Model: 1
Wednesday, 27 September 2017

Margins.

Is every process profitable?

Software companies have famously high margins, because they have no cost of goods sold. We have humans doing work, so our margins will be lower than pure software companies: it costs us money to do work for our clients.

In the short term, the question is: is every process profitable? Even without technology, if every process is profitable, then that guarantees our survival, buying us time to build technologies which increase our margins over time.

In the long term, the question is: can we achieve margins worthy of a technology company?


Models

To answer these questions, I’ve built models for four scenarios.

1. Base. No efficiency gains. 
2. Conservative. Small efficiency gains.
3. Moderate. Solid efficiency gains.
4. Extreme. Huge efficiency gains.

None of these models include fixed costs or sales & marketing costs. Our fixed costs are small: payroll is ~$17K per month at present. As for our sales & marketing costs, we’re still figuring that out, and so we’ll leave that for a future modeling exercise.


Tabs

Here’s a quick explanation of each tab in the models above.

Revenue

Price per process x number of processes. 
(For our pricing tiers, see our website.)

Execution Cost

This is the raw labor cost of paying operators to execute a process.

Coordination Costs

These are the other variable costs involved in executing each process.

There are supporting roles that other agents play. 
There is a cost to making process upgrades.
And there are various software costs.

Gross Profits

Revenue minus execution and other variable costs equal gross profits. 
Margins decrease at every pricing tier and increase with every efficiency.


Execution Efficiencies

Let’s compare execution cost efficiencies across models.

Base

The Base model assumes a simple economic efficiency. The more processes we do, the more we can take advantage of unskilled labor. Skilled labor is reserved for executing the initial instances of each process, but then unskilled labor can be trained to replace them.

So there are seven (7) labor tiers in the Base model:

PRC 1–4 =$5/hour.
PRC 5–10 =$4.5/hour.
PRC 11–25 =$4/hour.
PRC 25–50 =$3.5/hour.
PRC 51–100 =$3/hour.
PRC 101–200 =$2.5/hour.
PRC 201+ =$2/hour.

Conservative

The Conservative model maintains the 7 labor tiers from the Base model, 
but it introduces efficiency gains.

Although it introduces efficiency gains, it places a -25% buffer on them. Because of this, there are efficiency losses until the 11th process, where it plateaus, and then begins improving.

Moderate

The Moderate model maintains the 7 labor tiers from the Base model, as well as the efficiency gains from the Conservative model. The only difference is the removal of the -25% buffer.

Extreme

The Extreme model reduces the cost of labor per hour by 25¢ at every tier — the lowest tier, at $1.75, is still more expensive than our actual base operator pay, which is $1.50.

The Extreme model also increases the velocity of efficiency gains, growing it to 15% per pricing tier to cap at 90% efficiency.


Coordination Efficiences

Let’s compare coordination cost efficiencies across models.

Base

The Base model puts a -25% buffer on coordination costs.

Here are the efficiency assumptions:

PRC 1–4 =-100% efficiency.
PRC 5–10 = -50% efficiency.
PRC 11–25 =0% efficiency.
PRC 25–50 =+25% efficiency.
PRC 51–100 =+50% efficiency.
PRC 101–200 =65% efficiency.
PRC 201+ =75% efficiency.

Conservative

No difference from the Base model.

Moderate

The Moderate model removes the 25% buffer. It also increases efficiency assumptions, but caps out at the same rate of efficiency.

PRC 1–4 =0% efficiency.
PRC 5–10 =15% efficiency.
PRC 11–25 =30% efficiency.
PRC 25–50 =45% efficiency.
PRC 51–100 =60% efficiency.
PRC 101–200 =70% efficiency.
PRC 201+ =75% efficiency.

Extreme Model
The Extreme model lowers the base costs and increases efficiency assumptions even further, growing 15% per tier, until capping out at 90%.

PRC 1–4 =0% efficiency.
PRC 5–10 =15% efficiency.
PRC 11–25 =30% efficiency.
PRC 25–50 =45% efficiency.
PRC 51–100 =60% efficiency.
PRC 101–200 =75% efficiency.
PRC 201+ 90% efficiency.

Results

Margins on the first process…

Base. 7.5% margins.
Conservative. 1.25% margins.
Moderate. 33% margins.
Extreme. ~40% margins.

On the 400th process…

Base. 28.65% margins. 0% efficiency. 
Conservative. 41.15% margins. 50% efficiency. 
Moderate. 50.83% margins. 60% efficiency. 
Extreme. 67.95% margins. 90% efficiency.

In all models…

Every process is profitable.
Every hour of operation is profitable.
Price per process decreases as quantity increases.
Margins increase over time.

Conclusion

Because of Amara’s Law, these scenarios model increasing margins over time. Our margins today are in the Conservative bracket — already better than the Base model. Over the next three to five years, we’re confident we can achieve the efficiency gains in the Moderate model. In the five to ten years following this, we hope to achieve the gains in the Extreme model.

To return to the two questions from the beginning: we see that every process profitable, and that margins improve steadily over time — until they asymptotically approach software margins.

To an investor, the basic value proposition of synthetic intelligence — that humans are involved in execution — would seem to imply low margins. But we see that at scale, a synthetic business model can achieve astonishing margins.