Algorithmic Thinking Applied to Sales

After reading the book Algorithms to Live By, by Brian Christian, I decided to incorporate some of the lessons of the book into a sales lecture I gave to a group of entrepreneurs at the Wotso co-working space in Sydney. Below are some ideas that technically-minded people may find familiar, which should help to de-mystify sales as a process.

  1. Sales as a stochastic process

In sales, you are always working with incomplete information because you only have access to the information the prospective client will share with you. Without complete information, you have two options: 1) create assumptions or 2) build a tolerance for incomplete information into your process.

There are several problems with the first option. First, people often tell us what we want to hear, whether they are trying to be agreeable or to not show all their cards. Second, if we are trying to sell something, our assumptions will be biased towards the outcome that we want – completing the sale. Therefore, the first lesson we can gain from algorithmic thinking in sales is to create a sales process that is non-biased towards potential leads.

This means treating every new lead the same way. Even when a client says, “I’ve already been through 10 demos, just send me a quote!”, you need to have a healthy skepticism about how good of a client they will be. Even if they do buy from you right away, there must be a reason that they went through 10 other competitors before finding someone to work with.

Having a sales process in place not only protects you from bad clients, but it gives you a series of touchpoints to gather data that can be used to qualify leads. Someone who agrees to have a call with you, then a meeting, then brings his manager in for a review meeting is more likely to buy than someone who simply requested a quote and cannot be bothered for a phone call. One question that remains in my mind is: how do we follow up with leads that don’t show much interest today, but may buy in the future?

  1. Exponential Backoff

Taken from Algorithms to Live By, exponential backoff is a method for handling unreliability by exponentially decreasing the level of communication. We tend to do this instinctively (for example, an unreturned text message results in a precipitous drop in communications), but the key is to not stop communicating completely. If a prospect seems very interested, then suddenly stops replying to email, perhaps you wait a few days before getting in touch, then a week, then two weeks. Moreover, it is critical to be increasingly direct. The series of messages may look like this:

  1. “Hi Joe, our software team just fixed a bug that you had pointed out earlier, just wanted to let you know. Also, we are still waiting to hear back from your team on Q3’s project. If you need anything from us to make a decision either way, please let me know” (3 days)

  2. “Hi Joe, we are getting ready to finalize our schedule for Q3, if you were still interested, we have September 1-9 available and October 14-21 available. Please let me know so that I know whether to add your group or not.” (7 days)

  3. “Hi Joe, were you still interested in moving forward with the project? If you are unable to at this time, no worries, but it would really help for planning purposes if you could confirm.” (14 days)

Although we are fairly certain that Joe will not purchase in Q3, we still do not know what he will do in the future. For that reason, it is important to keep communications going and try to obtain more information that can help us build our future pipeline. With the exponential backoff method, we save time by not chasing dead ends and can focus our time on more productive deals.

  • Explore vs. Exploit

If you are looking to go after new business, you may be exploring different lead generation techniques. The question is, how do you decide when to stop exploring different techniques and settle on the ones that work? One approach is to implement the optimal stopping algorithm. In practice, this means first defining the length of time (or money) you have to work with. Let’s say that you have one year before you run out of operating capital. Optimal stopping says that you should explore different lead generation techniques for the first 12/e months (e=2.718) or 4.5

months. After this, you chose the best technique you have and run with it to maximize expected returns. Optimal stopping is sometimes called the 37% rule, because 1/e roughly equals 0.37.

Of course, this model is a bit overly simplistic to rely on for important strategic decisions. However, it can be a good starting point to think about whether you are on pace to achieve your desired outcomes. Generally speaking, if you do not have a revenue-generating scheme in place within 37% of the time at which you will run out of money, that is not a good sign.