Which robot should you buy? A use-case guide for warehouse operators

Operators ask us one question more than any other: which robot should we buy? Here is our honest answer, in two parts. First, what the market data says about each type of robot. Second, three rules to use before you sign anything, drawn from our own testing. One thing up front, so you can judge the rest: we test robot control software, not commercial robot fleets. We have scored 60 trials on an open benchmark ourselves. We also host a third-party 50-run policy evaluation on cable insertion, but its source data records no outcomes, so none of it is scored by us. The robot types and prices below come from sourced market research, not our lab. When a number is an estimate, we say so.

The short answer, by use case

Find your job below. Each card shows the robot type, what it costs, how proven it is, and the one thing to demand before you buy. Prices are market estimates [1]. No robot vendor publishes a price list.

Two more facts to hold while you shop. In a July 2024 survey of 300 robot buyers, 80% expected payback within three years, though the same firm reported in 2026 that 61% now want it inside 12 months [9]. And average warehouse pay hit $26.66 an hour in May 2026, up 4.7% in a year [10]. That wage math is real. It is also in every sales deck. That is why the rules below matter more than which robot you pick.

Three rules before you sign

Rule 1: set the pass mark before the test, and use the standard one. Our own evaluation shows why both halves matter. We fixed our bar before scoring, which was right, but we picked a looser bar than the benchmark's, and that alone turned a dataset where nothing clears the standard into a 48.1% headline. We corrected it in public. If the bar gets set after everyone watches the robot run, or set lower than the accepted one, the number bends against you. So write pick rates, uptime, and how often a person must step in into the pilot contract first.

Rule 2: record every trial, and check that your detector actually works. Ours did not. It fired on 8 of 27 failing trials, and on 10 of 25 passing ones, so it flagged good runs more often than bad ones. We published the correction rather than the original finding. The lesson stands even though our example broke: compare the hit rate against the false-alarm rate, or you cannot tell a detector from a coin. The full study and the correction are in Evaluation #1 on this site.

Rule 3: ask for error rates, including the dashboard's own. Our own alarm system flags 40% of passing trials as bad, and we published that number along with the fact that it is measured in-sample. A dashboard that catches every problem and quotes no false-alarm rate is marketing, not monitoring.

Check the vendor's health

Robot companies are failing while demand grows. In the last 14 months:

• Zebra paid $301M for Fetch in 2021, then wound it down and sold the business in March 2026 for $20M of total consideration, about a 93% write-down [11]
• Attabotics filed for bankruptcy protection in July 2025, after raising over $165M [12]
• One tracker counted about $16M across two disclosed AMR rounds in early 2026, even as robotics venture funding overall hit a record [13]
• Geek+ became the first listed AMR company [14]

So protect yourself in the contract:

• Ask for code and spare-parts escrow, so you keep access if the vendor folds
• Add exit and continued-support clauses
• Prefer robot types where several vendors can replace each other
• Read the vendor's balance sheet as closely as you watch the demo

What this guide is not

It is not a vendor ranking. The real answer depends on your building, your products, and your software. It comes from a test in your own warehouse, with the pass mark set first. Running that test is the service we sell. The three rules are free, and they work without us.

One last point from the deployment data: software integration and change management come up far more often than robot performance when buyers explain unmet expectations [4][15]. One industry review of DHL's 2025 survey put it plainly: robots 'rarely meet vendor-quoted throughput in real conditions' [15]. Buy the robot your team and systems can absorb, not the one with the best demo reel.

Sources

1. Industry pricing estimates, labeled as such (no vendor publishes primary price lists): PickTheRobot 2026 cost guide; Robotomated budget guide; RaaS per-robot figures via industry aggregators.
2. The Robot Report, Locus Robotics surpasses 6 billion robot-assisted picks, Oct 2025 (picks made by people working alongside robots, not autonomous picks)
3. AutoStore Q4 2025 results, Feb 12 2026: about 1,900 systems in 65 countries; FY2025 revenue $538.6M and order backlog $557.0M (the two are different figures): news.cision.com/autostore-as
4. PwC 2026 Digital Trends in Operations Survey (767 US leaders; 89% say tech investments have not fully delivered; integration complexity a leading cause): pwc.com
5. ASTM Committee F45 publishes performance test methods for driverless industrial vehicles (NIST chairs it): nist.gov. The safety standard is ANSI/ITSDF B56.5 (2024), with ISO 3691-4 internationally
6. Boston Dynamics Stretch first commercial deployment at DHL, Feb 2023 (dhl.com press archive); Locus Array launched Apr 10 2026 (Business Wire); Locus acquired Nexera Robotics May 19 2026 (separate event)
7. Robotics & Automation News, Digit passes 100,000 totes at GXO, Nov 24 2025. Fleet size has never been disclosed by GXO or Agility; the agreement dates to June 2024, so the milestone is about 17 months in
8. Gartner press release, Jan 21 2026: through 2028, fewer than 20 companies will go live in production for supply chain and manufacturing use cases (this counts adopting companies, not vendors): gartner.com/en/newsroom
9. Interact Analysis Mobile Robots Buyer Survey, published Jul 27, 2024 (300 buyers), therobotreport.com/what-do-customers-expect-from-mobile-robots. Superseded in part by Interact Analysis Voice of Market, May 2026: 61% now expect payback within 12 months
10. US Bureau of Labor Statistics, series CES4349300003 (NAICS 493 average hourly earnings, all employees): $25.46 in May 2025 and $26.66 in May 2026, a 4.7% rise. Note this is the all-employee average; production and nonsupervisory workers averaged $25.92
11. Zebra Technologies SEC filings (CIK 877212): FY2021 10-K, Fetch acquired Aug 9 2021 for total consideration of $301M ($290M cash plus $11M for an existing minority interest); FY2025 10-K, decision to exit the robotics business in Q4 2025; Q1 2026 10-Q, sale to Skild AI completed Mar 27 2026 for $20M total consideration, announced Apr 15 2026
12. Global News and The Globe and Mail, Attabotics filed a Notice of Intention under Canada's Bankruptcy and Insolvency Act, Jul 2 2025 (a restructuring proceeding, not a bankruptcy assignment)
13. New Market Pitch AMR funding tracker (aggregator, treat as directional), 2026
14. TMTPost, Geek+ lists on HKEX main board, Jul 2025
15. DHL Supply Chain Insight 2030 survey, Nov 2025: only 34% of VP-level leaders fully satisfied with warehouse robotics deployments: warehouseautomation.ca/news/dhl-report