عربي

The missing half of MENA's venture capital data

The missing half of MENA's venture capital data

An article by Nour ElShaeri, startup and VC journalist and CEO of Investment Analyst AI (IA)

A merchant of the last century would have thought our bookkeeping peculiar.

He kept a daybook, a ledger and a correspondence file, and he kept them in double entry, because he had learned at some expense that a house which records only what it buys has written down half of its own history.

The cargo he declined at Alexandria, the consignment he judged overpriced at Marseille, and the young trader he thought unserious, who was shipping on his own account within three years, were all entered, dated, and kept. Not from sentiment. This is based on the plain commercial understanding that a house improves only by consulting the record of its own judgement.

The venture houses of this region have become rather adept at recording what they buy. What they decline is another matter.

A ledger kept on one side

We now have admirable data on what gets funded. Monthly and quarterly funding reports have made the regional record far better than it was five years ago. A founder in Riyadh or Cairo can see the size of rounds, the active funds and the sectors drawing capital.

Every figure in that record, however, sits on one side of the ledger. It is a register of subscriptions. It is not a register of judgement.

Consider what a fund may not systematically keep.

The companies it declined, who funded them afterward and at what price. An honest comparison between the thesis it states to its limited partners and the thesis its checks actually describe. The founder who came through the door two years ago under a different company and a different name. The pattern of where its own partners disagree, which is often the most instructive paper in the house.

Each of these is a fact about the fund itself, not about the market. No external provider can furnish it in full.

And where these records are not kept or revisited, the most expensive lesson a firm purchases each year, the lesson of its own refusals, is bought and then discarded.

I have spent the past several months speaking with more than 15 senior investors across Saudi Arabia, the UAE and Egypt in the course of my work for Arab News. In those conversations, I have yet to encounter a partner who could readily answer a simple question: What did we pass on in the last 24 months that was raised elsewhere, and what does that tell us about how we decide?

The question is not an accusation. It points to a weakness in how investment decisions are remembered, examined and, eventually, improved.

The thesis nobody wrote down

There is a reason the book is so often incomplete, and it is more fundamental than neglect. The criteria themselves are not always written down in a form that can be consistently applied.

Ask a partner what his firm invests in, and you will receive a fluent and confident answer, delivered rather differently by each partner you ask. The thesis lives in conversation and in the body. It is instinct, and it is held, correctly, to be the partner's finest instrument.

The difficulty is that instinct is being spent at the wrong station.

It is a superb instrument for conviction: for the reading of a founder across a table, for the decision to commit capital against an uncertain future, for the moment when the evidence runs out and somebody must nonetheless choose.

It is a poor instrument for the first layer of screening, where the work is not judgement alone but consistent application. Several hundred proposals, read to one standard, in one order, with the same questions put to each.

Instinct does not scale to 300. It does not survive a long afternoon, and it cannot easily be audited afterwards, having left no record of itself.

This matters to founders as much as to investors. A company may return to the same fund with a stronger business, a different proposition and evidence that addresses its original weaknesses, only to be assessed without reference to the previous conversation. A fund that cannot remember why it declined a company is poorly equipped to recognise when that decision deserves reconsideration.

The first layer of screening therefore matters more than its status in the firm suggests. It is the station at which a house's discipline is either applied to everything or quietly abandoned. Everything downstream inherits what that layer let through.

A firm that leaves it entirely to instinct has not economised on process. It has declined to have one.

The excuse that has expired

For much of this industry's life, there was a reasonable excuse. The labour was prohibitive.

A clerk able to read several hundred proposals a quarter, record the grounds of each refusal against the house's stated criteria, and recall two years later that a founder had been seen before under another name would have cost rather more than the error he prevented.

That excuse has expired.

The arrival of competent AI screening and analysis tools for venture capital has made much of the clerk's labour cheaper for the first time. What was principally a staffing question has become a question of institutional will.

The capability worth having, however, is more particular than the phrase "AI for VCs" suggests.

The first requirement is that the criteria belong to the house. A screening layer built on somebody else's definition of a satisfactory company is a borrowed opinion. The configuration must be the fund's own thesis, set down explicitly and therefore auditable against the fund's actual behaviour.

The second is that evaluation proceeds in layers rather than in a single verdict so that a proposal is sorted rather than sentenced, and a partner can see at which layer and on what grounds a company fell.

The third is depth on demand: the ability to take any company that merits attention and produce a substantiated analysis with its sources attached so that the reasoning survives the meeting and can be reread when the market proves it wrong.

The fourth, and the one this industry has barely begun to want, is memory.

A system that knows what the house has already seen, already refused and already reasoned about is our merchant's ledger restored, in a form that scales to the volume of deal flow a modern fund receives.

None of this requires a firm to surrender its judgement to a machine. Nor does institutional memory begin with AI. A disciplined record, faithfully maintained and periodically consulted, is already an improvement. Technology merely changes the economics of maintaining that record at scale.

Here I should be plain about my own position. Alongside my work as a journalist, I run IA, Investment Analyst AI, an AI screening and evaluation platform built for VC funds and accelerators in this region. More than 1,000 pitches have passed through it, and the company participates in NVIDIA's Inception program.

I have an interest in this argument. I would make it in any case, and I would rather make it openly than have it inferred.

Two cautions

Machines sort; partners decide.

The merit of a screening layer is that it applies the house's own criteria consistently to everything, including the proposals that arrive at the wrong hour through the wrong introduction.

It is not a substitute for conviction, and a firm that treats it as one will succeed only in automating its own mediocrity.

And a record is worth keeping only if it is read. A ledger that nobody consults is just a drawer.

The houses that compound an advantage here will not necessarily be the ones that adopt the newest analysis tools first. They will be the ones that decide their judgement is an asset worth recording.

Our merchant understood as much with a pen and a bound book. We have considerably better instruments and, thus far, a worse record.

Thank you

Please check your email to confirm your subscription.