Simulated Annealing Algorithms for Outbound Sales

This is a guest post from Rob Snyder, Incendium's Founder in Residence.

Hi all —

Let’s talk outbound sales.

For early-stage founders, outbound sales is *the* best approach to getting traction. Someday you’ll probably invest in things like SEO, ads, community, etc. etc. etc.

But outbound is what really matters in the early stage for three reasons:

  • You get rapid feedback (unlike SEO, which can take months)
  • It’s relatively low cost (unlike ads, where you can spend a bunch of money and NOT learn anything)
  • It helps you find demand without thinking very hard (unlike other channels where you need to have a thesis and/or a strategy about where demand is)

This last point is underrated. Wildly underrated. Here’s why.

We think we know a lot of things:

  • Who our target buyer is
  • What they need to achieve
  • What they want to buy
  • What message will work with them
  • What they are likely to search in Google

But how do we know these things? Two ways we might know:

  • Theoretical knowledge: We believe it’s the case because we’ve written it down and thought it through
  • Empirical knowledge: We’ve proven it’s the case through a repeatable selling motion that converts customers predictably, every month

If your answer falls into “Option 1,” you don’t *actually* know. You have theories, but no real answers.

Don’t worry - you’re not alone. Most startups fall into Option 1. Many Series A - B companies have Theoretical knowledge that they articulate beautifully, but don’t map to reality.

How do you get from Theoretical knowledge to Empirical knowledge? You sell, albeit with a different approach to selling than what’s typically taught.

Simulated Annealing vs. A/B testing

If you’re smart, you’re likely thinking about creating effective testing methodologies to “prove” where to focus. You’re considering A/B tests, pulling from the hypothesis-driven “Lean Startup” approach or that systematic “product-market fit survey” you saw from that VC blog.

This is all bullshit.

It is the worst kind of scientism, because it provides the veneer of “science” in a domain where the rigors of the scientific method aren’t helpful.

You know this if you’ve tried to design effective outbound experiments: There are too many variables to be able to get real signal from any sort of A/B test or designed experiment.

Don’t believe me? Just look at all the targeting filters in Apollo, and consider that tweaking each individual targeting filter slightly could constitute an A/B test, holding all else equal. The problem? You don’t know which variable actually matters - or if none of them do - and you don’t have 1,000 years to test them all.

Don’t worry though, because there’s a simpler approach. You just have to understand simulated annealing algorithms:

File:3D TSP solved with simulated annealing 2.5 MB.gif - Wikimedia Commons

Looks complicated? It is. Here’s the explanation for us MBAs: You do a random search of a multi-dimensional space, and double-down on what seems to be the global optimum.

How does this relate to outbound? You do a high volume of relatively untargeted outbound, figure out who replies, and double down on that kind of person / company through continued iteration. You let the market pull you, rather than coming with strong hypotheses that are likely to be wrong in some unknowable way ex ante.

This means thousands and thousands of outbound emails that lead you to a more refined search radius. It means almost-daily debugging of your search radius to find the “search” that actually works consistently. And occasionally, you stop to think, “why the hell did THAT work?”

You can also just do a TON of cold calls to get even faster feedback.

Once you have empirical knowledge of WHERE demand is (aka, you’re booking a meeting or two per day pretty consistently with a semi-refined search), you then layer on more targeted outreach approaches:

The Outbound Portfolio

Outbound isn’t just one thing that works… it’s a portfolio of different approaches. In time, everything degrades and needs to be replaced.

Fortunately, it’s now relatively easy to set up a portfolio of approaches that can be more or less automated. Here’s an example of an outbound portfolio:

  • Simulated Annealing: Continuing with the large-scale, relatively untargeted outbound, hitting contacts with a new sequence every 6-ish months (unless they’re in a more targeted list below)
  • Perfect-fit list: A set of 100 accounts per rep that are EXACTLY LIKE our perfect-fit customers. Reps network into these accounts, do cold-calls, personalized emails and LinkedIn approaches
  • New hires with relevant titles: An automated outbound campaign that sends “congratulations” messages and relevant content to new hires in good-fit companies that fit certain titles
  • Hiring for a relevant title: An automated outbound campaign that sends “advice” messages and relevant content to certain titles whose teams are hiring for certain roles
  • Lost a key employee: An automated outbound campaign when employees with certain titles leave
  • Engaged with certain LinkedIn posts: An opportunistic approach to relevant LinkedIn posts that have gone viral within our target buyer segment
  • Quarterly / Annual planning: A two-email campaign that goes out to relevant accounts about the time of quarterly / annual OKR planning
  • Join our podcast / industry report: An email from the CEO doing research for an upcoming podcast episode or industry report

You can probably think of a bunch of other ones too. The cool thing is that 1-2 people can set up and manage a portfolio like these. It’s actually kinda fun, and what I’m doing with a few companies now!

But before you jump into the portfolio approach: These things ONLY work once you know where demand is. Empirically, not theoretically.

Introduce yourself!

What are you interested in working with us on?
Multiple choices possible!
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.