THE ADD-BACK · episode 02

The Baseline Is Perishable

Arc 1 · Days 1–100
In the room: Operating partner, incoming CFO, functional leads
The question: What do I have to measure before I'm allowed to cut?

The position

SightingA hundred days of political capital and a VCP with an automation workstream in it. The instinct is to spend the capital on the cut, because the cut is what shows up in the first board pack.

The consensus readMeasure later. The data will still be there.

The mechanismIt won't. The pre-automation baseline is the only period in which the process runs one way, and it stops existing the day you change it. Exception volume, cycle-time distribution and who actually resolves failures are observable now and unreconstructable afterwards. Nobody logs a corridor conversation retroactively.

The exposureAt exit you argue value creation from a model rather than a measurement, and a buyer discounts a model.

The testAsk your CFO for the pre-implementation exception baseline on last year's automation. If it doesn't exist, that's the whole episode.

The bridge — claimed to realised

Framed as provable value creation, because that is the currency at exit — a dollar of margin you cannot attribute is not worth a dollar in a bid. The operating case draws on the same composite as Episode 01; cited research is real and sourced separately.

LineMovesRunning
Claimed — VCP automation workstream underwritten at $1.2M run-rate savings across the hold$1.2M claimed / $0 provable
1. What a buyer will pay for changes the story

An acquirer's diligence accepts a value-creation claim on the same basis a QoE provider accepts an add-back: documented, realised, correctly characterised, net of offsetting cost. The operative word is documented, and documenting an improvement requires two measurements — before and after. Most ops teams capture the after; it is the one that goes in the board pack. The before is captured only if somebody decided in advance it would be needed, and in the first hundred days nobody is thinking about a diligence process four years out. Half of every value-creation claim has to be collected before the work starts, and the collection window closes without warning.

−0%$1.2M claimed, $0 provable
2. The four measurements that expire changes the story

Each cheap now, impossible later. (1) Exception volume and location — where failures go and at what weekly rate; without the pre-change number, 190/week at month 11 is not evidence of anything, it might always have been that. (2) Cycle time as a distribution, not a median — median, mean and 90th percentile monthly, or you cannot later show the tail is new rather than old. (3) The informal resolution map — who gets asked, by whom, for what; no system of record holds it anywhere, and it is consistently the largest driver of post-automation exception load. (4) Time allocation outside the target cost centre — the control group for your offsetting-cost search; without it you have an assertion and the buyer has a discount.

+$0$1.2M claimed, $0 provable — but now collectable
3. Why the window is a hundred days and not longer changes the story

Two independent mechanisms close it. The process changes — obvious, and the smaller of the two. And the people leave: the informal resolution map lives in the heads of the individuals whose roles the VCP is targeting, and mid-market portfolio companies see meaningful voluntary attrition in the twelve months after a sponsor transaction, concentrated among exactly the experienced operational staff who hold that map. You are racing the departure of your own instrumentation. That produces an uncomfortable sequencing point: the interview capturing the resolution map has to happen before the person knows their role is under review. Do it inside standard onboarding diligence in the first sixty days, framed as process documentation, because it is.

−$0unchanged
4. Where I was wrong — I scoped baseline capture to the systems where I was wrong

On a prior engagement I treated baseline capture as an analytics task and scoped it to the systems: pull the ERP data, pull the ticketing data, build the dashboard. Defensible, cheap, and it produced a genuinely good pre-change dataset on everything the systems recorded. It missed the entire informal layer, because by construction the informal layer is what is not in the systems. The correction is not more data — it is four to six structured thirty-minute interviews with the people who currently handle exceptions, asking one question: when this goes wrong, who do you ask? A day of work, and the highest-yield instrumentation available in the first hundred days.

What it cost: Six figures on a live case. When exception volume climbed post-implementation we had an excellent baseline for the transactions and none at all for who used to resolve the hard ones, so the offsetting-cost argument was unwinnable in both directions — I could not prove the load was new and management could not prove it wasn't. I under-scoped it because interviews do not look like measurement.

−$0unchanged
5. What the baseline is worth at exit

The bridge resolves to a binary rather than a range, which is unusual and is why this episode sits where it does. With the baseline, the workstream's contribution is documented in the form a buyer's diligence accepts and defends its place in adjusted EBITDA. Without it the savings are real, the margin is real, and the claim is unsupportable — so it either comes out of the adjustment or survives at a discount, depending on how hard the buyer pushes. The asymmetry is the point: capture costs a few days of analyst time and a day of interviews, in a period when you have maximum authority and nothing else competing for that specific resource.

+$1.2M provable, or −$1.2M$1.2M provable for ~$15–25K, or $0 provable for free
Realised$1.2M provable for ~$15–25K of first-hundred-days effort, or $0 provable for free

Not an overstatement — an unprovability. The savings can be entirely real and still fail to survive a buyer's diligence, because half the evidence was never collected.

The downside

Value at risk · multiple 9x, illustrative

A $1.2M run-rate workstream at an illustrative 9x. Capture cost is a few days of analyst time plus ~20 structured interview-hours — call it $15,000–$25,000 inside a hundred-day plan already spending multiples of that on advisors. Substitute your own comps.

$10.8M of enterprise value — defensibility turning on whether somebody spent a week on measurement in month two

When it surfaces. Three to five years later, in a buyer's confirmatory diligence — at which point the remedy costs nothing because there is no remedy. This is the only episode in the set where the failure is not detectable at the time it occurs, which is precisely why it belongs on the hundred-day calendar rather than the exit calendar.

THE ADD-BACK · episode 02 · diligence pack

The Baseline Is Perishable

The question: What do I have to measure before I'm allowed to cut?
Paste into a request list or a management agenda. Each question resolves to an artifact, not to a characterisation.

The diligence pack

  1. Freeze and export twelve months of history from every channel, shared inbox and ticket queue that currently receives exceptions for each process named in the VCP.

    Artifact: Raw exports, archived outside the operating systems.

    Do this before scoping, not before go-live.

  2. Produce transaction-level cycle time for each target process, monthly, twenty-four months back, reported at median, mean and 90th percentile.

    Artifact: The transaction table.

    Dashboards get rebuilt and lose history.

  3. Conduct and transcribe thirty-minute interviews with every individual currently resolving exceptions in the target processes, on one question: when this fails, who do you ask and what do they know that you don't?

    Artifact: Transcripts.

    The item that expires with attrition. It must happen before the person knows their role is under review.

  4. Provide a current time-allocation sample for every team adjacent to the target process — customer service, account management, finance ops — sufficient to establish a pre-change control.

    Artifact: A two-week sample, not an estimate.

  5. For each process, list every data element required to complete a transaction that is not present in the transaction record, and where it currently comes from.

    Artifact: Written response, from operations.

Disqualifier

Sequencing rule: items 1 and 3 before any vendor is engaged, and item 3 before any role is flagged in the VCP. If the baseline work is scheduled after scoping, it has already failed — the window closes when the process changes, not when the project ends.

Sources — every measurement with its sample

ClaimSourceSampleClass
Pre-change exception rate ~40/week rising to ~190/week by month 11; informal resolution via an undocumented corridor dependencyComposite, built to the shape of a real quote-to-cash automation (same case as Episode 01)n/a — compositecomposite
Meaningful voluntary attrition in the twelve months after a sponsor transaction, concentrated among experienced operational staffGeneral mid-market pattern, stated as a pattern rather than a measurementn/a — asserted pattern, not measured hereargued
Baseline capture cost of $15,000–$25,000Estimate, exposed so it can be substitutedn/a — illustrativeillustrative
9x entry multipleIllustrative mid-market compn/a — illustrative, arithmetic exposedillustrative

One standing caveat. Every number in this show is somebody else's measurement, and I'll tell you whose, with the sample. None of it is diligence on your deal. Do that yourself.

Episode theaddback-02-the-baseline-is-perishableVersion 1.0Dated 2026-08-03Canonical https://addback.enthropysystems.com/02-the-baseline-is-perishable/
Enthropy Systems · addback.enthropysystems.com