The Reach for Growth – What Happens When Software Companies Spend Margin to Grow Faster [1]
Quick Summary
Across 121 episodes where a healthy software company let its margin fall by three points or more, the growth rate two years later was 4.1 points lower than where it started. For the 123 companies that held margin steady, it was 2.8 points lower. Spending did not help; if anything it hurt.
Forty percent of the companies that spent grew faster afterwards. So did thirty-nine percent of the companies that did not. The two groups are indistinguishable.
The median company converted a point of margin into minus 0.42 points of growth. Only 29% achieved enough to justify the spend even on the most generous valuation of growth, and 22% on a stricter one.
This is not proof the move never works — margin often falls because growth already slowed, not as a choice. But it shifts the burden of proof. There is no sign in seven years of data that cutting margin is followed by faster growth.
Introduction
Every software company that slows down faces the same proposal. Growth has come off, the market has noticed, and someone argues for investing through it — more sales capacity, more product, more market development — accepting a thinner margin now to restore the growth rate later. It is the most common strategic move in the industry. This paper asks whether it works, using the companies that actually did it.
What the trade has to return
Spending margin to buy growth is a trade, and it is worth knowing the terms before making it. Comparing software companies of the same size, the market values a point of revenue growth at roughly three points of adjusted EBITDA margin. Growth is worth more than margin. That part is not in dispute, and it is why the proposal gets made.
Which gives a break-even test that any management team can apply to its own plan. Take a company with $500 million of revenue. Giving up three points of margin means spending $15 million a year it used to keep. To come out level it needs that spending to add about one point of growth, which is $5 million of additional annual revenue. Three dollars of margin for one dollar of new revenue. Beat that and the trade creates value; fall short and it destroys value.
That is the forgiving version of the bar. Measured among companies already growing at a healthy clip rather than across the whole industry, the market's terms are closer to two to one, which demands roughly twice as much growth for the same margin. Everything that follows uses the easier test, so the results are the generous reading [2].
What actually happened
To test it, take every software company that was already public before 2017, so that no post-flotation ramp distorts the picture. Find each occasion when a healthy business — profitable, and not already shrinking — let its adjusted EBITDA margin fall by three points or more in a single year. Then measure its growth rate two years later, allowing time for the spending to reach revenue. Compare that against the companies whose margin held within a point over the same stretch.
The two distributions sit on top of each other. Both groups decelerated, which is what the industry did over this period. The companies that spent ended 4.1 points lower than they started; the companies that held ended 2.8 points lower. Spending bought nothing, and the spenders have the fatter tail of severe declines.
Measured against the break-even test, 29% of these attempts were worth making. On the harder version of the bar, 22% were. Only 40% produced any additional growth at all, and the median company produced none: it gave up the margin and ended up growing more slowly than before [3].
The payoff is also lopsided, which matters more than the hit rate. Among the companies whose growth rate moved at all, the median gain was 6.2 points and the median loss 10.6 — the downside was close to twice the upside. A management team weighing this is not facing a coin flip with even stakes. It is facing a bet that is lost three times in five, and lost bigger than it is won.
Why a good price and a bad trade are not a contradiction
If growth really is worth three times margin, why does buying it fail so often? Because the price is set by comparing companies that already sit where they sit. It says a company growing 16% at a 22% margin is valued like one growing 15% at 25%. It does not say the second company can turn itself into the first.
Three things break that inference. Companies that grow faster may do so because they have better opportunities, not because they spent more — the spending and the growth share a cause. The timing is asymmetric: margin falls in the quarter the money goes out, and growth arrives several quarters later if at all. And the companies visible at 16% and 22% are the ones whose spending worked. The ones where it did not are sitting at 15% and 18%, and that is a population which does not advertise itself.
So the price is a ceiling on what reaching can be worth, not a forecast of what it will be worth. What actually happened is the better guide to the odds, and the odds are poor.
What a management team should take from this
The finding is not that investment is futile. It is that buying growth with margin is a transaction, it has a price, and most companies cannot transact at that price. A company contemplating the move should be able to say what conversion rate it expects and why it believes it sits in the top quartile rather than the median, because the median outcome is no growth and a permanently thinner margin.
It is also worth treating this the way the same company would treat any other use of capital. A software business weighing an acquisition, a new facility or a product line applies a hurdle rate, and would not fund something whose expected return merely matched the cost of the money. The same discipline belongs here, with one adjustment. The three-to-one price is already a discounted figure — a market capitalization is the market's own view of future cash flows, so the price of growth already reflects the risk and the waiting the market attaches to it. A cost of capital should not be layered on top of it; that charges for the same discounting twice.
What the price does not contain is the company's own execution risk, and that is what the record supplies. Fewer than a third of attempts cleared the bar, and the typical failure cost close to twice what the typical success gained. A hurdle built on those odds sits well above the market's price, which means a plan that only just clears three-to-one is not really clearing anything. The timing runs the wrong way too: the margin goes immediately and in full, while the growth, if it arrives, arrives later and only pays for itself if it is sustained.
The uncomfortable corollary is that a company which has genuinely run out of growth may be better served by accepting it than by spending down its margin to chase it. That is close to the opposite of the advice such a company usually receives, and it is the kind of conclusion that is easier to accept from a distribution of 121 attempts than from an argument.
The obvious objection, and what the data says about it
The first thing a careful reader will say is that these were not decisions. A margin decline is frequently a symptom rather than a choice: growth slows, operating leverage runs backwards, and the margin follows. On that reading this test is measuring companies already in trouble and attributing their trouble to a decision they never made.
The objection is partly right, and it can be measured rather than argued about. Going into the episode, fifty-five percent of the companies that went on to spend margin were already decelerating, against forty-five percent of those that held it. The median company in the spending group had seen its growth rate fall over the previous year; the median company in the holding group had seen its rise, by 0.5 points. The two groups are not identical going in, and some of what looks like failed investment is a company that was already sliding.
But the confound can be removed, by keeping only the companies whose growth was steady or rising at the moment the margin fell — cases where the cut looks like a choice rather than a consequence. That subgroup is about half of each population. The quantity to compare is the penalty for having spent — how much worse the spenders did than the holders they are measured against — and it is worth working out twice.
Across all the episodes, spenders ended 4.1 points lower and holders 2.8 points lower. The penalty for spending is the difference between those two: 1.3 points.
Among only the companies that were not already slowing, 54 spenders ended 9.5 points lower and 67 holders ended 5.7 points lower. The penalty there is 3.8 points.
So removing the confound raises the penalty for spending from 1.3 points to 3.8, nearly tripling it. Stripping out the symptom effect does not shrink the difference between spending and holding. It widens it. Whatever that effect is doing to this data, it was masking the finding rather than manufacturing it.
A related worry is timing. A company spending into a decelerating market will see its growth fall whatever it does, and the comparison would be measuring the market rather than the decision. The industry as a whole was not decelerating over this period — aggregate growth ran between 8.7% and 17.9% from 2017 to 2025 with no trend in it — but the two groups are not spread evenly across those years. Spenders are over-represented in 2021, just before the 2023 reset, and holders in 2023, just after it. Comparing each spender only against holders that began in the same year removes that skew, and the penalty comes out at 1.5 points rather than 1.3. Matching on market segment instead, so that a company is measured against others selling into the same demand, gives 2.4 points. Every way of holding conditions constant makes the penalty larger, not smaller.
What this does not establish
The episodes are still not experiments, and the tests above narrow the objections without eliminating them. The segment matching in particular is thin: several of the fifteen market segments hold too few episodes to pair at all, so a company whose own niche was maturing faster than its segment average would not be caught. Episodes from the same company in consecutive years are not independent. Two years may be too short for investments with long payback. And the test cannot see what the spending bought that is not revenue — retention, product, or a position that pays later. What it does establish is that a company planning to spend margin on growth should not expect the industry's experience to support it.
Endnotes
[1] Based on active public software companies that were listed before 2017 · annual results 2017 to 2025 · Source: EODHD, Worldlocity analysis
[2] What a point of growth is worth relative to a point of margin is the ratio c to d from log(market capitalization) = a + b·log(revenue) + c·growth + d·margin, where log(revenue) holds size fixed. It is two fits, not one. The three-to-one figure is 3.2 to 1 across 235 companies, excluding the most extreme two percent at either end on both growth and margin; across all 257 it is 2.9 to 1. The two-to-one figure is 1.8 to 1 across the 174 companies growing between 5% and 30%. As break-even conversion rates those are 0.31 and 0.56 points of growth per point of margin. Market capitalization is used rather than the revenue multiple because that multiple shares its denominator with margin, which would tie the two together mechanically.
[3] Per point of margin given up, the median company converted minus 0.42 points of growth and the upper quartile plus 0.48 — enough to clear the easier bar, not the harder one.
[4] In this article margin means adjusted EBITDA, which equals EBITDA plus stock-based compensation.
Method note: the population is active public software companies that were listed before 2017, which keeps out the post-flotation margin ramp that would otherwise read as a company improving. Within it, an episode is any year in which a company that was profitable and not shrinking — an adjusted EBITDA margin of at least 5% and a growth rate at or above zero — then saw that margin fall by three points or more over the following year. The comparison group is every year in which such a company's margin instead held within one point. Each episode is scored on the change in its revenue growth rate between the year it began and two years later, which allows time for spending to reach revenue. That gives 121 spending episodes and 123 holding episodes across 2017 to 2025. A company contributes more than one episode where it qualifies in more than one year, so the episodes are not fully independent of each other. What a point of growth is worth against a point of margin is estimated separately, from a single cross-section of the 257 companies carrying a growth rate, a margin and a market capitalization, as set out in the endnotes.
Comments