SandBox Logistics Trial

See the well running low before it does

Three short videos from your SandBox Logistics sandbox in Nash, built on your own June 5 Permian day. The first shows your team setting up an agent by answering five questions. The other two show the truck planner and the early warning you asked for on 21 October. Every agent recommends, and your dispatchers decide in SBS.

Your June 5 Permian day, re-timedRecommend only6-week free trial

The three videos

01 · Set it up yourself · 1:51

Build an early-warning agent in five questions

"I want to create a new custom agent. Interview me with a few short questions, then propose it for my approval."

Nash asked five questions about a well-health export, wrote the check as a fixed script, tested it on the export, and created Well Sand Check once it was approved. The new agent runs on each upload and only reports.

5questions to set it up

Each uploadruns when a new export lands

Riverbend Cheddarcritical on the test run, with Hutchison as the donor

Created in the portal by answering Nash's questions, then tested on a well-health export before anything was saved.

02 · Forecast versus actual · 0:59

How many trucks does each job need tomorrow?

"It's the evening of June 4, 2026. How many trucks does each job need tomorrow, and how did yesterday's plan do against what actually happened?"

The planner plans tomorrow from the stages scheduled in each job's frac design, then grades yesterday's plan against what actually ran. No job was over-trucked, so it suggests no moves between jobs.

230trucks across 5 jobs tomorrow

3 of 5jobs under-trucked yesterday

13 shortHutchison, against a plan of 58 trucks

Daily Pace & Truck Planner, on your load history with a synthetic frac design.

03 · 6, 12 and 24 hours out · 1:01

Which well should dispatch watch first?

"Give me the outlook for every well at 6, 12 and 24 hours from now. Which well should dispatch watch first?"

The new agent runs its script on the export, then extends the same formula to 6, 12 and 24 hours. Riverbend Cheddar needs attention now, and Hutchison and PPL carry most of the volume needed by tomorrow.

Riverbendwatch first, short for its next stage

17 boxesRiverbend needs within 6 hours

161 and 118boxes Hutchison and PPL need by +24 h

The agent from video 1, on the same export and the loads inbound in Nash.

Set one up yourself

In Nash Agent, click Create a custom agent, or say in a new chat that you want one.
Answer five questions: what it does, when it runs, where results go, when it notifies, and whether it only recommends.
Attach a recent export. Nash writes the check as a fixed script and test-runs it on your file.
Read the test result and the setup it proposes. Nash points out any overlap with agents you already have.
Approve. The agent appears under Agents and runs on its schedule or each time a new file lands. To change it later, ask in a chat.

Agents already in your sandbox

Daily Pace & Truck Planner. Trucks per job for tomorrow, and yesterday's plan graded over, under or about right.
Hot-Load Dispatcher. When a well falls behind, ranks the three best trucks by ETA and proppant on board.
Well Health Watch. Puts a card on the Execute page when a well will run dry before its inbound loads land.

The six weeks, as proposed

Week 1

Logins for Sean, Rob and Chris. The first well-health export and a daily load export. We set up the early-warning agent together.

Week 2

Agents run on scheduled CSV drops. First weekly sync on what fired, what was right and what was noise.

Weeks 3 to 5

Tune thresholds and wording with Rob's dispatchers in mind. Add SBS stage times if IT can export them, and pick an integration path.

Week 6

Readout against the measures below, then a decision on rolling out to dispatchers.

Measures to agree on Thursday

Warning lead time: hours between the agent's alert and the well running dry, for every Lack of Trucks event.
Truck plan accuracy: job-days the planner calls over- or under-trucked, against what actually ran.
NPT hours from Lack of Trucks, against the same weeks before the trial. Rob's bar on 21 October was 10 hours saved per job per year.

What we need in week 1

What	Format	What it drives
Logins	Names and emails for the pilot group	Access to this sandbox
Well-health export	CSV or Google Sheet every 30 minutes, one row per well: current stage, boxes on site, boxes per stage, stages per day, next stage start	The early warning and the 6, 12 and 24 hour outlook
Daily load export	The SBS load report you sent in October, once a day	Forecast versus actual
Stage logs or frac designs	Per Permian job: stage, scheduled date, proppant, pounds per stage	A forward-looking truck plan
Well and plant coordinates	Latitude and longitude per job and plant	Real distances and ETAs
SBS contact	An IT contact for a read-only export or API	Replacing the CSV drops later

Questions for Thursday

Who is in the pilot group now, and who joins later?

Which export can you send first, and how often?

Where should alerts go: the portal, email or Teams?

What result in week 6 would make this worth rolling out to dispatchers?

Should drivers ever see a recommendation, or only dispatch? The load pool Sean described is a later track.

How to read the videos

From your data

The June 5 Permian day, the load history and the downtime log you sent in October, shifted onto today's clock.

Reconstructed or synthetic

The well-health export is rebuilt from your June 5 records, not a live SBS feed. The planner's frac design is synthetic, built from your one stage log and your pace curve. Well positions sit at each job's recorded route miles, because the export has no coordinates.

Because the frac design is synthetic, we don't claim the planner beats "same as yesterday" until a real export arrives. Agents in the trial never contact drivers or change anything in SBS.