The cost of a failed batch

The short answer: on our model, a failed 30-day iPSC neuron differentiation destroys $3,800 to $6,500 of directly attributable cost before you have replaced anything, and the replacement cycle costs the same again, plus six to eight weeks you cannot buy back.

This page is a model. Every unit price in it is real and sourced. Every quantity, hour and yield is our assumption. It is a structure you can put your own numbers into, which is the thing nobody publishes. Treat the total as illustrative rather than as a quotation.

Why this page exists

Every cost discussion in this market prices success. Suppliers quote what a vial costs, cores quote what a differentiation costs, and buy-versus-build models assume batch one works.

Failure is where the money actually goes, and it is invisible in every price list because no supplier and no core facility publishes a failure rate. We searched for one and could not find a single published figure from a commercial or institutional source. So this page does two things: it prices the consequences of a failure using real sourced unit costs, and it shows what the answer does across a range of failure rates rather than asserting one.

The batch being modelled

ASSUMPTIONS, all of the following are our choices, not sourced figures.

The unit prices

SOURCED, every figure below was read from a named supplier storefront or a published rate card on 2026-09-01.

For the consumables line we carry forward $1,567.33 per batch, the figure built line by line from STEMCELL Technologies published list prices in the buy-versus-differentiate model. The derivation of that number lives on that page and is not repeated here.

Layer 1: what a failure destroys directly

MODELLED. A batch that fails at day 30 has consumed everything a successful batch consumes and produced nothing. There is no partial credit: a differentiation that yields the wrong cell type, or too few cells, or cells that will not attach, has cost full price.

That is the figure most people mean when they say a batch failed. It is not the whole cost.

Layer 2: the starting material

MODELLED. A failed differentiation usually consumes a vial of the parent line, and if the failure cause is unknown, it often consumes a second one during troubleshooting.

If the line is irreplaceable, a patient-derived line, a bespoke edit, a disease line with no current stock, this row is a project risk rather than a price, and the correct response is a banking policy rather than a budget line. See cell banking costs.

Layer 3: diagnosing the failure

MODELLED. This is the layer people forget entirely, and it is frequently larger than the consumables.

When a batch fails you do not know why, and the differential diagnosis costs money to work through. Using published assay prices:

Add expert time. At the Gates Center’s published consulting rate of $175.88 an hour, or Baylor’s $53.75, four hours of someone senior working out what went wrong is $215.00 to $703.52.

A proper failure investigation is a $2,000 to $3,300 event. Most laboratories do not run one. They repeat the batch and hope, which is a decision with a cost that shows up in Layer 5.

Layer 4: what the failure took down with it

MODELLED. This is where failed-batch cost stops being linear.

If the batch was destined for a downstream experiment that had already started, the failure destroys that too. Using the sourced figures above, a common configuration:

Notice the shape of that. The glia vial and the plate are lost the moment the neurons are, and neither is recoverable. The astrocytes at $1,081 cost more than the plate, more than the labour, and more than a catalogue vial of the neurons they were supposed to support.

This is the single most common way a “cheap” failure turns expensive: the failure is discovered at week three of a six-week maturation, by which point three other purchased inputs have been committed to it.

The totals

MODELLED, the sum of Layers 1 to 4. Not a quotation.

The headline range at the top of this page, $3,800 to $6,500, is deliberately narrower than these totals, because it covers only the consumables and cabinet time, which is what most people count. The full attributable cost is two to four times what most people count.

The schedule, which is not priced here

We are not going to put a dollar figure on lost time, because we cannot source one. Anyone who tells you a week of programme delay costs a specific amount is either quoting your own internal opportunity cost back at you or making it up.

What we can give you is the elapsed time, from published sources.

A failure discovered at the network-activity stage puts you roughly ten weeks back, not thirty days: the differentiation again, plus the maturation again. If the parent line itself is the problem and has to be re-derived or re-edited, add the 8-to-16-week editing window on top.

The published detail that matters most here comes from the Lemieux study: spontaneous firing appears around 11 DIV with no network activity at that point, and the mean firing rate only begins to climb after roughly 17 DIV. That is a five-to-six-week window during which a healthy culture and a failing culture look similar. It is why failures in this field are discovered late and expensive rather than early and cheap.

Failure rates: what is published, and what is not

We looked for a published differentiation failure rate and found none from any supplier or core facility. We are not going to invent one. What does exist is adjacent evidence about how often the inputs are compromised, which is a lower bound on how often things go wrong:

• Mycoplasma. The National Center for Advancing Translational Sciences reported that in the first year of its cell culture quality control programme, 13 percent of cell line samples were positive (Roth et al. SLAS Discovery 2020). An independent survey of NCBI’s RNA-seq archive found 11 percent of series with cultured samples had at least one sample with 100 or more reads per million mapping to mycoplasma (Olarerin-George and Hogenesch, Nucleic Acids Research 2015). Two very different methods, one contaminating a laboratory’s own incoming lines and one contaminating the published literature, converging on roughly one in eight.
• Line-to-line variability. Volpato and Webber (Disease Models & Mechanisms 2020) set out guidelines for addressing variability in iPSC-derived disease models specifically because differentiation outcomes vary between lines in ways that confound experiments. Variability is not the same as failure, but a line that differentiates poorly produces batches that fail.

Use those as a floor, not a rate. A mycoplasma prevalence around 11 to 13 percent tells you that contamination alone accounts for a meaningful fraction of failures before you count protocol drift, reagent lot variation, operator error and line-specific differentiation propensity.

What the failure rate does to your cost per usable batch

MODELLED. The effective cost of a successful batch is the batch cost divided by (1 − failure rate). Using the Layer 1 direct cost at academic rates, $6,309.60:

At the mycoplasma-prevalence floor of 13 percent, in-house differentiation at commercial labour rates is already $822 per batch behind buying the catalogue vials, before Layers 2, 3 and 4 are added, and before any other failure mode is counted.

That is the real case for buying, and it turns on who absorbs the failure rather than on quality. When a supplier’s batch fails, they eat it. When yours fails, you eat all four layers.

The comparator: what a core charges, failure included

Boston Children’s Hospital IDDRC publishes a differentiation service fee schedule, and the comparison is instructive because a core’s price includes their failures:

$2,334 for a batch of three-week-old NGN2 cortical neurons, delivered. That is roughly a third of our modelled Layer 1 direct cost, and the core carries the failure risk rather than you.

The yield is not stated in the same units as our model, “one batch = two 10 cm dishes” is not 10 million cryopreserved cells, and the two are not directly comparable. But the direction is clear and worth taking seriously: for anyone who needs neurons rather than needing to make neurons, an institutional core that has run the protocol hundreds of times is cheap, and its price already includes the batches that did not work.

Reducing failure cost: what actually moves the number

Ranked by effect on the totals above:

1. Test the inputs before you commit downstream material. A $134.27 mycoplasma test on the parent line before you thaw a $1,081 astrocyte vial and coat a $488.80 plate is the highest return in this entire model. At a 13 percent contamination prevalence, that test has a positive expected value on the Layer 4 loss alone.
2. Stage the commitment. Layer 4 exists because glia, plates and instrument time get committed before the neurons are known to be good. Any checkpoint that defers those purchases past the point of no return removes $2,683.80 from the failure cost.
3. Bank the parent line properly. Layer 2 becomes a project risk rather than a cost when the line is irreplaceable. Banking is cheap relative to re-derivation: contract expansion and banking is a 4-to-6-week job, CRISPR re-editing is 8 to 16 weeks.
4. Keep an isogenic or catalogue fallback identified before you need it. Knowing that a $643 vial can rescue a timeline is worth more at the point of failure than at the point of planning.
5. Run the diagnosis rather than repeating the batch. A $351.27 minimum workup is cheaper than a second $6,309.60 batch that fails for the same unidentified reason. Laboratories skip this constantly and it is straightforwardly false economy.

Using this model with your own numbers

The three inputs that move the total most, in order:

1. Whether downstream material was committed. Layer 4 is $0 or $2,683.80 and nothing else in the model has that switch-like behaviour.
2. Your labour rate. $143 against $179 an hour is $864 on Layer 1 alone.
3. Your failure rate. Between 0 and 33 percent, the effective cost per successful batch moves by 49 percent.

Everything else, reagent brand, exact media volumes, which mycoplasma assay, moves the answer by less than any of those three. If you refine one number, refine your failure rate, and refine it by actually recording it rather than estimating it.

How we can help

We built this model because clients kept describing failed batches in weeks and we wanted to describe them in pounds and dollars. If you send us your batch size, your labour rate, your observed failure rate and what gets committed downstream, we will run the same arithmetic on your numbers and put current quotes against the fallback options.

Frequently the useful output is the observation that a $134.27 test sitting in front of a $2,683.80 exposure is the cheapest thing on the page, and that nobody had sequenced the workflow that way.

Send us a specification.