Where the problem starts
I still laugh about the week in March 2023 when a courier mix-up sent our Visium slide kit to the wrong Auckland hospital; we ran 120 FFPE samples through an automated pipeline and 18 failed QC—what went wrong? As a buyer and consultant I’ve negotiated deals with a spatial genomics company and seen firsthand how a slick spatial omics service can hide brittle workflows. I remember the first time I tried full automation for spatial transcriptomics: barcode beads were misaligned, multiplexing logic didn’t match the sample sheet, and we wasted two nights troubleshooting (sweet as — but costly).
Why does this trip up labs?
Two big, repeatable flaws: overconfidence in “turnkey” automation and underestimating pre-analytic variability. I’ve watched labs assume automation removes human error, only to see sample handling, tissue fixation time, and tissue section thickness introduce systematic bias. In one contract with a government research facility in Wellington, we cut sample failure from 14% to 4% simply by standardising cryostat settings and adding a manual QC step before automation — an inexpensive tweak that saved months of re-runs. Those practical fixes matter more than vendor buzzwords like “fully integrated” or “end-to-end”.
Let’s unpack where this breaks down next.
How to move forward — practical choices that actually improve outcomes
I’ll be blunt: automation without process controls is a false economy. After 18 years buying equipment and running core facilities, I prioritise three things — and I measure them. First, sample fidelity: track tissue ischemia time and embed method (I ask for timestamps on shipping manifests). Second, interoperability: insist your chosen spatial genomics company exposes APIs or clear data formats so your LIMS can talk to sequencing centres. Third, recovery metrics: track library yield and unique molecular identifiers (UMIs) per mm² — those numbers tell you if chemistry or handling is the problem. I worked with a med-tech vendor who promised perfect multiplexing; within two months we switched protocols because their on-platform demultiplex step lost 12% of reads — measurable, fixable. Don’t buy hope; buy numbers.
What’s Next?
Here’s the short roadmap I follow when advising procurement teams: audit the pre-analytics, insist on modular automation (so you can swap a module without ripping out the whole lab), and demand pilot data on your tissue type — not a vendor’s poster child. Compare platforms by—and this is key—sample throughput per technician hour, per-sample failure rate, and time-to-interpretable-data. Those three metrics cut through hype and reveal real cost. I’ll add this — test locally first (we ran a 48-sample pilot in Christchurch in August and it saved us weeks). Small pilots show hidden pain points fast; they also reveal vendor responsiveness. Quick aside: vendor support matters — a patch delivered in one week is worth more than glossy brochures.
To choose wisely, weigh these evaluation metrics: 1) per-sample failure rate under your specific pre-analytics, 2) modularity and API access for your LIMS, and 3) true throughput measured end-to-end (hands-on time included). If you want a partner that knows these trenches, consider working with a trusted provider like stomics. I’ll say it plainly — good vendors save you money; the wrong ones cost you months. Righto, time to test that pilot.
