You can put a number on it. Most organisations never have, which is why it is the first thing cut.

Every other function in an organisation can answer the question "what are you worth?" Sales points at revenue. Procurement points at savings. Even functions with indirect value — legal, IT — have conventional arguments about risk avoided and productivity enabled.

The impact function usually cannot answer it. It can describe what it produces: reports, evaluations, indicators, learning. It generally cannot say what those are worth in money.

This is why it is the first thing cut in a difficult year. Not because leadership does not care about impact, but because a function that cannot state its value is, in a budget conversation, indistinguishable from one that has none.

The objection, and why it does not hold

The standard response is that impact value cannot be monetised without distorting it — that reducing social outcomes to currency is precisely the thinking impact work exists to resist.

That objection is about valuing the impact. This is a different question. We are not asking what the outcomes are worth. We are asking what the function is worth: what changes financially because the organisation has an impact function that works, versus one that does not.

That is an ordinary business question and it has ordinary business answers.

Where the return comes from

In the engagements we have modelled, value concentrates in a handful of places:

  • Decisions made earlier. The largest single source, and the least discussed. A programme corrected at month six rather than month eighteen saves twelve months of spend on an approach that was not working. This requires evidence to arrive before decisions, not after — which is an operating property, not a measurement one.
  • Work not repeated. When learning does not travel between teams, insight is produced repeatedly and compounds nowhere. The cost is real and almost never counted.
  • Bids won and retained. Credible, well-evidenced impact is increasingly a scored criterion. This one is directly attributable and organisations usually already track it, in a different department.
  • Reporting effort avoided. Organisations with no operating layer produce bespoke reporting for each funder, repeatedly, by hand. A coherent function collapses much of that into one apparatus.
  • Risk not realised. The overstated claim withdrawn, the failing partnership ended earlier. Genuinely hard to attribute, and worth including only when it can be evidenced.

The honesty problem

Here is where most ROI analysis of this kind loses credibility: it blends what has actually happened with what is projected, presents a single confident figure, and invites a finance director to find the seam. They will find it.

The discipline that matters is separating the two and saying which is which. In our own work we label every figure as realised or modelled, and we do not add them together.

For one engagement, we modelled the return on the impact function at AUD $5.0 million to $15.2 million, a ratio of 8:1 to 25:1 against the upper bound of investment. That is a modelled figure, produced with the client on their own cost base and their own programme data. It is not a realised return and we do not present it as one.

A range labelled honestly survives scrutiny. A single number that blends realised and projected does not survive the first serious question, and takes the rest of the argument down with it.

Building the case

The method is not exotic. Establish the fully loaded cost of the function. Identify the decisions it has plausibly changed, or would change if the evidence arrived in time. Attach the financial consequence of each with the client's own numbers. Separate realised from modelled. Present a range with the assumptions visible.

The hardest step is the second, and it is diagnostic in itself: if you cannot identify decisions the function has changed, you have not found a valuation problem. You have found the operating problem, and the valuation was never going to be flattering.

Which is the useful thing about running this exercise. Either you end up with a defensible number for a function that has been undervalued, or you learn precisely why there is no number to find.