Book a demo
Menu

Planning for advanced therapies

One patient, one batch, one plan.

Provenesis plans cell therapy networks the way they actually run. It lines up apheresis collection, courier windows, manufacturing slots, QC release, and infusion for every patient, and prices each trade-off in dollars.

Book a demoRead the CAR-T planning brief
One plan across the networkSix patient rows on a shared timeline. Each row starts with a collection, followed by a manufacturing bar labeled with its suite, a QC release bar, and an infusion. Below, a QC release capacity lane shows the testing workload against a capacity line. Where three batches are in release at once, the workload reaches capacity and the window is highlighted as binding, with a dollar value per added analyst shift.PATIENTHORIZONcollectinfuseP-01S1QCP-02S2QCP-03S3QCP-04S1QCP-05S2QCP-06S3QCQCRELEASELOADcapacityBINDINGλ $ per analyst shift
  • Manufacturing slot: one patient, one batch
  • QC release testing
  • Binding constraint: shared QC release capacity
Fig. 1One plan across the network. Each row is one patient's batch, from collection through manufacturing and QC release to infusion. Shared release capacity sets the pace.
  • Autologous and allogeneic
  • Vein-to-vein scheduling
  • Slots, suites, staff
  • QC release capacity
  • Priced in dollars

01The problem

Patient-specific supply chains break MRP logic.

Most planning tools were built for discrete manufacturing: averages, fixed lead times, and a bill of materials. An autologous therapy has none of that slack. Every batch starts with one patient's cells and ends with that patient's infusion.

  1. 01

    Every batch is a patient.

    There is no safety stock to draw on. A batch that fails release can mean a second leukapheresis for a patient who is already sick. The plan has to protect each patient's path, not a fill rate.

  2. 02

    Many clocks run at once.

    Collection dates, courier windows, apheresis material stability, manufacturing duration, QC test durations, and the treatment center's lymphodepletion and infusion schedule all have to line up for each patient.

  3. 03

    Averages do not hold.

    Fixed lead times assume every run takes the typical time. Expansion yields, turnaround times, and failure rates vary from patient to patient, and the plan breaks on the patients who are not average.

  4. 04

    Capacity is more than suites.

    A slot is only real if the suite, the trained operators, the QC analysts, and the vector lot are all available at the same time. Most tools plan one of these and leave the rest to email.

  5. 05

    The real plan lives in spreadsheets.

    When the planning system cannot represent the process, planners rebuild it in spreadsheets and absorb every disruption by phone. That works until volume grows.

  6. The technical brief covers how to model a CAR-T network for planning: slots, staff, QC capacity, and variability.

    Read the technical brief

02How it works

From your network to decisions you can act on.

  1. Step 1

    Model your network.

    Apheresis sites, courier lanes, manufacturing sites, suites and slots, operators and shifts, QC capacity, vector supply, and treatment centers.

  2. Step 2

    Add your patients.

    Confirmed and expected patients from your referral pipeline and demand forecast, plus the dates already promised to treatment centers.

  3. Step 3

    Optimize.

    Mixed-integer optimization finds the best plan that meets every constraint, not just the first plan that fits.

  4. Step 4

    Act on decisions.

    A schedule for each patient and each suite, the trade-offs behind it, and what relaxing each constraint is worth in dollars.

See how it works

03What is different

Three ideas behind every plan.

01

Economic realism.

Every decision is priced in dollars: an idle slot, an added shift, an expedited courier. Supply chain, manufacturing, and finance compare one number instead of negotiating between their own KPIs.

  • Dollar-denominated plans
  • The cost of every constraint
  • One number across teams

02

De-averaging.

Expansion yields, manufacturing failures, QC retests, and transit times vary. Provenesis plans with that variability in view, so the plan holds up when a run comes in long or a batch fails release.

  • Variable yields and turnaround
  • Failure and retest rates
  • Plans tested against the range

03

Decisions, not dashboards.

A dashboard shows you the bottleneck. Provenesis tells you what to do about it: which patient takes which slot, when to collect, when to release. Feasible is not the same as optimal.

  • Mixed-integer programming
  • Concrete schedules
  • Explicit trade-offs

04Trade-offs

See what every constraint costs.

Every plan comes with the reasons behind it. Provenesis reports which constraints bind, what each one costs, and what relaxing it would be worth.

That turns a capacity debate into a decision: another slot, another shift, or another QC analyst, each with a dollar value attached.

Constraint detail

Illustrative

constraint
qc_release_capacity[site, d]
form
Σ tests[b, d] ≤ analysts[d] × tests_per_shift
status
binding in weeks 2 and 3
value of relaxing
λ[d] $ per added analyst shift
drivers
sterility_hold, retest_rate, release_queue
Binding constraints
Which limits set the plan, by day, by site, and by resource.
Value of relaxing
What one more slot, shift, or analyst is worth in dollars.
An inspectable model
Sites, suites, and constraints are named and readable, so your team can see why the plan chose what it did.

05Proof

Already planning in production.

Provenesis grew out of a planning system we built that runs in production at a major pharma company's advanced therapies business, where it replaced a legacy planning system.

Built from years of hands-on planning work across process-industry clients.

  • Results

    [TODO: quantified result, needs approval]

    For example, slot utilization or planning cycle time, stated with its assumptions.

  • References

    [TODO: confirm whether a reference conversation can be offered]

    Client names stay confidential. A reference would be arranged privately.

06Team

Built by optimization people who have worked inside plants.

The team behind Provenesis comes from optimization and operations research, with years of planning work at large process manufacturers [TODO: approve naming former employers]. We build planning systems that run in production, and we know what a plan has to survive on a Monday morning.

About the team

Next step

See Provenesis on a network like yours.

Book a working session with the founder. Bring your hardest planning week and we will show how Provenesis would plan it.

Book a demoGet the CAR-T planning brief