Short answer: Choose the gripper from the parts outward. List every part the robot must handle — geometry, weight, surface, fragility — then match the gripping principle to the hardest parts on the list, plan how parts will be reoriented between pick and place, and only then open a catalogue. Our overview of gripper types covers the principles; this guide is the selection process for the two applications where most grippers are bought: bin picking and machine tending.
Step 1: Inventory the parts, honestly
The output of this step is a table, not an opinion: every part number, its dimensions and weight, the surfaces a gripper may touch, the surfaces it must not (finished faces, threads, seals), and how parts arrive — oriented on a tray, nested, or random in a bin. Random arrival is the dividing line: bin picking adds a perception and grasp-planning problem on top of the gripper, which the companion article how bin picking works covers as a system. Parts with delicate surfaces or wide size ranges will decide the gripper family; design for them, not for the average part.
Step 2: Match the principle to the parts
Parallel-jaw grippers with shaped fingers handle the rigid, prismatic majority of machined parts and give the most predictable grip. Vacuum excels on flat, sealed surfaces and large light parts, and fails instructively on porous or oily ones. Soft and adaptive fingers forgive geometry variation and delicate surfaces at the cost of precision. Magnetic gripping is superb on ferrous parts and forbidden near chips, swarf and anything magnetically sensitive. Established suppliers such as SCHUNK span several of these families in one catalogue, which helps when the part list refuses to fit a single principle — the directory's end-effector section is the shortlist source.
Step 3: Plan reorientation and the second grip
Machine tending has a geometry problem bin picking does not: the part often must change orientation between the raw stock and the chuck, and again after machining. Solutions, in rising order of hardware: a re-grip station (a simple fixture the robot sets the part into and re-grasps), shaped fingers that present the part differently, or a double gripper carrying raw and finished parts in one visit — which also halves door-open time. Decide this before the gripper purchase, because a double gripper is a different product from a single, and retrofitting one mid-project is a schedule event.
Step 3b: Budget the gripper's own weight and moment
A gripper is payload before it grips anything. Its mass, plus the part at full arm extension, must sit inside the robot's rated payload and wrist moment — with margin for the acceleration of the actual cycle, not the static case. Heavy grippers on small robots are a quiet cycle-time killer: the cell runs, but derated, forever. Check the gripper's weight from its datasheet (not its marketing page), add the heaviest part from your step-1 table, and verify against the arm's payload curve at your working reach. If the numbers are tight, the answer is a lighter gripper or a bigger robot — discovered now, on paper, rather than at commissioning.
Step 4: Fingers are a design item, not an accessory
For rigid parts, the fingers carry the engineering: contact geometry matched to the part, compliance where the bin demands it, materials matched to surface sensitivity, and a quick-change interface when several part families share the cell. Budget engineering time for fingers and expect two iterations. Suppliers will design fingers for you — supply them the full part table from step 1 and real sample parts, and judge their proposal by how it handles your three hardest parts.
Step 5: Trial on your parts, in your bin
Close the selection with a witnessed trial: your parts, your container, your lighting, a pre-agreed success rate over an agreed number of picks, including the known problem parts. Record failure modes, not just the rate — a gripper that fails by dropping is a different risk from one that fails by declining to pick. The trial result, with the part table, becomes the acceptance basis in the purchase documents.
Frequently asked questions
What information does a gripper supplier need from me?
The full part list with drawings or samples, weights, materials and surface restrictions, how parts are presented (tray, conveyor or random bin), the target cycle time, and the machine or process the robot serves. A supplier quoting from less than this is guessing with your project.
One gripper or two for machine tending?
A double gripper — one side for the raw part, one for the finished — usually pays in machine tending because it cuts door-open and idle time per cycle. It is a different, heavier product, so the decision belongs in the selection phase, not as a commissioning surprise.
Can one gripper handle all of our parts?
Occasionally, when parts share geometry and surfaces. More commonly the honest answer is one gripper body with interchangeable fingers per part family, or two grippers with a changeover plan. A selection process that starts from the full part list surfaces this early, while it is still cheap.
How should a gripper trial be organised?
Your parts, in your container, under your lighting, against a success rate and pick count agreed in advance. Include the parts you know are difficult. Record how failures happen, not only how often — the failure mode decides whether the risk is acceptable.