Question

My Per Developer Timelines stretch across a wide range of dates. How can I steepen this slope?

Answer 1

There's an elephant in the room.

It's very likely that you have one or more cases with a large estimate (over 40 hours).

When we predict your ship dates, we run a bunch of simulations. An elephantine estimate injects a huge amount of variability into this process, because one thing matters more than everything else:

Which "velocity" was applied to that big estimate for this simulation

Here's a simplified example:

---

Let's say you have two cases with estimates:

• Update Logo - 1 hr
• Rewrite code from scratch - 40 hrs

Let's say you have these two estimates in your history:

• estimated: 1hr, actual: 30m (You finished in half the time.)
• estimated: 1hr, actual: 2hr (It took twice as long as thought.)

These two estimates translate into two velocities:

• 0.5
• 2

When we apply the first velocity to a case in EBS, we expect the case to take half as long as its estimate. When we pick the second, we expect it to take twice as long.

---

Simulation 1

• Update Logo draws 2, so: 1hr * 2 = 2hrs
• Rewrite code from scratch draws 0.5, so: 40hrs * 0.5 = 20hrs

  Total: 22 hours

---

Simulation 2

• Update Logo draws 0.5, so: 1hr * 0.5 = 0.5hrs
• Rewrite code from scratch draws 2, so: 40hrs * 2 = 80hrs

  Total: 80.5 hours

---

So it's either going to take half a week or two full weeks.

Now, let's break up that big case into five smaller cases, each representing a day of work.

• Update Logo - 1 hr
• Investigate rewriting from scratch - 8 hrs
• Convince mgmt it's a bad idea - 8 hrs
• Find bugs - 8hrs
• Fix them - 8hrs
• Hit the beach - 8hrs

Same number of total estimated hours. Different breakdown.

---

Simulation 1

• 1hr * 0.5 = 0.5hrs
• 8hrs * 2 = 16hrs
• 8hrs * 0.5 = 4hrs
• 8hrs * 2 = 16hrs
• 8hrs * 0.5 = 4hrs
• 8hrs * 2 = 16hrs

  Total: 56.6 hours

---

Simulation 2

• 1hr * 2 = 2hrs
• 8hrs * 0.5 = 4hrs
• 8hrs * 2 = 16hrs
• 8hrs * 0.5 = 4hrs
• 8hrs * 2 = 16hrs
• 8hrs * 0.5 = 4hrs

  Total: 46 hours

---

This is a very simplified example, but as you can see, the spread between the high and low estimates is greatly reduced by breaking up cases into smaller increments.