Building Conservative Return Assumptions

The problem this solves

You present a business case with an impressive return, and the finance reviewer asks where each number came from. Two answers are “an industry benchmark” and “our other customers.” The reviewer discounts the whole model by half and the deal stalls. Strong claims did not hurt you; unsupported ones did. The remedy is conservative return assumptions the customer has validated themselves.

Make the customer the author of the inputs

A number the prospect supplied is far harder to challenge than one you supplied. For each input, ask the owner to state it, then write down who said it and when. Wherever possible use their internal data: ticket counts, hours logged, billing records. When they offer a range, take the cautious end and say why.

Apply haircuts openly

Even validated inputs carry uncertainty. Apply an explicit reduction, such as crediting only a portion of the expected improvement in year one, and show it as its own line. Hidden caution looks like padding; visible caution looks like discipline. Then ask the reviewer whether the haircut is too light or too heavy, which turns an attack into a discussion.

Choose the order of the benefits

• First, savings that appear directly in a budget line and can be audited.
• Second, capacity or productivity gains, credited only partially.
• Last, strategic benefits shown outside the total, as upside the customer may include.

Keeping upside outside the headline return means the headline can never be accused of inflation.

Run the case against its own weakest input

Find the single assumption that, if wrong by half, hurts most. Show what happens to the payback when it is cut in half. If the case still clears the customer’s hurdle, you have shown robustness without being asked.

What this means for the SE

Keep a log of every assumption, its source and its owner. When finance challenges a figure in the meeting, you can answer in one sentence and name the person who gave it.

Decision and check

Release the model only when every input has a named source, all haircuts are visible, and the case still meets the customer’s payback rule with the weakest input halved. If it fails, present the smaller, conservative return assumptions you can defend and offer a proof of value to earn the rest.

This is part of the 80% of Solution Engineering that stays constant whatever the product. The SE Ignite curriculum teaches these mechanics for free, so you can practice them before your next deal.