Playbook › Playbook
Buyback-manufactured EPS growth screens identically to operating growth, and the market pays the same multiple for both
Claim
Adjusted EPS growth answers "what did a share earn?" It does not answer "did the business earn more?" Those two questions diverge completely when the share count is falling fast, and every screener, every headline and most sell-side models report only the first.
The natural experiment — three prints, one sector, one day
On 2026-08-04, Gartner, Leidos and Broadridge all reported. Yahoo classifies all three as Information Technology Services. The market's reaction ranked them in exactly the inverse order of their operating performance.
| IT | LDOS | BR | |
|---|---|---|---|
| Move on the print | +19.0% | +8.8% | +5.6%* |
| Revenue trend | guided −1.9% | +7% | +8.6%, accelerating |
| Operating income, 3–4yr | +6.3% ($1.11→$1.18B) | +86% ($1.13→$2.10B) | +57% ($760M→$1.19B) |
| Operating margin | 20.3% → 18.2% | 7.8% → 12.2% | 13.3% → 17.3% |
| FCF | guided flat | +23.5% CAGR | +41.8% CAGR |
| Headline EPS growth | +24% adj | +1.6% underlying | +12% adj |
| Buyback ÷ FCF | 169%, w/ $800M new debt | ~58%, cash-funded | ~13%, cash-funded |
| Book equity | $320M, P/B 199× | $4.92B | $2.66B |
*since 7/29, not a single session.
The biggest one-day gain in the group went to the only company whose operating income has not grown, whose revenue is guided down, and whose free cash flow is guided flat. Gartner's +24% adjusted EPS is arithmetically sound and economically hollow: FCF guided $1.185B against $1.18B (+0.4%) while the share count fell 70.85M → 66.95M in two quarters (−5.5%).
The mechanism
EPS = earnings ÷ shares. Hold the numerator flat, shrink the denominator 5% a year, and EPS compounds 5% a year forever — on a screener, indistinguishable from a business growing 5%.
Three things make it look better than organic growth rather than merely equivalent:
- It is smoother. Buyback pace is a management decision; demand is not. The EPS series has lower variance, which reads as quality.
- It survives revenue decline. Gartner's revenue guide is negative and adjusted EPS is +24% in the same release.
- It flatters every per-share ratio at once — EPS, FCF/share, book value per share (downward, which raises ROE). Gartner's ROE is 95% because equity is nearly gone.
Why the end date is invisible
The programme runs until the funding does. Gartner spent 169% of FCF on stock in FY2025 and issued $800M of new debt; book equity fell $1.36B → $320M. None of that appears in a P/E, a PEG, an EPS growth rate, or a forward estimate. The metric that ends the story is absent from every metric that tells it.
This is not a claim that the buyback is wrong. Retiring a 19%-FCF-margin subscription business at ~10× FCF is defensible arithmetic. It is a claim about what the growth is made of and how long it can last.
Diagnostics
- Operating income CAGR vs EPS CAGR. The gap is share count. If operating income is flat and EPS is +24%, you have found it. One line, catches it every time.
- Buyback ÷ FCF. Under 100% = funded by the business. Over 100% = funded by the balance sheet, and check debt issued in the same year.
- Book equity trend. A collapsing equity base with rising ROE is the same fact reported twice, and the second reporting looks like quality.
- Revenue guide vs EPS guide. Divergence in sign is conclusive.
- Compare FCF per share growth against FCF total growth. If total FCF is flat and per-share FCF is compounding, the compounding is the denominator.
The inverse is also a signal
LDOS and BR both grew operating income faster than revenue and funded buybacks entirely from cash flow — and both trade below their fair-value centres while the manufactured story trades above its own. This pattern is as useful for finding what the market has underpaid for as for flagging what it has overpaid for.
Sibling claim: [[pitfall-unrealized-equity-marks-break-headline-pe]] — that note covers a fake numerator (non-operating marks); this one covers a shrinking denominator. Both produce EPS growth without business growth, and both are invisible to a P/E screen. Check for both.
Related: [[pattern-ai-build-inflates-earnings-while-destroying-fcf]] · [[pitfall-gate-cleared-by-adjective-not-arithmetic]]
History
- 2026-08-04 — Established from three same-day, same-sector prints, which is an unusually
clean control: same tape, same day, same classification, opposite economics. Filed
slowrather thanstatic— the mechanism is permanent arithmetic, but the prevalence depends on a rate environment where debt-funded buybacks are affordable. Re-check the Gartner instance specifically after FY2026 closes: with book equity at $320M and the buyback at 169% of FCF, the funding constraint should become visible within four to six quarters, and that will either confirm or refute the "computable end date" half of this claim.