In this article
Five prices hide in one bill
On 19 October 2026, a software services firm we work with ran one lead search against a $20 budget.
Every row below is the same bill. A seller can quote any of them. Only the kept and ready rows say what you bought.
| Unit | Count | Price per unit |
|---|---|---|
| Search query | 226 | 2.3 cents |
| Post found | 2,090 | 0.25 cents |
| Post judged | 1,635 | 0.32 cents |
| Lead kept | 70 | 7.5 cents |
| Lead ready to contact | 21 | 25 cents |
97% went to finding, not judging
Searching cost $5.08. Judging 1,635 posts cost $0.15.
Cheap judging changes the design. When reading a post costs a hundredth of a cent, you judge everything and drop freely.
The costly habit is the opposite: finding less, then defending each find because it cost money to fetch. Which sources earned the finding money is in the breakdown's table.
Cutting one plan by a quarter
The same day, a cost check went line by line through the firm's monthly plan of repeated searches. The $20 above is one run's cap. This $11.00 is a month's plan. The check removed paid lookups whose answers cheaper steps already gave.
| Line | Planned | After the pass |
|---|---|---|
| Funding checks | $3.48 | $1.50 |
| Job-ad checks | $3.95 | $3.14 |
| Whole plan | $11.00 | $8.21 |
Coverage stayed the same. A plan bigger than the budget is what turns that cut into sums, not argument.
What the 25 cents leaves out
The ready row buys a graded lead with evidence, not a meeting. A person still checks each keeper, writes to real people and answers replies. Reading the 70 keepers alone takes an afternoon, which costs more than the $5.23 machine bill at any salary.
So the question is never whether 25 cents is cheap. It is whether the 21 leads deserve the human hours that come next. That is what the pilot scorecard measures.
If 1 of the 21 becomes a held meeting, that meeting's machine cost is the whole bill. If 5 do, it is $1.05 each. We publish no meeting counts yet, because that data is too young.
Ask every seller for this ladder
Ask for one real run's bill, split the same five ways, and ask which counts feed it. Ours carry tests, a habit from a count that once lied to us.
Our seat price is public: $1,000 a seat a month. Mach isn't open yet. The ladder is what sits under a number like that.
A low seat price with no kept leads is the most costly choice of all.
Common questions
Is 25 cents a benchmark I should expect?
No. It is one run for one firm in one market. The part to reuse is the ladder, not the price.
Why is judging so much cheaper than finding?
Searching pays outside services for each request. Judging reused each post's own text with fixed questions and a fixed scoring formula. That costs almost nothing per post.
Where the numbers come from
One bill and one run's stored counts. One firm we work closely with; run of 19 October 2026; exported 20 October 2026.
