · Virginia · caspase.ai