Insights
Most efficiency claims in this market have no before. We have one, because we ran the account by hand first — then built the system that replaced most of that work, and measured both across four months.
Case study 11 min read
Four months of measured production on one account: nine people on nine devices became two people on two devices, monthly usable output more than doubled, and cost per usable piece fell 3.99×.
3.99× cheaper per usable piece
Source OAP client in the USA — production records, May to August 2026
Read the case studyArticle 8 min
Nine questions that decide whether a number means anything, written by working through the ones that weaken our own. Includes two of our reports on the same four months that do not reconcile.
95.4% vs 91.1% the same measurement, scoped twice
Report 6 min
Four months of metered and modelled power on one content account. Nine devices drew 136 kWh a month; two draw 12. The uncomfortable part is the figure nobody can measure.
91.1% less electricity per month
Article 6 min
Cost per piece can be produced by firing people. Output per person-hour cannot, because it only improves if volume holds. Four tests to run on any production claim, including ours.
0.16 → 6.60 usable pieces per person-hour
Report 7 min
Every content budget has a rework line nobody puts in the model. On one measured account it moved the real cost per piece from $7.49 to $12.48 — a 67% difference hiding inside a quality metric.
$7.49 → $12.48 the same month, priced honestly
Every entry above is drawn from the same four months of records.
Bring your monthly volume, your current headcount and your rejection rate to the call. We will tell you plainly whether a build moves them at your size, or whether it does not.
Book a build callWe keep three entries on your device so you are not asked this again. Nothing here profiles you, and nothing is shared with an advertiser. Change it whenever you like, or ask us to delete the record entirely.
Remembers this choice so you are not asked again, and keeps the record that proves we asked. Cannot be switched off.
Would remember preferences between visits. Nothing uses this today.
Would build an advertising profile. We do not do this, and will ask again if that ever changes.
