Pruning Your Decision Trees

Smart people like to solve hard problems. But success often goes to the disciplined person with the right temperament for the circumstance. I’m often guilty of making things overly complicated. I get fascinated with the beauty of complex adaptive systems and the challenge available in trying to tame it to my liking. However entertaining this might be, it’s rarely profitable. The situation almost always bears more fruit after wise pruning to fewer key decisions that are accurate, different and non-obvious to the the crowd. This creates sustainable edge that is difficult to copy because it’s adaptive to the changing circumstances. There is little energy or cost associated with analysis and maintenance.

I’ll walk you through a decision tree I developed for soil sampling many years ago. I have others today… and more that need pruned. The results of this simple decision tree were exponentially more work per person, per year compared to our competitors who often assigned an individual technician to a single region. It was a simple insight I had and became strangely simple to manage… which afforded me the ability to work on other things while earning exceptional profits. My only problem is it’s scale was limited to geography. I solved that limitation a different way. (more on that another time.)

The following is from memory. I had a one page graphic in our employee handbook/manuals that crew leaders and technicians could reference. It was front and center (and only slightly more nice to look at.)

  1. Is it raining?
    1. Yes = Move regions
    2. No = Check next decision
  2. Is current region complete?
    1. Yes = Move regions
    2. No = Check next decision
  3. Have you been working this region for more than 3.5 days?
    1. Yes = Move regions
    2. No = Keep working this region

The three rules were “self-healing.” They were fractal. They allowed the business to grow and expand like a forest soaking up sunshine and moisture. It was simple and beautiful to implement. No debates. Just action, properly ordered. How? Why? Let me break it down in detail and try to explain why and how the process might be applied in your own business.

Decision #1 (Is it raining? Then, move): Soil sampling is an outdoor activity. You drive an all-terrain vehicle across a farm field with global position satellites guiding your way to a specific location. A probe goes in the soil and the soil goes into a labeled bag. All of this activity becomes terribly ineffective if it’s pouring rain. Not only is it miserable working conditions for the employees, the probes simply don’t function well in wet soil. It’s better to move to a different region (where it is not raining… and transport during the rain.) In my business, we ultimately divided our work into two distinct regions. At one point we tried to have four regions but it was overkill.  We simply shifted the way we defined “region.” It became more fluid and could have different factors quickly added, re-weighted or removed. One office administrator could regroup fields in a few minutes based on their own decision tree. This simple, custom software synchronized with the remote employees’ devices. In our case the regions tended to be of a north-south flavor. However, the criteria was based on density of suitable conditions.  It’s easy for a person not in the field to see this.

Application to your business: Sometimes conditions change that are outside of your control. A lot of smart people work really hard to modify the situation… or sell you products to tame the circumstance,. Perhaps it’s more efficient to simply change your method? By simply dividing your world into good/bad groups you can more easily switch your circumstances from bad to good. It’s even better if you know the groups will revert back if/when you give the zone room to revert back to good. If the bad never becomes good again, just don’t go back. Keep splitting.

Decision #2 (Is current region complete? Then, move): Our crews would move regions when they were done. This is one simple rule/decision that is good to publish. It’s not first because this is an energy optimization algorithm. The manager cares about the total magnitude of success, not just isolated success. It’s better to win the war and lose a couple of battles. But, if you win a battle, move on. At the crew level, it was cause for mini-celebrations. “We caught up!” It feels good. Celebrate in the truck… while moving regions.

Application to your business: Getting your work done is the goal. Simply by splitting groups of customers into two portions you can have mini celebrations of being done on a very regular basis. As a manager this space filling aspect allows you redirect energy in a natural way. Why send more resources into a region if another could benefit from your new found excess? Remember to define when to move on.

Decision #3 (Have you been away too long? Then, move): Our objective was to complete all the work before the change of seasons. In the fall, this was frozen ground. In the spring, it was tall crops. We could hammer through frost and walk through tall crops… but the profitability evaporated. So, once the work was done in a region we would move on. Making this rule somewhat more beautiful was the way our customers released jobs to us. In the fall, crops would be harvested and our farmer customers would let us know “Field A is ready.” Other businesses face similar situations. One contractor must wait on another to complete their task. Or, a vendor must wait on a customer to call. As a result of this irregular release pattern we would often catch up with the volume in a region. As harvest progressed we would fall behind. This is when this rule seemed counter-intuitive. Many of our competitors spent the whole of their business in a single region or with a single customer profile. We had a diverse customer type and location. We grouped customers by location and type. These groupings became our regions. Like many other industries, few service customers will tolerate a prolonged absence even with regular communication. They want to see real, tangible progress on THEIR work. This is where the construction contractor can probably nod his head… and the homeowner or developer will probably shake their head in disgust. The MUST move to a different customer in order to retain the sufficient workload. Ideally, your work is non-blocking to other vendors. Ours was non-blocking. If we failed to finish a region before moving, it increased our costs but in no way reduced the cost to the customer… until a breaking point. The fear we would miss the breaking point was why we had to move.

Application to your business: Group your customers by type and environment. We used retail, farmer-direct and geographic proximity. Grouping like kinds allowed us to keep everyone happy enough. Define what happy “enough” looks like. This will let you optimize your profit without putting your relationships at risk.

In conclusion, I think it’s also important to point out the interconnectedness of each decision. Move or stay are the two outcomes. Moving is the only diverting outcome. A decision to move reset the decision tree because the core circumstance was changed. You found yourself in a different region. Plus, the region you just left would change while you were away.

I’ve built much more complex decision trees that have different variables and different outcomes. This one stands out because it was pruned repeatedly. It was also part of a larger process. The administrator had their part. It was simplified over time, too.

This decision tree is complete but not exhaustive. It wasn’t a computer program. I left out things like “Are all regions complete?” because that came up naturally. “Uh, boss, we’re done.”

Good luck building your own operational decision trees. I recommend considering the following variables: External conditions (that you don’t control), Time sensitive variables (some things blow up if left alone for too long… and are inefficient to spend constant time/attention on at all points), and finally, space-filling conditions… when you’re done, you’re done… but sometimes things change, and there’s more.


Share the Post:

Related Posts