On Clairvoyance

2025.11.23

CXXVII

[Hard Work Pays Off…; ...if We Define Hard Work & Pay Off; Beating the Market; A Less Crowded Crowd; Math is Clairvoyance]

Thesis: Useful models help us predict the future.

[Hard Work Pays Off...]

We can use models of the world to predict the future. For this reason, a very accurate model of the world is incredibly valuable... who wouldn't want to predict the future?!

The model doesn’t have to be an exceedingly complex or obviously “mathematical” one, either. We can view heuristics as predictive models, too.*

Let's take a very simple example:

Hard work pays off.

This, in a way, predicts the future: if you work hard now, you will get some sort of pay off or reward at a later time.

We can reword it as:

If HardWork, then PayOff.

In this context, both HardWork and PayOff are variables that can either be "True" or "False".

The above model is shorthand for…

If HardWork is True, the PayOff is True.

When you evaluate the statement, you replace HardWork with either True or False. So, if HardWork==True, it reads like:

If HardWork==True, then PayOff==True

So, if we make HardWork True, then we definitely get a PayOff!**

As you can see, our little model makes an attempt at predicting the future under the conditions of HardWork being true.

_{*This is something I’ve viewed to be true since I was a child—the wisdom of the ancients is a distilled predictive model for helping us understand cause and effect very efficiently}

_{**Alternatively, if HardWork==False, we don't actually know anything about PayOff (it may or may not be true).}

[...if We Define Hard Work & Pay Off]

The above HardWork model is cool because in a lot of situations, it's right. HardWork does lead to PayOff.

But, it's not always true. It really depends on what we mean when we say HardWork or PayOff.

If you think HardWork is gardening for an hour in the sun one time and PayOff is being a billionaire, then the model totally fails to predict the future! And, obviously, models that don’t actually predict the anything aren’t as useful as the ones that do.

Now, if we replace HardWork & Payoff with more narrowly defined terms, we'll reduce the number of situations the model claims to work in, but we'll also increase it's accuracy.

Maybe we want to say that if you take 1000 hours of sales calls (HardWork), then your win rate will double (PayOff). Now, this is very objective:

If (SalesHours > 1000), then WinRate = WinRate * 2

Whether or not this is true is debatable. But, the point is, now we have a lot more clear definition to iterate on and tweak.

The trade off is, now our model doesn't apply to gardening or coding or swimming or writing or flying or a lot of other things. In this case, we’ve traded generality for accuracy.

Sometimes, that's okay. If your goal is to predict the win rate of your sales reps, gardening and being a billionaire shouldn’t matter so much.

We can go a step further and keep making the model more and more specific:

If (SalesHours > 1000) and (WinRate < .10), then WinRate = WinRate * 2

We've even more narrowly scoped the model & it's prediction to only apply to sales people with a WinRate currently under 10%.

This cuts off a lot of sales people who have a win rate above 10%, but it is a simple tweak that makes the model “correct” a much higher percentage of the time. This is because generally, doubling a win rate from, say, 15% to 30% is a lot harder than doubling it from 5% to 10%. And, doubling from 30% to 60% is in the realm of “basically impossible” in a lot of people’s minds.

We can keep increasing the predictive power of our model by adding other conditions and variables, like level of focus during the sessions, past experience, contract value, and sales cycle length, etc.

And, it doesn’t have to be so binary, either. We could make a function that, given any combination of those numbers or features ,outputs an expected increase in win rate. Such a model would be more useful, because it could tell us about a lot of different cases.

[Beating the Market]

Investors use super complex models all the time to predict the value of a company.

There are big spreadsheets with many rows that take in information at the top, and then go through a bunch of equations and assumptions, and get a number out at the bottom.

A simple version might be:

Value = (Revenue - VariableCost - FixedCosts) * 5

The really cool thing about these models is even if they are imperfect (they always are), they can help you understand the impact of different variables on the number you care about (value).

As an example, if the price of copper goes up, you can use the model to predict how much that will effect the variable costs of a computer company that uses copper, and likewise the final value.

In this way, you can make a prediction like:

If the copper price increases by 10%, the Value of Computer Co will decrease by $1000.

So, if your model is somewhat accurate, and you can gather data on it's inputs, than you can predict what will happen to the value & can possibly make money.

A word of warning, though particularly in finance:

Models are wrought with assumptions and can easily be inaccurate. Tread with caution, and use such things as guides, not gospel.

[A Less Crowded Crowd]

A lot of people do financial modeling. I think this is maybe because there are so many numbers that are widely available and lend themselves to a model that can be evaluated overtime against what actually happens.

For that reason, there are a lot of really strong financial models, and it is a competitive space.

When a space is more competitive, it is harder to get outsized returns.

That’s part of the reason that working with sales teams via BirdDog is so interesting to Jack & I; we have all of the ingredients for a pretty compelling model:

• A valuable set of target variables we want to predict and optimize {win rate, contract value, sales cycle length}

• A data set that has a) an actual causal impact on said target variables b) can help make decisions that positively impact the target variables

When you have these two things, then it becomes a question of connecting one to the other with your model. If you can predict the targets based on the data set, and can make different decisions based on the data set, then you can improve the target variables.

As I hinted at last week, win rate, contract value, and sales cycle length all lead directly into revenue. CEO’s and boards care about revenue, but they might not care about the open rate of your emails.

The really exciting part about all this is that no one else has the data sets that we have.

Sure, the data sets can be copied, but it’s not as easy as going out and buying them. After all, we did not spend the last 15 months building a commoditized data set. We spent the last 15 months building an engine that builds a bespoke data set on demand.*

And now, over the last month or two, we’ve just started making some relatively trivial models to evaluate the causal relationship between these unique data sets and the variables we’re optimizing for.

It’s early, but it’s pretty clear to us that there is a strong causal relationship between our data and the target variables… Already, the insights have been impactful enough to make it up to the CEO at one our clients with high 8 figures in yearly revenues & be part of the C Level conversation for their 2026 planning.

We’re just getting started, but are very optimistic that we can articulate and scale this predictive power across our client base & use it to open up doors with new ones, too.

_{*This is, in itself, a model that generates models, if you will}

I’m here every Sunday, please subscribe if you enjoyed the post!

[Math is Clairvoyance]

I’m obsessed with accurately modeling the world and am really grateful that my current project gives me a chance to do that.

I’m super excited about it, because we’re really just scratching the surface.

To tie it together, the commonality between all of the above models, other than their ability to predict the future, is that you can express them with math.

That's a good reminder that math is a powerful way to view the world, and it's a pretty beautiful one, too.

Live Deeply,