Arron Bennett, Founder of Bennett Financials
That's my latest video, on the three ways fast growth breaks a team. The first is hiring too late. This post turns that idea into a small script you can run against your own numbers.
The idea is simple. Most teams hire when everyone is already full. But a new person takes months to get good. So by the time they're productive, the team has been overloaded for a while and customers have felt it.
The fix is to start hiring when the team will be about three-quarters full, looking ahead by however long your last hire took to ramp up. I'm bad at math in my head. That's why I use spreadsheets. Or in this case, pandas.
What does the script need?
Four inputs and a history:
- Monthly workload history. Clients, tickets, projects, whatever one person carries.
- Headcount of fully trained people.
- Capacity per person. How much one trained person handles well, from your own best months.
- Trigger. I use 75%.
- Ramp time. How many months your last hire took to work without close checking.
In the video the example is a person who manages about 10 clients. At seven, you start hiring. Below, let's say it's a payroll services bureau with nine specialists and $2.2 million of revenue, where one trained specialist runs about 45 client payrolls well. It's hypothetical. The numbers are made up to show the method.
The script
import pandas as pd
# Hypothetical payroll bureau: monthly active client count, last 12 months
history = pd.Series(
[228, 233, 236, 242, 247, 251, 258, 262, 268, 273, 279, 284],
index=pd.period_range("2025-11", periods=12, freq="M"),
name="clients",
)
SPECIALISTS = 9 # fully trained staff today
CAPACITY = 45 # clients one trained specialist handles well
TRIGGER = 0.75 # start hiring at 75% of capacity
RAMP_MONTHS = 3 # how long your last hire took to get good
# Average monthly growth over the last six months
growth = history.diff().tail(6).mean()
# Project the next 12 months on that trend
future = pd.period_range(history.index[-1] + 1, periods=12, freq="M")
forecast = pd.Series(
[history.iloc[-1] + growth * (i + 1) for i in range(12)],
index=future, name="clients",
).round()
load = forecast / (SPECIALISTS * CAPACITY)
crossing = load[load >= TRIGGER]
print(f"Growth: {growth:.1f} clients a month")
print(f"Load today: {history.iloc[-1] / (SPECIALISTS * CAPACITY):.0%}")
if crossing.empty:
print("No hire needed in the next 12 months")
else:
month = crossing.index[0]
print(f"Team reaches {TRIGGER:.0%} in {month} ({load[month]:.0%})")
print(f"Start the hire by {month - RAMP_MONTHS}")
Output:
Growth: 5.5 clients a month
Load today: 70%
Team reaches 75% in 2027-02 (76%)
Start the hire by 2026-11
So the bureau should open the role this month. I know the reply. "Arron, we're busy, but we're not that busy." Right. The team isn't full. It's at 70%. But it crosses 75% in February, and a new specialist takes three months to get good.
Why 75% and not 100%?
Because the last quarter of capacity is where quality goes. At full load there's no slack for a sick day, a big new client or someone training a new hire. Mistakes go up, turnaround slips and customers notice. Hire so the new person is ready at 75%, and the team stays out of that zone.
Change TRIGGER to 0.85 and run it again. The answer moves to "Start the hire by 2027-07". That's eight months later. And it means the new person shows up when the team is already stretched. That gap is what hiring too late looks like in numbers.
What are the limits of this script?
It's a straight-line trend. It doesn't know about seasons, a big client landing or a client leaving. So treat the output as a prompt to look at the pipeline, not a decision.
Easy improvements:
-
Use the pipeline instead of the trend. Replace
growthwith signed and likely work by month. - Add seasonality. For seasonal businesses, compare each month with the same month last year.
- Check cash. A new hire costs full pay from day one and produces less during the ramp. Add a line for that cost so you know the cash you need.
- Run it monthly. It takes seconds once the history is in a CSV.
Hiring ahead only works if the basics hold
The video makes a point I'd repeat here. Hiring early only makes sense with reliable leads coming in and a steady close rate. Without them, you can hire for work that never shows up and squeeze your cash. So pair this script with a look at where new work comes from.
And hire one at a time where you can. If the script says you need two, stagger them. Two people trained properly beat four trained badly.
How do you find your capacity number?
Look at months where work got done on time without overtime, and complaints were low. Divide workload by headcount for those months. That's your real capacity. Don't use your busiest month. That's what overloaded looks like.
If your history is in a CSV with one row per month, you can find those months with pandas too. Filter to months where overtime was below some threshold and on-time delivery was above another. Then take the median workload per person. The median matters here. One really quiet month shouldn't set your capacity. Neither should one heroic one.
Capacity isn't fixed either. Better tools, better training and cleaner processes raise it. Check it again every six months, and update CAPACITY when the evidence says your team can carry more without quality slipping.
How do you measure ramp time?
Ramp time is the number most teams guess. Measure it instead. For each of your last few hires, note the month they started and the month they first carried a normal workload without someone checking everything. The gap is the ramp. Use the longest recent one, not the average, because the whole point of the script is not being late.
If you've only hired once, that one number still beats a guess. If you've never hired, start with three to six months. That's the range I see most often in service work. Then correct it after your first hire.
What does the cash side look like?
A new hire costs full pay from the first month but produces less until the ramp is over. Add a small block to the script that multiplies monthly cost by the ramp months and by the share of output still missing. Print it next to the hire date. In the bureau example, a specialist costing about $5,500 a month all in, at roughly half output over a three-month ramp, means about eight grand of cost not yet matched by work. Knowing that before you post the job is what lets you hire early without squeezing cash.
Why run this every month?
Because the answer moves. A big client signing, a client leaving or a change in the growth rate all shift the date. Run the script on the first working day of each month and write down the answer. Over a year, that gives you a simple record of when you knew a hire was coming. It's also a good way to check if your trend assumptions are any good. Garbage in, garbage out.
For more on this, I've written about whether a service business can afford its next hire, and our labor efficiency research has the benchmarks behind the capacity numbers.
Top comments (0)