If you've read even one PE diligence memo, you've seen an "adjusted EBITDA" number that's meaningfully higher than the reported figure. Every buyer adjusts. The interesting question is which addbacks a serious analyst signs off on and which ones would embarrass them in a bank meeting.
The all-time embarrassment is WeWork's "Community-Adjusted EBITDA." It's the metric that made a whole industry pause and ask what exactly EBITDA is supposed to represent — and it's still the cleanest teaching example of what happens when addbacks stop pretending to be adjustments and start doing structural work.
What Community-Adjusted EBITDA actually was
WeWork Companies debuted "Community-Adjusted EBITDA" in the prospectus for its first bond offering — a Rule 144A high-yield deal that priced in April 2018. The definition, in plain English, was earnings before you subtract almost anything.
The metric started from GAAP net loss and then added back:
- Interest, taxes, depreciation, amortization (the standard EBITDA build)
- Building- and community-level operating expenses — rent, utilities, internet, the salary of the community manager greeting members at the door, building amenities
- Sales and marketing expenses
- General and administrative expenses
- Growth and new-market development expenses
- Pre-opening community expenses
Read that list again. WeWork's core business was operating buildings and filling them with members. The primary cost of doing that — the lease, the coffee, the manager on the floor — was being added back to arrive at "profitability."
For 2017, the resulting number was a positive $233 million on $886 million of revenue, alongside a GAAP net loss of $933 million. So the same year the company lost nearly a billion dollars, one line item on the prospectus said the business generated $233 million in earnings.
That is not adjusted EBITDA. It's contribution-margin math with a marketing label.
Why the metric existed at all
To be fair to WeWork's finance team, there is a real accounting question buried in here.
A newly opened WeWork location loses money for months. It signs a 10-year lease, spends heavily on buildout, and takes time to stabilize occupancy. If half your locations are less than 18 months old, your consolidated P&L will look terrible even if the mature locations are cash-generative. Contribution margin at the property level is a defensible cut for investor discussion — as long as you call it what it is.
The problem was the label and the placement. By calling this metric "EBITDA" and putting it at the top of the covenant discussion, WeWork blurred two very different concepts:
- Property-level contribution margin (before corporate overhead) — a real internal KPI
- Company-level EBITDA — what lenders and PE buyers use to size debt
Bond investors sizing coverage ratios don't care about property contribution before G&A. They care about the cash left over to service interest — which is why interest coverage tests are built on operating income, not on a hand-crafted metric that strips out the largest cost lines.
Covenant Review's Adam Cohen, quoted in press coverage at the time, told the FT the phrase was one he had never seen before. He was right — nobody had, because it was invented for that document.
The April 2018 bond deal — where the metric was born
Context matters. WeWork upsized its debut bond from $500 million to $702 million on the back of that prospectus, with the offering reportedly five times oversubscribed. Coupon-hunting credit investors were willing to underwrite the story.
Here's the pattern the S-1 pages surfaced, using the 2017 numbers WeWork disclosed:
| Line item | 2017 (reported) |
|---|---|
| Revenue | $886M |
| GAAP net loss | ($933M) |
| Community-Adjusted EBITDA | $233M |
A single non-GAAP line converted a nine-figure GAAP loss into a nine-figure positive. The magnitude of that swing — over a billion dollars between two rows on the same page — is what made analysts stop and re-read the definition.
What happened when the S-1 landed
The We Company filed its Form S-1 with the SEC on August 14, 2019. By then WeWork's 2018 numbers had doubled — revenue of roughly $1.8 billion and a GAAP net loss of roughly $1.9 billion — and the disclosure standards of a public offering were about to be applied to the community-adjusted vocabulary.
The S-1 quietly retired the "Community-Adjusted EBITDA" label. In its place, the company reported "Adjusted EBITDA" (using more conventional addbacks) and disclosed property-level economics under a "Contribution Margin" heading. The substance was closer to the same math — the label was different.
It didn't matter. By the time the roadshow started, the rest of the S-1 — the related-party real estate transactions, the tri-class supervoting stock, the "ClassCorp" trademark licensed back to the company by the CEO — had made "Community-Adjusted EBITDA" a synecdoche for the whole filing. The valuation reportedly collapsed from ~$47 billion in the initial marketing to under $10 billion.
The Registration Statement was formally withdrawn on Form RW on September 30, 2019, six weeks after it was filed.
What this teaches you about EBITDA addbacks in PE diligence
WeWork was extreme, but the underlying question — which addbacks are legitimate and which ones aren't — comes up in every PE deal. When a seller hands over a Quality of Earnings report, the "Reported EBITDA to Adjusted EBITDA" bridge is where you focus first.
There are four categories of addbacks, and only two of them are usually defensible.
Category 1: Non-recurring items (defensible). One-time legal settlement, a natural-disaster loss, an executive severance from a departure that isn't going to repeat. If the buyer can look you in the eye and say the cost genuinely won't exist in NewCo, the addback holds.
Category 2: Owner-related adjustments (defensible with scrutiny). The founder's above-market salary, the corporate jet, personal expenses run through the business. NewCo is going to replace the CEO or normalize comp; that saving is real. But you need to see the underlying detail — "$800K of owner comp addback" without documentation is where the bridge starts sliding.
Category 3: Run-rate adjustments (be careful). Annualizing a new customer that signed in December. Adding back synergies you haven't captured yet. These aren't wrong on principle, but they're forecasts dressed up as historicals. Lenders will discount them heavily; sponsors should too.
Category 4: Operating cost addbacks (rarely defensible). Rent. Marketing. G&A. If it's an ongoing cost of running the business, it's not an addback — it's an operating expense. This is the WeWork failure mode: dressing up "we don't want to count rent" as "community adjustment."
Category 4 addbacks are the "smell test" line. A quality of earnings that leans on them is a quality of earnings that isn't going to survive lender diligence.
Where creative EBITDA actually kills deals
The real damage isn't disclosure embarrassment. It's covenant math.
Every leveraged loan defines EBITDA precisely in the credit agreement. The definition matters because the interest coverage covenant, the leverage covenant, and the excess-cash-flow sweep all reference it. If the borrower's "adjusted EBITDA" includes categories that the credit agreement's definition doesn't, the borrower is going to trip a covenant sooner than the model says.
This is where PE buyers pay real attention to EBITDA definitions. A deal underwritten with 6× leverage on management's adjusted EBITDA might actually be 8× on the credit agreement's definition. That's the difference between comfortable coverage and a covenant breach on the first bad quarter.
The mechanic here — sizing debt against EBITDA and then testing coverage — is the same one we walk through in the debt capacity calculator. You set an EBITDA number, pick a leverage cap and an interest rate, and see whether the binding constraint is the leverage multiple or the coverage ratio. It's a 30-second sanity check that would have flagged the WeWork bond covenants as tight if anyone had run them off GAAP EBITDA rather than the community-adjusted version.
How to read an EBITDA bridge in practice
If you're on the buyside and someone hands you an adjusted EBITDA number, the workflow is roughly:
- Start from GAAP operating income. Don't start from adjusted anything. Get to a real operating income number from the audited financials.
- Add back D&A explicitly. You now have unadjusted EBITDA. Write it down.
- Walk each addback line, one at a time. Ask: is this genuinely non-recurring? Is there support (invoice, contract, board minutes)? Would a bank definition allow it?
- Recompute the leverage ratio at the unadjusted number. If the deal only pencils at "adjusted" EBITDA, you're taking model risk that the addbacks stick.
- Model a downside where 50% of the addbacks fail. If half the adjustments are contested at exit, does the return still work?
Junior associates skip step 4 and 5 all the time. Senior partners never do. And the reason is exactly the WeWork lesson — you can label anything "adjusted EBITDA," but the number the credit agreement enforces and the number a strategic acquirer pays for at exit are both closer to the unadjusted line.
The takeaway
Community-Adjusted EBITDA became a punchline because the label was too obviously trying to do the work the numbers wouldn't. But every PE model has an addback bridge, and every bridge has at least one line item that would look worse under scrutiny. The discipline is knowing which lines you'd defend to a lender at 5 p.m. on a Friday and which ones you'd quietly remove first.
If you want to see how EBITDA definitions ripple through returns math, our All-in-One PE Model has explicit inputs for reported EBITDA, adjusted EBITDA, and the leverage covenant EBITDA, all feeding a single S&U table and coverage schedule — so you can see what a 5% swing in addback approval does to your return profile before you send the model up. The related MOIC vs IRR discussion covers why the same swing hits IRR harder than MOIC on shorter holds. See the sheet preview before buying.
The label matters less than the underlying number. But if you're calling it EBITDA, at least don't add back rent.
Sources:
- WeWork Companies Inc., Form DRS (draft registration statement), SEC EDGAR, sec.gov/Archives/edgar/data/1533523/000162827918000370 — WeWork Companies filing detail
- The We Company, Form S-1, filed August 14, 2019, sec.gov/Archives/edgar/data/1533523/000119312519220499/d781982ds1.htm — the IPO registration statement
- We Co., Form RW (registration withdrawal), September 30, 2019, sec.gov/Archives/edgar/data/1533523/000119312519258913/d808793drw.htm — official IPO withdrawal
- Quartz, "Get ready for some creative accounting in WeWork's IPO filing," qz.com/1685919 — Community-Adjusted EBITDA definition and context
- CNBC, "WeWork says revenue more than doubled last year to $1.8 billion, but so did its net loss," March 25, 2019, cnbc.com — 2018 revenue and net loss figures
- Crunchbase News, "WeWork Loses $1.9B On $1.8B In Revenue During 2018," news.crunchbase.com — 2018 net loss and revenue detail
- Fortune, "WeWork IPO Filing Withdrawn," September 30, 2019, fortune.com — IPO withdrawal and valuation collapse