Programmable Revenue · Issue 01 · Founder Briefing

The Instrument Issue

The research behind better pipeline, in plain English. An instrument is just a number you can name and manage. Below are five sales reflexes turned into five numbers you can run, the same briefing we present at the weekly founders call.

Every finding here comes from a real, recent academic paper, not a LinkedIn post. Each one swaps a gut reflex for a number you can manage. The point is not to read the papers yourself. It is to see the shift, with the evidence attached.

1. We read the research so you do not have to

The study. Each week the newsroom runs the same search across five research databases where scientists publish (arXiv, SSRN, NBER, OpenAlex, Semantic Scholar) and pulls the go-to-market papers worth reading. One recent window returned 45 of them. None were trending on social.

What it means. The research runs about a year ahead of the vendor pitches and the feed. Four fronts are moving fast right now: ranking leads by what they do, grading what good selling looks like, software that runs real outreach, and proving which touches caused the sale. You do not need to read the papers. You need to see these shifts before your competitors do.

The move. Stop running on folklore. For one quarter, name the evidence behind every big GTM bet your team makes. No study, no scale.

2. More dials will not fix the funnel. One rate will.

The study. Researchers reviewed 286 real penalty kicks. Goalkeepers dive left or right almost every time, yet staying in the center stops more shots. Doing something feels safer than doing nothing, even when nothing is the better play [1]. "Just make more calls" is the same dive.

What it means. Pipeline is four rates multiplied: reach 40%, conversation 15%, qualify 20%, meeting 70%. They land at 0.84%, just 84 qualified meetings per 10,000 worked contacts. Lift the one conversation rate from 15% to 20% and you get 112, the capacity of a 33% headcount increase, with zero hires. Volume multiplies the whole chain by a little. One named rate bends it.

The move. Write your four rates on one line, circle the weakest, and coach that one, not the call count.

3. Lead response time is the most-proven number in sales

The study. In 2011, MIT-affiliated researchers audited how fast 2,241 companies answered a web lead, and a companion study followed 1.25 million leads [2]. Contact within the hour was about 7x more likely to qualify the lead than an hour later, and more than 60x more likely than the next day.

What it means. Fifteen years on, the typical reply still takes 42 hours, and that number only counts companies that replied at all (in 2011, 23% never did). Everyone knows speed wins. The median never moved. That gap is your edge.

The move. Pull your last 90 days, measure your median minutes to a first human reply and your slowest 10%, and put one person's name on that number.

4. Score leads on what they do, not who they look like

The study. A car company with 6.14 million leads ranked them not by who looks likely to buy, but by the buying signals each lead gives off while deciding (a test drive booked, financing opened, a car configured), then ran it live for 132 days [3]. The leads showing the signals bought far more.

What it means. Your reps can only call the top of the list, so the order decides who gets reached. Of the model's top leads, 25.76% actually bought, versus 14.41% the old way, and the live test lifted conversions 9.5%. On your top 100 calls, that is roughly 14 buyers becoming 26: same reps, same hours. Judge a lead score by who sits at the top, not a textbook accuracy chart.

The move. Decide what you rank on (the final sale, or the steps before it) and what you show reps (a grade, or how many of their top leads buy).

5. AI is sharp at some tasks and quietly bad at others

The study. Harvard and BCG ran a controlled experiment with 758 consultants doing real work, half with AI and half without [4]. The result was not "AI helps." It was split: on work inside the AI's strengths the AI half pulled ahead (about +40% quality), and on one task just outside them the AI half got more confident and more wrong (about 19 points less accurate). The model was GPT-4. Newer models move that line. They never erase it.

What it means. AI hands you hours on the work it is good at, but every competitor got the same speed boost for free, so more volume gains you nothing. Put the hours into sharper targeting and better messaging, and keep a human on the judgment calls until you have re-mapped the edge for the new model.

The move. Run your own evals. Give the latest model ten real tasks from your world, with your data, and score where it is reliable and where it is confidently wrong. The test resets with every model.

Five reflexes, five numbers you can manage

  • Running on LinkedIn folklore every play backed by a real study.
  • "Just make more calls" fix your one weakest funnel rate.
  • "We respond fast enough" answer new leads in minutes, not hours.
  • "Trust the lead score" count how many of your top leads buy.
  • "Do more with AI" re-test what AI is good at, bank the time.

Tenbound defines the model. graph8 runs the system. CIENCE delivers the outcome. The point is simple: name the number, then manage it.

References
[1]Michael Bar-Eli, Ofer H. Azar, Ilana Ritov, Yael Keidar-Levin, Galit Schein (2007). Action bias among elite soccer goalkeepers: The case of penalty kicks. Journal of Economic Psychology, 28(5), 606-621. doi:10.1016/j.joep.2006.12.001. https://doi.org/10.1016/j.joep.2006.12.001 · accessed Jun 12, 2026
[2]Oldroyd, J. B., McElheran, K., Elkington, D. (2011). The Short Life of Online Sales Leads. Harvard Business Review 89(3). https://hbr.org/2011/03/the-short-life-of-online-sales-leads · accessed Jun 12, 2026
[3]Zhang, C., Liu, Y., Sun, Y., Zhang, X., Cao, Y., Jiao, J. (2026). Rethinking Sales Lead Scoring with LLM-based Hierarchical Preference Ranking. arXiv preprint, cs.IR. arXiv:2606.04387. https://arxiv.org/abs/2606.04387 · accessed Jun 12, 2026
[4]Dell'Acqua, F., McFowland, E., Mollick, E., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., Krayer, L., Candelon, F., Lakhani, K. (2023). Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality. SSRN 4573321; Organization Science. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4573321 · accessed Jun 12, 2026