The Buyback Flattery Test

Knowledge Bank · Shelf: Financial Analytics

A buyback reduces the share count. That is the whole mechanism, and it is worth stating slowly, because almost everything that follows is arithmetic set in motion by it.

Every per-share figure is a total divided by a count. Earnings per share, above all. Shrink the denominator and the per-share figure rises even if the total does not move. The business can be standing still — same net income, same operations, same year as the one before — and the per-share line still climbs, because there are fewer shares underneath it.

This piece is about a comparison we call the flattery test. It has two steps: lay total growth beside per-share growth, then read what management actually paid for its own shares. We set out the method, the two engine functions that carry it, and the limits — stated inside the piece, as we state results.

The source documents are the house's "RA Analytical Engine: Master Plan" (implementation record, 26 September 2026) and the owner's "Frameworks & Concepts Index" (dividends versus buybacks and the capital-allocation decision tree, from his finance coursework).

Step one: total growth beside per-share growth

The first step is a placement, not a calculation. Put a company's total growth — total net income, first year to last — next to its per-share growth over the same span. Earnings per share is the usual per-share figure, and the one most often headlined.

Then read the gap. If the total grew 10% and the per-share figure grew 40%, the distance between those two numbers is share-count arithmetic, not operating growth. Nothing in the business had to accelerate to produce the second figure; the denominator did the work. The flattery is in the headline a reader stops at.

Our quant engine, ra-quant, carries a function for exactly this divergence, named flattery. It also carries a per-share version of any statement series, so the total and the per-share line can be laid side by side without hand assembly. The comparison is the function's entire job. It does not tell anyone what the gap means for a decision; it shows the gap.

Step two: what was paid

The second reading is price. The engine's allocation module computes buyback effectiveness: the share-count trend, and the implied average price management paid for its own shares, set against the current price. The plan's own phrasing for the question is "did they buy high?"

The arithmetic behind that question is simple. A company that retires shares at high prices transfers value to the shareholders who sold. One that retires them cheaply concentrates the remaining owners' claim. The engine states which happened; it does not judge intent.

One convention to know before reading any output: buybacks are recorded as positive magnitudes — cash outflows as positive numbers — one documented convention across all modules. A reader who assumes a negative sign will misread the series.

A capital-allocation act, read as one

The owner's finance coursework treats dividends and buybacks as the two doors of the same room — returning cash to owners — with a capital-allocation decision tree for choosing between them. We read a buyback the same way: as a capital-allocation act. Cash spent here was not spent on plant, debt, or dividends. That framing matters because the flattery test measures only the share-count effect. The opportunity cost sits outside the measure, in the decision tree, and a reader who wants the whole picture needs both.

What the test is not

A buyback is not wrongdoing, and divergence is not deception. Companies disclose buybacks. The flattery is not in the filing; it is in the headline a reader stops at. Earnings-per-share growth is the growth number most directly manufactured by the denominator — which is precisely why it is headlined so often. Those two facts sit together, and the test exists to keep them in view at once.

The limits, stated plainly

The engine's growth module for organic-versus-acquired decomposition was deferred. The flattery measure therefore covers only the buyback-divergence case: growth bought by acquisition is not separated by this test.

The implied average price paid is derived from statement aggregates — buyback spend and share-count change — so it is an approximation across the period, not a trade-by-trade record. It cannot tell you what was paid on any given day, only what the period's aggregates imply.

Share counts also move for other reasons: option exercises, issuance, splits. The divergence attributes the net effect of all of them, not each cause. A gap between total and per-share growth is share-count arithmetic, but the test does not sort that arithmetic into its components.

The test reads completed periods only. Announced buyback authorizations are intentions, not retirements. An authorization is not a share retired, and nothing in this measure treats it as one.

Status of the tooling, in two parts: the allocation analysis — buyback totals included — runs live in Research Analyst today. Both functions — flattery and buyback effectiveness — are also carried in ra-quant, the house's standalone package of the same machinery, test-verified on its development branch; the package itself has not been merged as a separate artefact, and its own interface panel is not wired.

That is the method and its edges. The comparison is available to anyone with two statement series and a share-count line; the product's engine does this work today, with the conventions held consistent. What the numbers mean for any particular company, and what anyone does with them, is work we leave where it belongs — with the reader.