IMP — Impact Developer & Contractor S.A.

Deep-value report (REFRESH): Impact Developer & Contractor S.A. (BVB: IMP)

Report date: 2026-09-03 · Reference price: … (previous BVB close, yfinance IMP.RO, 2026-09-03; intraday currentPrice 4.00) · Market cap: 476.5 mn lei · Shares: 118,247,071 · Reporting currency = price currency = RON (no conversion risk in the indicators)

Regime: REFRESH on top of the thesis of 2026-07-24 (41 days). The mechanical triage (deep_delta.py) flagged a single unverifiable falsifier — “the 82% sales slowdown in Q1 2026 is transitory” — which required two consecutive quarters of revenue growth data and had only one. In the meantime the company published the H1 2026 semi-annual report (August 19, 2026), which supplies exactly the missing quarter. This report re-derives what the new filing moved and explicitly carries over the rest.

Source and issuer-identity note (read before any figure)

Symbol collision, resolved: IMP is Impact Developer & Contractor S.A., a Romanian real estate developer listed on BVB (Yahoo IMP.RO), confirmed via query_tracker.py indicatori IMP → “name=Impact Developer & Contractor S.A., sector=Real Estate/Construction, exchange=BVB.” The issuer has no SEC EDGAR CIK; the EDGAR step was not and will not be run for it.

The web research brief supplied for this session (research-IMP-20260903.md, 9,136 characters) is entirely about a different company — Impala Platinum Holdings (JSE: IMP), a South African platinum-group-metals producer, with FY2026 results in South African rand, dividends of 1,855 cents/share and mines at Rustenburg. Not a single figure in it relates to the issuer analyzed here. I rejected it in full and used no value from it; this isn’t a detail discrepancy, it’s a company-identification error. All the context research in this report was redone by me point by point, with a source for every figure.

The primary sources actually used (all downloaded and extracted to text by me in this session):

• bvb.ro/infocont/infocont26/IMP_20260819181446_Raport-S1-2026-IMP.pdf — H1 2026 semi-annual report, 125 pages, consolidated interim financial statements at June 30, 2026, approved by management on 08/19/2026, unaudited (the press release states this explicitly). The dominant source of this refresh.
• bvb.ro/infocont/infocont26/IMP_20260819181430_Comunicat-S1-2026-RO.pdf — the accompanying press release.
• m.bvb.ro/infocont/infocont26/IMP_20260526181524_Raport-T1-2026-IMP-RO.pdf — Q1 2026 Report, 126 pages (for the Q1↔Q2 bridge and for the comparative debt/inventory notes).
• data-pack-IMP-20260903.md (yfinance, 8 quarters) — used only where the filing doesn’t give a quarterly breakdown (EBITDA for Q3/Q4 2025), and flagged as such.
• GBL tracker: query_tracker.py scor IMP → 65.8% — INTERESTING (Graham 6.0 / Buffett 5.5 / Lynch 7.5; total 19.0/30; F-Score 5.0), assessed 2026-07-24. This is the PRIOR score, which I compare against in the final chapter.

Exchange rate used: …/EUR, derived from the H1 2026 report itself (revenue 112,156 thousand lei = 21,807 thousand euros), not from an external quote. All conversions in this report use it.

What I carry over from the 07/24/2026 report, without re-verification (nothing material has changed and there’s no new filing that contradicts them): the description of the project portfolio and its history; the management history (CEO Dan Sebastian Câmpeanu since 06/01/2024, CFO Claudiu Bistriceanu since June 2024); the dividend history (last one in 2020, none 2021-2026) and the policy announced for 2027; the historical financing structure and the 20:1 share consolidation of April 2025; the EcoCivica litigation; the impossibility of reconstructing 5-year historical multiples due to the 20:1 consolidation. What I fully re-derive: earning power, the FCF bridge, earnings quality, the balance sheet over the last 4 quarters, the valuation, the Monte Carlo simulation and the verdict.

A process correction on the previous run, explicitly flagged: the Monte Carlo simulation frozen in the delta package is arithmetically invalid. mc_dcf.py takes --oe in millions; the July run was given … (absolute lei), producing a “median intrinsic value” of 6,380,880 lei/share and a median MOS of “ ….” It’s not a reading error, it’s a factor of 10⁶. The corresponding JSON file didn’t even exist on disk when this session opened. I re-ran the simulation from scratch, with correct units; all Monte Carlo figures in this report come from mc-IMP-20260903.json generated today.

Executive summary

What changed, in one sentence: the reference thesis’s central falsifier broke, and the semi-annual report shows the slowdown wasn’t a one-quarter transitory shock, but a contraction that repeated identically the following quarter and drained the company’s cash engine.

The falsifier, tested against primary figures. The July thesis assumed the 82% collapse in Q1 2026 unit sales (16 units vs 87) was a deferral shock tied to the VAT rise to 21%. The test required two consecutive quarters. We now have them, from the 08/19/2026 filing: consolidated revenue fell … in Q1 2026 (55,415 vs 86,934 thousand lei) and … in Q2 2026 (56,741 vs 89,715 thousand lei, derived as H1 minus Q1). It’s not a rebound, it’s the same band, two quarters running. Residential unit sales in H1 2026: 31 units, versus 153 in H1 2025 — …; area sold 2,738 sqm vs 12,302 sqm (…); the cost of residential units recognized in cost: 13,985 vs 72,992 thousand lei (…). The falsifier doesn’t hold. The “transitory shock” thesis is refuted by data, not opinion.

A discovery more serious than the slowdown: reported profit is entirely revaluation. Consolidated operating profit in H1 2026 is 20,240 thousand lei — but it contains a line of “Gains from investment property revaluation” of +21,635 thousand lei, non-cash, IAS 40. Excluding it, H1 2026 operating profit is −1,395 thousand lei, i.e. an operating loss, and pre-tax result becomes −10,572 thousand lei. The company itself confirms the figure in its own actual-vs-budget table: budgeted investment-property gains 0, actual 4,207 thousand euros, in a half where actual EBITDA was 4,518 thousand euros. “Real” EBITDA, ex-revaluations, for H1 2026 is 311 thousand euros ≈ 1.5 mn lei — on a company with 476 mn lei market cap.

This retroactively invalidates the July valuation base. The previous report capitalized an EBITDA of 106.2 mn lei over 4 quarters (Q2’25–Q1’26). Of that window, 62.88 mn lei was IAS 40 revaluation gains (29,132 thousand lei in H1 2025 + 33,748 thousand lei in H2 2025; the annual total of 62,880 thousand lei is confirmed in note 7 of the Q1 2026 report, Colliers revaluation at 12/31/2025). The 106.2 mn lei base was thus inflated 59% at the very date it was used. On the current window (Q3’25–Q2’26) it’s worse: EBITDA ex-revaluation 19.9 mn lei.

The cash engine has stopped. Consolidated CFO: −17,603 thousand lei in H1 2026, versus +109,860 thousand lei in H1 2025. On the current TTM window (Q3’25–Q2’26), CFO = 14.2 mn lei, versus the 95.4 mn lei on the window aggregators still use. FCFE, after interest (reported under financing, as with any IFRS issuer), lease principal and dividends to minorities, is −5.1 mn lei. Cash fell from 40,402 to 28,471 thousand lei (…), while current debt rose to 95,192 thousand lei — cash/current-debt coverage 0.30×, with 50.2 mn lei of bonds due in December 2026 and February 2027.

Estimated value — a range that’s split in two. Earnings-based and asset-based methods no longer converge at all, and that’s the information, not a model flaw:

Median of the four: …. Range: … … …. Monte Carlo on the earnings arm (20,000 scenarios): median …, probability of undervaluation 0.0%.

Verdict: WATCH — thesis suspended. Downgrade from “INTERESTING, with reservations” (07/24/2026). The discount to net assets remains real and large (0.46× IFRS book, 0.35× Colliers market NAV), but the single mechanism that would close it — apartment sales converting inventory into cash — has stopped for two consecutive quarters, and the company has simultaneously entered a … construction cycle with negative FCFE and 50.2 mn lei of bonds to refinance in eight months. This isn’t a short thesis: the assets are real, valued by an external appraiser, 82% unmortgaged, and pre-sales at ARIA Verdi (34 units, … mn) show the premium segment hasn’t been hit. It’s a thesis that lost its proof and should only reopen on two verifiable conditions: the IMP26E refinancing before 12/24/2026 and a quarter with revenue growth exiting the −36/… band.