The sales book I started in 2018

In the fall of 2018 I started writing a book. The working title was “Automating Sales,” and the draft calls itself “The Beginner’s Guide to Sales Automation.” There is an outline from October 22, a chapter on buyer personas from October 26, and a long draft last edited on December 11. It was written for anyone running a sales process, starting from none.

Most of it was about what you would expect: personas, cadences, a stack of tools. The parts that hold up best are about something else. They are about checking work done by people you can’t watch.

The method

The book’s loop was “Sale → Measure → Learn,” which the draft admits was “Appropriately stolen from the lean startup methodology Build → Measure → Learn,” Eric Ries’s loop. Attempt the sale, measure the engagement, learn, and go around again. The case for putting sales first leaned on a CB Insights figure the draft cites: 42% of small businesses fail because they can’t find a market need.

Chapter 3 laid out the cadence I recommended for a cold lead: six to eight touches, by email, social messages and calls, on days 1, 3, 7, 14, 28 and 58. Chapter 4 listed the tools. That list has aged the way tool lists do.

The part about people

Chapter 5, “Automate Lead Gathering,” is the one I would still hand to someone. It is about paying freelancers to find the right contact at each company on a list, and most of it is about verification.

Start with the price. The draft says a reliable contractor should produce a lead for between $0.30 and $0.50, and that I liked a target of $150 for 500 leads with a three-to-five-day turnaround.

Then test before you trust. Post the job, hire three to five freelancers for a small batch, and keep each worker’s submissions separate “so you can track their work and measure each one’s performance.” The draft is blunt about the odds: “Generally, if you hire 5 workers, 1 maybe 2 should pan out as strong performers.” Keep about 50% more workers on the roster than you need, because people come and go.

Then check the output against the world. Run every list through an email validation service before you send to it, and take the invalid addresses back to the contractor. The sample job post told workers this up front:

We will be testing bounce rates to determine quality before payment. If you cannot find the email address it is better to move on instead of entering a false value.

Sample job post, “Automating Sales,” 2018

Then label everything. Before leads go into the CRM, give each one hidden fields: the freelancer’s name and profile, the date, how long the batch took, the cost per lead and a link to the original job post. The draft says: “These details may seem redundant as you get up to a repeatable process, but I can promise you they will become valuable down the road.”

What I see in it now

I wrote that chapter about people paid by the batch, where a guess could look exactly like a finished lead. Eight years later I run agents, and an agent’s “done” can hide a guess just as well. The rules carried over almost one for one.

• A blank beats a guess. “Better to move on instead of entering a false value” is the same rule as one Excellent keeps: a check that cannot run stays visible. It is never counted as a pass.
• Check before you accept. Bounce rates were tested before payment. The workflow Excellent is building re-runs the checks on finished work before it counts as done.
• Keep each worker’s output separate. I tracked freelancers one by one. My agents each get their own identity, derived from the process tree, so an agent can’t approve its own work by posing as someone else.
• Label the provenance. Name, date, time, cost, source. That is a receipt, years before I used the word.

The draft also had a warning I still agree with: buying a lead list was “The absolute last resort (and we do not recommend it).” Start from the cleanest data you can get, because every later step inherits its problems.

Where it stops

The draft in my notes stops there. After a section called “Scaling out the process,” it ends on three headings with nothing under them: “Scaling out your success,” “How to monitor the process” and “How to involve Apres.”

“How to monitor the process” is, in a different form, what I work on now.