Use case:

Workload Distribution

Here’s how we replaced a $1500/mo project manager with a few high school math equations (that were also 30% more accurate than the human)…

Our clients with larger teams (of writers, editors, salespeople etc) often have a project manager on staff to decide who should get the next project.

Whether this process is documented or not, it’s often done as a factor of each team member’s…

  • Current workload and overall output over the last several weeks (“Macro” availability)
  • Experience level (the most complex tasks go to the most experienced team members, but if there is a basic task it generally goes to someone new to get them trained quickly)
  • Planned holiday/sick
  • “Micro” availability (did they JUST get a project?)
  • Topic/project type specialties (ie: anyone COULD do this project, but Dave just finished a very similar one…)

And so on…

We first encountered this problem with a content agency, and after working together for several years, they confided in us that this project distribution process was a pain point.

Challenges with Workload Distribution

We had helped this agency automate lead flow, onboarding, and team communications (to update project status), and this had all helped the agency scale to a team of 30-50 writers. But, now the “simple” task of deciding who would get the next project was starting to become a challenge at their new scale:

  • Writers were paid by the project, so would often complain to the project management team if they weren’t getting as much work as they wanted
  • Project managers would assign a project to a writer, only to have them turn it down due to availability (which meant further delaying the start of the project)
  • And, the team was growing quickly, and sometimes new writers were left wondering when they’d get their first project to learn the ropes

When the founder came to us with this problem, he thought it was an operations need (NOT an automation one). He assumed this process couldn’t be automated at all, and instead, he (and his project management team) assumed they just needed better training, checklists, and so on.

They had good reason to believe so too — the project managers were considering seemingly-complex factors like:

  • Writer specialities for certain project types
  • Always-changing writer capacity (ie: a writer who’s great one week may get overwhelmed or demotivated the next, may get better over time, and all of this would influence how much work could be reliably assigned to any given writer)
  • The inaccuracy of some project types (ie: their clients would often mislabel project types)

We recognized this process not as a behavioral problem, but a mathematical one.

We knew it would be possible to break down the emotional judgement call the project managers were making, and express it as a math equation – we’d estimate the availability of each writer, rank their quality score, rank their timeliness, and determine the topic of each project. In so doing, we could determine the list of writers best for any given project.

The founder was skeptical (but decided to trust us), despite his project managers insistence that this process could never be done accurately with an automation.

How We Automate Workload Distribution

First, consider the emotional decisions/judgement calls in your business – “who should get the next project” “which project should I quality check first” “which of these thousand candidates should I interview” etc.

All of these judgement calls can simply be expressed as a combination of logical factors.

For example

Available writers =

  • Writers where their current workload + current project workload < estimated capacity (macro capacity)
  • AND writers that are not sick/on planned holiday (availability)
  • AND writers that have not denied a project recently (ie: micro-capacity)
  • AND based on the topic, writers that could cover this topic

Best writers =

  • Available writers AND writers ranked by the following factors:
  • Timeliness (how on-time they are as a factor of project size and how long they had to do the project and how late the project was submitted, for the average project over the last few weeks)
  • Receptivity (how quickly they confirm they can take on a project)
  • Quality (how few mistakes they make and severity of these mistakes for an average project they submit)

Emotional = Logical + Logical + Logical…

So, it appears possible to represent an emotional judgement call with a series of math equations…but why would we bother to do this?

If a human does this decision-making process, it introduces several problems:

  • Inconsistency: Each team member uses different decision-making criteria (ie: one person thinks submitting projects late is a huge problem, but another prioritizes responding quickly to start a project):
  • Cost/time and mental energy: it takes your time to train new team members on your preferences, and time for them to think through the right decision, again and again. All this leaves less time for you and your team to think strategically, solve problems that only humans can solve, and so on.
  • Inaccuracy: Humans simply can’t consider as much information as a computer can, which leads to forgetting information/mistakes etc.

So, we know the high-level strategy and the motivation for replacing a human judgement call with math…here’s how we do this:

  1. Decide on an output variable (ie: probability a writer can take on a project now)
  2. Define the variables to be used (ie: total time elapsed since a writer denied a project)
  3. Define the “shape” of the decision (ie: if a writer JUST denied a project, odds are they are not available, but after a day or so, the probability should go back to normal)
  4. Find a math equation to represent the output given the inputs (ie:

Then, repeat 1-4 for each factor of the decision-making process, and validate as it’s used.

We have since found that a similar strategy can be used for any business decision making. Here’s how we implement this for ANY process:

  1. Define the decision-making output (best writer, best editor, best salesperson, next project to check, and so on)
  2. Interview the decision-makers (project managers etc) and break the judgement call into logical parts
  3. Define a math equation for each part
  4. Design an Airtable base that allows tracking of each input – ie: “timeliness” requires the deadline for each project, actual submission time, when the project was assigned and the total workload (since an easy project with plenty of time given should be penalized much more than a difficult project assigned last minute), and the team member assigned to each project)
  5. Ensure each input is tracked automatically
  6. Program the math equations into Airtable, Google Sheets etc, linked to real data
  7. Show the ranking produced to the project management and leadership team – see if it makes sense to everyone (the best way to validate a decision-making algorithm is to show the output score for ALL team members)
  8. Change the equation if EVERYONE agrees that the factors are incorrect (or, what often happens, is address the disagreement in priorities and get buy-in on one common way to perform this decision)
  9. Tweak weighting as the system is used and track results over time (ie: is the math equation more accurate than the humans? If not, adjust as necessary to account for the factors not included in the equation).

Similar Use Cases:

Want Something Similar In Your Business?

Book a free operations consultation below