# LeverOps is new. I'm not.

Source: https://leverops.co/track-record/
Last updated: 2026-10-13

I have spent a decade inside tech startups, close to the operating detail, including two startup exits. I currently work as a vendor to a private-equity-backed software company, rolling out AI with guidance from its PE partners.

LeverOps has no client case studies yet, and I will not dress that up. What follows is the same work done inside a much larger company: where I looked, what I found, and what I built. The company is not named. The numbers are real.

**How to read the numbers.** Counts and percentages are measured from the company's own systems. Dollar figures are the size of the problem found, not money recovered. Where something is an estimate, it says so.

## Problems I found

### Leads were waiting more than a day

I measured what happened to every new lead after it arrived. The median lead waited 27 hours for its first touch. Only 8.9% were touched within an hour. 17% never got any action from a rep at all. The dashboard had not shown it, because it only counted the leads that were worked.

### Most leads never reached a salesperson

Only 16.2% of the leads created were ever assigned to a rep. Nearly all conversions, 99.8%, came from one stage of the process, which converted at 89.2%. Another 46,295 leads sat in a stage converting at 0.43%, and 54,751 sat in one converting at zero.

### Open deals had gone quiet

133 qualified, open deals worth $31.3M had no recent activity and no one chasing them. The fix is an alert, the cheapest item on the list. It pays for itself with one recovered deal.

### Lost deals nobody went back to

$92M of pipeline was sitting closed-lost with no process for revisiting it. At the company's own win rate that pool is worth roughly $15M in bookings. That figure is an estimate.

### "Rep behaviour" that was a scheduled job

Leads appeared to be reworked by reps 19.6% of the time. 44% of those reworks landed on exactly day 66 of a lead's life: a scheduled sweep, not a person's judgement. 88.9% of ownership changes were bulk operations. The team had been managing a behaviour that did not exist.

### Lead cost varied 237 times by source

One purchased data source supplied 56.5% of all leads at $7.54 each. Leads from chat and support cost $1,785 each. And 82% of revenue had no lead source recorded, so nobody could say which spend was working.

## A build I stopped

The plan was to make reps fill in required discovery fields before a deal could advance. Before anyone built it, I tested the idea against 2,027 past deals. Deals with none of the fields filled were won 15.8% of the time. Deals with three or four filled were won 8.5% of the time. The build was dropped before it cost anything.

Sometimes the right answer is to change nothing. I will tell you when it is.

## Numbers I took back

Three of my own early estimates did not survive a second check: a dead-pipeline model and two renewal figures. I withdrew them before anyone acted on them. If I cannot defend a number, you will not see it in a memo.

## What I built

### A research agent for the sales team

About 35 business development reps were researching each new lead and account by hand. I built a hosted automation that does the research and hands the lead to a person whenever the AI cannot make a confident call. Run it twice on the same lead and the output is identical.

My estimate is about 10 minutes saved per lead. Across roughly 1,000 leads a month that is about 2,000 hours a year, close to one full-time person. That is an estimate, not yet a measured result: the agent is in preview and account matching is still being improved.

### Call records tied to deals

Matched 99.5% of 15,338 recorded sales calls to the deal they belonged to, so managers could see what was said on the deals they were forecasting.

### A forecasting tool in software the team already owned

Built in Excel and Power Query, with nothing new to buy or learn.

## What this means for a business your size

The numbers are bigger there. The problems are the same ones: leads that wait, work that gets re-typed, a report nobody trusts, a fix that would not have helped. Finding them takes the same method, which is to follow the real work and count.

That is what a [Workflow Snapshot](/workflow-snapshot/) does for one workflow in your business, for $950 in a week. [Book a free operations review](/book/) to talk it through, or see [every price](/pricing/).