Start With Your Vision — Not With Goals. Measure What Really Matters.
Goals Without Vision Are Worse Than Useless. They can lead you in an expensive wrong direction.
Welcome to the Scarlet Ink newsletter. I'm Dave Anderson, an ex-Amazon Tech Director and GM. Each week I write a newsletter article on tech industry careers, and specific leadership advice.
Free members can read some amount of each article, while paid members can read the full article. For some, part of the article is plenty! But if you'd like to read more, I'd love you to consider becoming a paid member! As a side note, I wrote a version of this article years ago. Decided to re-write it because I like the topic, and I want to keep all my archives high quality!
Let’s say you take a shovel out to your backyard. You pick a spot of the way, and start digging a hole. Oh man, you’re working hard. You spend a few hours in the sun, shoveling hard.
Did you accomplish something?
Well, that depends on if you had a goal to have a hole in the ground.
If you wanted a hole in the ground, you’ve accomplished your goal. If you didn’t want a hole in the ground, this was a waste of time.
How do you decide if you want a hole in the ground? Well, in real life, that might be straightforward. You want a pond in your backyard, and you’d like to build it yourself. You investigate pond building, and decide that a hole is the next step in the pond building process. Hole digging time.
That’s easy. But I’ve repeatedly seen people at Amazon mess up their goal setting process. Let’s first talk about how things go wrong, and then what you can do to improve matters.
For a few years of my Amazon career, I worked in the Amazon Marketplace group. This is a platform which allows 3rd party sellers to sell their products through Amazon's retail website. Through this platform, sellers can set prices, inventory levels, discounts, and more.
This story was during my time there.
The Situation.
3rd Party Sellers run entire businesses through Amazon. As I said, they have to set prices, inventory levels, discounts, messages to buyers, etc.
There was an action our 3rd parties wanted to perform regularly. For the sake of keeping things simple, let’s say that it was setting prices. Each seller has a private queue for price updates, and each seller was throttled to a specific speed based on their business size. So a large seller might be allowed 1000 price changes a minute, and a smaller seller was allowed 100 price changes a minute.
The issue the pricing team was dealing with was that there was a long list of sellers who were running pricing changes at their maximum rate of price changes. They’d submit price changes faster than the system could process them, and they would fill up their queue of pending price changes until the system started rejecting future price changes.
For example, while a large seller was allowed to send 1000 price changes a minute, they were submitting 1500 price changes a minute, and the remainder of price changes were being tossed away by Amazon's system.
That sounded like a bad problem. The team in charge of price changes was concerned. They said that sellers needed to change prices quickly, and our price change system was limiting them.
The Solution.
The price change team set a goal to improve the speed of processing price changes for each seller.
They figured if they could process price changes more quickly, they could increase the queue size for each seller, and they could handle the incoming rate of changes.
They carefully measured the pace that sellers were submitting price changes (the failed rate). They calculated that doubling the processing speed (and queue sizes) across all of Amazon’s systems would remove the constraint on sellers.
In other words, if they could increase large sellers to 2000 price changes a minute (and small sellers to 200 price changes a minute), the problem would be resolved.
At first glance, this made sense. Da da dum. Foreshadowing!
The team properly set a SMART goal, outlining exactly how much processing speed would need to be added across Amazon’s system, and the increase in queue processing speed. They explained how this would resolve the issue that sellers were encountering.
Months went by. Work was completed across many systems. Finally, the team declared success. They had massively improved the architecture of the price change system across Amazon. They’d threaded improved processing speed from seller all the way to the retail website. Not by the 2x planned, but by 3x! A huge success.
There were parties, and literal champagne was poured. The team sent out an excited note to all their partner teams across Amazon that they were going to roll out triple limits for all sellers to remove their price bottlenecks.
Except, as the new limits rolled out, the bottlenecks didn’t change at all. This was initially met with confused frowns, and then panic started to set in.
Almost every single bottlenecked seller stayed bottlenecked, even when their limit was increased to 3x their previous limits. Emergency investigation tickets were opened. We’d just spent hundreds of engineer hours changing systems across Amazon, and something went sideways.