Unit price is a purchase order number, cost per reported result is a--HPLC Filter

Last quarter a QC analyst showed me a spreadsheet where her cost per reported result was $4.18 against a filter unit price of $0.71. Same lot, same method, same instrument. The gap was not freight or markup. It was 11 reruns out of 60 injections, each one carrying 22 minutes of prep and a fresh 2 mL aliquot through a HPLC filter. That is what an hplc filter cost per test calculation looks like when you stop buying filters and start buying results. This article is about how to build that number honestly, and why two filters that differ by 40 cents on the invoice can differ by three dollars on the bench.

Unit price is a purchase order number, cost per reported result is a production number

The invoice price answers one question: what did the box cost. The cost per reported result answers a different one: what did it cost to put a defensible number in front of a reviewer. Those two figures diverge the moment a filter contributes to a rerun, a reject, or a prep step that has to be repeated.

The working formula I use has five terms, and only the first one appears on any quote:

Filter unit cost divided by filters consumed per injection (usually 1.0, but not always, see the syringe filter case below)
Prep labor per injection, in minutes, times the loaded hourly rate
Instrument time per injection, including equilibration if the run aborts
Rerun cost = rerun rate times the full cost of one injection
Reject cost = reject rate times the cost of re-sampling from the source, which is always higher than a rerun because the original aliquot is gone

A 60-injection batch on a typical binary HPLC with UV detection costs roughly $38 to $52 in instrument and analyst time per injection once you load the full burden rate. If your rerun rate is 8%, that is an extra $3.04 to $4.16 per reported result. The filter itself contributes $0.40 to $0.90. The rerun term is four to eight times the filter term. That ratio is the whole argument.

I want to be blunt about where the biggest error usually sits. It is not in the filter price, which is easy to look up. It is in the rerun rate, which almost nobody measures by lot. When I audit a lab’s cost model, the rerun rate they quote is usually the one their LIMS computes across all methods and all matrices, and it is almost always lower than the per-method figure that actually drives cost on a single product line. A site-wide 3% rerun rate can hide a 14% rate on one method that runs 40% of the volume. That single method is where the money goes, and it is the one the aggregate number buries.

Why the prep time term is not zero

Filtering a sample takes 20 to 40 seconds of hands-on time if everything goes right. It takes 4 to 6 minutes if the filter clogs at 60% of the aliquot and the analyst has to swap to a second unit, re-label, and re-enter the vial position in the sequence table. On a 96-position autosampler that re-entry alone is 90 seconds. Multiply by 60 injections and the difference between a clean filter and a marginal one is 1.5 to 2.5 hours of analyst time per batch. At a $75/hour loaded rate that is $112 to $187 per batch, or $1.87 to $3.12 per result. That number dwarfs the unit price difference between grades.

There is a second prep-time term that gets missed entirely: the re-draw. When a filter clogs mid-aliquot, the analyst does not just change filters. They go back to the source vial, which may already be capped and returned to the autosampler tray, and they draw a fresh aliquot. On a stability study where the source vial has a limited volume, that re-draw can exhaust the vial and force a resample from the stability chamber. I have seen a single clogged filter on a 200 µL stability aliquot cost a full time-point resample, which is a $300 event in sample chain-of-custody paperwork alone. That is 400 filter units at Grade A pricing.

The two grades I benchmark, and what actually separates them

For comparison I use two real product classes that sit at opposite ends of the price spectrum for the same nominal duty. Call them Grade A and Grade B. Both are 13 mm, 0.22 µm, PVDF or PTFE membrane in a polypropylene housing. The difference is in the membrane lot control, the housing weld, and the QC sampling density.

Parameter	Grade A (controlled)	Grade B (commodity)
Unit price at 1,000 qty	$0.71	$0.31
Membrane lot traceability	Per-lot, with extractables data	Per-shipment, no extractables
Housing weld	Ultrasonic, burst tested to 8 bar	Solvent bond, no burst test
Hold-up volume	< 25 µL	35 to 50 µL
Extractable UV response at 210 nm	< 0.002 AU	0.004 to 0.012 AU
Puncture count (inlet septum)	≥ 40	18 to 25
Observed rerun rate (our data, 6 months)	2.1%	9.4%
Observed reject rate	0.4%	3.1%
Prep time per injection	28 s	41 s

Run the formula on those numbers at a $45 per-injection burden rate and a $75/hour analyst rate:

Grade A: $0.71 filter + $0.58 prep + $0.95 rerun + $0.18 reject = $2.42 per reported result
Grade B: $0.31 filter + $0.85 prep + $4.23 rerun + $1.40 reject = $6.79 per reported result

Grade B is 56% cheaper on the invoice and 181% more expensive on the bench. That inversion is not unusual. It is the default outcome whenever the rerun rate crosses about 5%.

Note what the table does not show: the cost of the column. A Grade B filter that sheds a septum core once per 500 injections puts a fragment into a $600 column roughly every second month on a busy method. Add that event at a 50% column survival rate and the Grade B per-test cost climbs another $0.60. The invoice never sees it, because the column is a separate budget line and the filter is not blamed.

The rerun rate is the variable you actually control

Most labs track rerun rate as a quality metric, not a cost metric. Flip that. A rerun rate measured per 100 injections, tracked by filter lot, tells you within two weeks whether a grade change is paying for itself. If reruns drop from 9.4% to 2.1% on a 60-injection batch, you save 4.4 injections worth of instrument time, which at $45 is $198 per batch. The filter price increase across that batch is $24. Net saving: $174 per batch. That is the entire business case, and it takes one month of data to prove.

The trap is measuring the rerun rate across a mixed population. If half your injections are clean water samples and half are high-turbidity soil extracts, the aggregate rate will look fine and the soil method will be bleeding. Split the rate by matrix before you draw any conclusion about a filter grade. I have watched a lab conclude that a premium filter was not worth it because their aggregate rerun rate was 2.5%, when the soil method alone was running at 19% and the water method at 0.4%. The premium filter fixed the soil method and did nothing for the water method, which is exactly what you would expect if you had split the data first.

Where the hidden cost hides: three failure modes that show up on the bench

Reruns do not come from nowhere. In my experience running these lines, three filter-related failure modes account for most of the variance between grades. Each one has a specific measurement that identifies it.

Extractable response at low wavelength

A commodity membrane can shed UV-active species into the filtrate. At 254 nm you will not see it. At 210 nm, where many pharmaceutical and environmental methods run, a Grade B filter can produce a baseline drift of 0.004 to 0.012 AU over the first 3 minutes of the run. That is enough to shift an integration baseline and trigger a manual re-integration, which is a rerun in everything but name. The measurement that catches it: run a blank filtrate through the same method and compare the first 3 minutes against a Grade A blank. If the difference exceeds your method’s baseline noise specification, the filter is contributing.

The extractables problem compounds with temperature. A filter stored at 30 °C in a warehouse for six months can shed two to three times the extractable load of the same lot stored at 20 °C. If your lab stores filters on a bench near a window or a heat source, your measured extractables will not match the datasheet, and your rerun rate will drift upward through the summer. I have seen a lab chase a “seasonal” rerun pattern for two summers before tracing it to a filter cabinet next to a south-facing window. Moving the cabinet dropped the summer rerun rate by 4 points.

Housing weld failure under backpressure

Solvent-bonded housings fail at 3 to 5 bar. Ultrasonic welds tested to 8 bar hold. The failure mode is not a burst, it is a slow leak at the inlet that shows up as a bubble in the filtrate and a phantom peak on the chromatogram 40 seconds in. The measurement: weigh the filter before and after a 2 mL filtration at 4 bar. A Grade A unit gains less than 5 mg. A failing Grade B unit gains 15 to 40 mg, which is filtrate that went around the membrane rather than through it. That is a reject, not a rerun, because the aliquot is contaminated.

The backpressure term is where syringe geometry matters. A 1 mL syringe with a 13 mm filter generates about 2 bar of backpressure at a normal thumb push. A 5 mL syringe on the same filter generates 4 to 5 bar at the same flow rate, because the plunger area is larger and the analyst pushes harder without realizing it. If your method uses 5 mL syringes, a solvent-bonded housing is already at its failure ceiling. That is not a filter defect, it is a system mismatch, and it shows up as a reject rate that nobody can explain until they measure the pressure.

Septum coring on multi-puncture inlet

If your workflow uses the filter inlet septum as the sample entry point (common on inline filter tips), the puncture count matters. Grade A septa hold ≥ 40 punctures before coring produces a visible fragment. Grade B septa start shedding at 18 to 25. A fragment downstream of the filter but upstream of the column is a column killer, and a column replacement is $600 to $1,200. One cored septum per 500 injections at a 3% fragment escape rate costs more than the entire year’s filter budget for that method. The measurement: inspect the septum under 10x magnification every 20 punctures and log the first visible core.

Two field cases where the per-test math changed the purchase decision

Case one: the 210 nm drift that looked like a column problem

A contract lab running a USP method at 210 nm had a 14% rerun rate and had already replaced two columns in six months. The symptom was a drifting baseline in the first 3 minutes, more pronounced on Monday morning runs. The measurement that identified it: a blank filtrate injection through the same method compared against a filtered mobile phase blank. The blank filtrate showed a 0.009 AU rise over 3 minutes that the mobile phase blank did not. The filter was a Grade B PVDF at $0.29. Switching to a Grade A PTFE membrane filter at $0.68 dropped the blank drift to 0.001 AU and the rerun rate to 2.4% within three weeks. The columns stopped failing. The lab had been attributing a $2,400 column replacement cost to the column vendor when the actual cause was a 39-cent filter difference per injection.

The Monday-morning detail was the clue that cracked it. Filters sit in the autosampler tray over the weekend with the method queued. The Grade B housing, solvent-bonded, was leaching into the residual liquid in the filter during the 60-hour idle period, and that leachate was the first thing the Monday run injected. The Grade A ultrasonic housing did not leach. Same method, same analyst, same instrument, 39 cents a filter. The fix took one purchase order.

Case two: the 96-well prep that lost 40 minutes a batch

A bioanalytical group filtering 96 samples per batch through 0.45 µm filters saw prep time balloon from 25 minutes to 68 minutes per batch. The symptom was filter clogging at roughly 60% of the aliquot on about 1 in 5 samples, forcing a second filter and a re-entry of the well position. The measurement: they weighed each filter before and after filtration and found the clogged units had retained 180 to 240 µL of a 500 µL aliquot, versus 40 to 60 µL on the clean units. The retained volume pointed to a membrane with a wider pore size distribution than specified. Switching to a lot-controlled filter with a stated pore size distribution (mean 0.45 µm, ±0.05 µm) brought prep time back to 27 minutes. The per-batch saving was 41 minutes of analyst time, or about $51, against a filter cost increase of $9 per batch. They also moved the filtrate into a vial insert to cut the dead volume term, which shaved another 8 seconds per injection off the prep step.

Building your own cost per test calculation in four steps

You do not need a spreadsheet model with 40 rows. Four inputs get you 90% of the answer.

Measure your current rerun rate and reject rate by filter lot. Two weeks of data is enough. Log the lot number on every injection in your sequence table. If your LIMS does not carry the lot, add it as a custom field. This is the single most valuable number in the calculation.
Time one prep step with a stopwatch. Not the average, the actual. Include the label, the transfer, and the sequence table entry. Most labs underestimate this by 40%.
Load your instrument time at the fully burdened rate. Include the analyst, the instrument depreciation, the facility, and the QC review. A typical number is $35 to $55 per injection for a binary HPLC with UV detection. If you are running LC-MS, it is $80 to $140.
Apply the formula. Filter unit cost divided by filters per injection, plus prep labor, plus instrument time, plus rerun rate times per-injection cost, plus reject rate times re-sampling cost. The output is your cost per reported result. Track it monthly by filter grade.

For reference on instrument-side assumptions, Agilent publishes method and consumable guidance for its chromatography consumables and columns that helps calibrate the instrument time term, and Thermo Fisher’s chromatography and sample handling documentation covers sample prep time assumptions that align with what I see on the bench. Restek’s technical resources are useful for the extractables side of the calculation. For regulated methods, EPA SW-846 sets the QC acceptance criteria that determine whether a rerun is required or optional, which changes the rerun rate term directly.

The decision rule for procurement: buy the grade whose rerun rate keeps the per-test number flat

Here is the rule I give to every procurement manager who asks me whether to switch to a cheaper filter. Compute your current cost per reported result. Then compute it again with the candidate filter’s unit price and your best estimate of its rerun and reject rates, taken from a 200-injection trial, not from a datasheet. If the candidate’s per-test number is within 10% of your current number, buy the cheaper unit. If it is higher, do not switch. The 10% band absorbs measurement noise in your rerun rate estimate and leaves room for lot-to-lot variation.

Two operational parameters matter more than any other in that trial. First, run the trial on the same method, same analyst, same instrument, and same sample matrix. A filter trial on a different matrix is not a trial, it is a coincidence. Second, log the puncture count and the hold-up volume on every unit. If the candidate’s hold-up volume exceeds 35 µL on a 13 mm device, the transfer loss alone will cost you more than the price difference. If its puncture count is under 30, you will see coring within the first 200 injections.

One more thing about the per-test number that most labs miss: the filter is not the only consumable in the prep path. If you are running COD or similar colorimetric methods alongside your HPLC work, the vial and the insert each carry their own per-test contribution. A COD vial at $1.20 with a 6% reject rate contributes $1.27 per reported result before you add any labor. The same logic applies. Build the per-test number for every consumable in the path, not just the filter, and then decide where the money actually goes.

A quick troubleshooting checklist before you blame the filter

When the per-test number spikes, work through these five checks in order. Each one has a specific measurement.

Blank filtrate at your detection wavelength. If the first 3 minutes show more than your method’s baseline noise, the filter is contributing extractables. Compare against a Grade A blank.
Filter weight before and after a 2 mL filtration at 4 bar. A gain over 5 mg means the housing is leaking filtrate around the membrane. That is a reject, not a rerun.
Septum inspection at 10x every 20 punctures. A visible core before 30 punctures means the septum grade is wrong for your workflow.
Retained volume on a 500 µL aliquot. More than 100 µL retained points to a pore size distribution wider than specified.
Storage temperature log for the filter cabinet. Sustained exposure above 25 °C raises extractables two to three times and shows up as a seasonal rerun pattern.

The decision rule is simple: a filter is worth its price only if its rerun rate keeps the cost per reported result within 10% of the incumbent. Below that band, buy on price. Above it, buy on performance. The invoice tells you nothing about which side of the band you are on. The rerun rate does.

References & further reading

Related resources from our lab consumables network:

External references and standards: