Normalising SME EBITDA

Give two competent accountants the same SME P&L and the resulting normalised EBITDAs are likely to be 15 to 25% apart. Not because either is wrong, but because normalisation is a series of judgments, and without a shared framework two reasonable people produce two reasonable answers that don’t agree. On a 5× multiple, that gap is €750k to €1.25M of valuation uncertainty before any multiple discussion has started.

A Belgian BV or Dutch BV P&L is fiscally optimised. That’s good accounting, not a flaw. But it isn’t economic reality. Between fiscal presentation and economic reality lies normalisation. Six categories where most of the divergence happens:

1. Market-rate owner compensation
2. One-off costs and benefits
3. Non-operating items
4. Working capital corrections
5. Maintenance capex vs. depreciation
6. Intercompany and related-party transactions

1. Market-rate owner compensation

Almost always the largest adjustment, and almost always wrong in both directions. Owner-managers are typically over- or underpaid relative to a market-rate non-owner CEO. The normalisation question: what would a professional CEO command for this size, sector and geography?

2. One-off costs and benefits

Hidden because they sit inside operating lines, not labelled "one-off." Add back legal disputes, restructuring charges, Covid-only costs, one-off IT projects, fines, asset disposal losses. Subtract one-off subsidy income, asset sale gains, provision releases, non-recurring project revenue. The trap: sellers want only the costs added back. A defensible normalisation is bidirectional.

3. Non-operating items

Structural P&L items that aren’t operating: family vehicles, owner-property rent above or below market, investment income, holding costs, sponsorships. The most-forgotten case: when the property stays with the seller and the new owner will pay market rent. Normalise current (often off-market) rent to market.

4. Working capital corrections

Doesn’t hit EBITDA directly but hits valuation hard via the working-capital target at closing. Use a rolling-12-month average, not the year-end snapshot. Strip seasonal noise. A retail business with €600k year-end stock but €1.1M rolling average means the new owner injects €500k of working capital not visible in the snapshot.

5. Maintenance capex

EBITDA strips depreciation. Capex isn’t in EBITDA but it’s very much in free cash flow. Two businesses with identical EBITDA but capex of €50k vs. €600k yield €950k vs. €400k of free cash flow. Estimate steady-state maintenance capex from a 5 to 10 year average, sector benchmarks (manufacturing 4 to 8% of revenue, services 1 to 3%) and any deferred investments that are really overdue.

6. Intercompany / related parties

Where most deliberate fiscal optimisation hides: management fees, royalties, intra-group loans, related-party rent, related-party purchases. Normalise these to arm’s length. Otherwise the buyer is buying a set of transactions that won’t survive on their own.

What makes a normalisation defensible

1. Bidirectionality. Adjustments go both ways.
2. External grounding. Every adjustment cites an external reference.
3. Logging. Full audit trail per item.
4. Consistency. Same rules across the 3 to 5 year history.

Conceptually this is not complex. It is execution-heavy. Three to five days for an experienced professional, on every single deal. That is exactly why it is a software problem. Not because the rules are hard, but because applying them consistently at scale is the only way the lower mid-market gets to a defensible valuation. That is what Upswitch’s normalisation engine and integrity guards exist to do. Cross-check every normalised EBITDA band against Upswitch Index sub-segment data, sectors with the highest normalisation spread (manufacturing, B2B services, hospitality) are exactly where the six categories swing the most.