We have two dynamics crashing into one: software companies with technically great distribution across verticals are cheap to buy, and the data of their enterprise clients is barely being used.
Horizontal software companies are currently valued at exit multiples of 3X. VCs have to write them down on their balance sheets.
At the same time, software itself is becoming increasingly commoditized.
So the question becomes: what are you actually buying when you buy a software company?
Increasingly, I think the answer is distribution.
And distribution gives you something else: telemetry.
Existing Software Consolidators
The talk of the town among SF fund managers is that there is an opportunity to leverage the data across the portcos of software consolidators in a different way.
Software consolidators already own companies that are deeply embedded in the workflows of highly specific target groups. Historically, these businesses have been valued largely on ARR. But recent pricing shifts suggest that ARR no longer carries the same significance it did 10 years ago. In software, ARR may increasingly become little more than a token spend or an extension of the underlying customer relationship, rather than the primary metric driving company valuations.
On the contrary the behavioral data across those companies is barely part of the equation and therefore rarely taken into consideration.
What makes this interesting is not just the data of one company. It is the patterns across the portfolio.
Can you use them to make better predictions and forecasts? Can that create some form of alpha? And if it does, does that increase the value of the portfolio beyond ARR?
At Octrace, my first company, we tried something adjacent using publicly available data.
Octrace taught me that scraping public data doesn’t give you alpha. If everyone can access the same information, the information itself isn’t the edge. That changes when you look at data that is difficult to access, proprietary, or embedded in closed systems. The harder the data is to obtain, the more defensible the potential insight becomes.
The more interesting question is whether the same approach works with behavioral data from existing software consolidator portcos. (under NDA and with the right consent structure)
If you can see what businesses actually do, rather than what they say they do, can you predict the market better?
And if yes, what is that data worth?
Distribution as an Asset
This leads to a different way of thinking about software.
Software may increasingly become the wrapper around a more valuable asset: access to a target group.
Distribution can be built or bought.
You can manufacture access through features and extensions that are non-monetized: free tools that multipliers adopt because they cost nothing, which builds reach without revenue and keeps the valuation closer to CAC than ARR.
Or you can buy software companies based on the value of their ICP and distribution.
At current software valuations, buying distribution may simply be cheaper than building it.
The most interesting targets are multipliers: firms that serve firms.
Accountants. MSPs. Property managers. Other businesses that sit in the workflows of many downstream companies.
One multiplier gives you access to more than one customer relationship. It gives you telemetry into a broader customer base.
That is where this becomes more than a software consolidation strategy.
Distribution creates telemetry.
Telemetry creates a record.
The record allows you to make predictions.
And better predictions create alpha.
What This Could Become
So what would a fund look like that buys distribution?
The fund would acquire majority stakes in software companies based less on the software itself and more on the target groups they reach.
Operationally, the goal would be to optimize those companies for market share.
Not necessarily revenue maximization but market penetration. The more specific the verticle the better.
The asset being accumulated is the seat inside the target group and the behavioral record that comes with it.
Over time, the operating software company and the record could become two separate assets.
The customer base and engagement can be sold to an existing consolidator.
The consented, aggregated telemetry layer stays behind.
That distinction matters because a software company can be sold, but longitudinal record compounds.
Which is why I don’t think the record should sit inside a ten-year fund. It should sit inside a permanent holding company.
The fund does the deals. The software assets can roll off. The record stays.
The Prediction Economy
My working assumption is that after Software as a Service, a new value creation layer emerges: the prediction economy.
Software itself becomes cheaper to build and easier to replace.
Distribution does not.
And proprietary ground truth does not.
The real moat may therefore not be the application layer. It may be owning access to a defined market and the running record of how that market behaves.
That record can improve investment decisions, pricing, underwriting, product decisions and forecasting.
The longer the record runs, the more valuable it potentially becomes.
Ground Truths
A) Distribution is the last moat in software.
B) After Software as a Service, a new value creation layer will emerge, which will be the prediction economy.
C) The worth of a target group is defined by market share and purity of ICP.
D) Multipliers are particularly valuable because they provide telemetry into a broader downstream customer base.
E) The software is not necessarily the permanent asset. The record is.
Execution
A) The fund takes majority stakes in software companies.
B) The fund is operationally involved in those companies, optimizing them for market share inside a defined ICP.
C) Companies are selected based on the quality of their distribution, particularly their reach into multipliers.
D) Once enough distribution and history have been built, the operating customer asset can be sold to consolidators.
E) The consolidator gets the customers and their engagement. The holdco retains the consented, aggregated telemetry feed.
F) The consolidator can have the option to roll into non-preferred stock of the fund.
G) The record sits in a permanent holding company, not in the ten-year fund. The fund does the deals and rolls the data asset into the holdco.
The open question is whether proprietary behavioral telemetry across enough companies actually creates durable predictive alpha.
I think it is worth finding out.
If you’re running a software consolidator, investing in one, or think I’ve got this wrong, write to me at viktoria@izdebska.com.