LinkedIn conversions: count view-through or click-only for bidding?

I keep going back and forth on the Insight Tag attribution setting for B2B lead gen. LinkedIn will credit a conversion to someone who only saw a sponsored post and converted later without clicking, and on long-cycle offers that can make the reported CPA look way better than it really is. My view is to bid on click-through conversions only, because that's the number you can defend when someone asks why spend went up. View-through can stay on as a secondary conversion so you can still see the assist effect, but it shouldn't drive the bid.

The counterargument is that a lot of B2B buyers see the ad on mobile, then search the brand on desktop days later, so click-only undercounts real influence. Fair, but I'd treat that as a reporting question, not a bidding one. Where do you land on it?

I wouldn't call this only a reporting question. On LinkedIn, whichever conversion action you mark as primary is what the bidding optimizes toward, so a reporting choice quietly becomes a bidding choice.

The practical problem with click-only on long-cycle B2B is volume. If you only get a handful of click-attributed leads a month, the algorithm has almost nothing to learn from, and you end up paying for noise. Before changing anything, check whether view-through is actually feeding the bid signal or just sitting in the report, because those two cases need different fixes.

Agreed on volume, but attribution windows matter about as much as click versus view. On LinkedIn both are set on the conversion action itself. Take a B2B software company with a 60-day sales cycle: if the view window is 7 days, most of the influence you're trying to measure falls outside it, so the view number looks tiny even when the campaign is doing its job. A long click window has the opposite problem, because it can credit leads that were really driven by something else. So before deciding what the bid should follow, I'd set the windows to match how long the buying decision actually takes.

I'd push back on tying the bid window to the full sales cycle. With a 60-day click window, the algorithm doesn't find out whether a click turned into a lead until two months later, so it keeps spending on signals that look fine for weeks before any feedback comes in. That loop is slow, and it hits the low-volume accounts Elena mentioned hardest. Keep the long windows for reporting, where you want the full picture, and use a shorter window on the conversion you actually bid on, even if it undercounts. Then reconcile the two in your monthly numbers.

Shortening the window doesn't fix this, because the delay comes from the event, not the window. If a lead only counts after a sales call, a 7-day window just gives you fewer conversions and the same wait on the downstream step. Bidding on an earlier event, like form submit or a qualified-lead flag, gets you faster feedback, and you can judge that event against pipeline later.

Undercounting isn't neutral either. A short window mostly keeps the fast deciders, who may not be your actual buyers, so the bid ends up learning the wrong profile.

Jun, I'm not sure a short window selects the wrong buyers. That only holds if fast converters differ from slow ones, and in B2B they often don't. Someone who fills a form in week one can easily be the person who signs in week six.

The real risk is the event itself. Say a software company's form fills come mostly from students or job hunters. The bid will get very good at finding them whatever the window is. I'd check lead quality by stage before deciding what the primary conversion should be.

Fair on the event mattering more than the window, and the student problem is real. I'd fix it at the form before touching the bid. Requiring a work email or adding a company size field cuts out a lot of job hunters, though it costs some volume, which ties back to Elena's point. Before settling on a primary event, I'd check whether the fills that reach opportunity stage look different from the rest. Do you have that split by stage, or is it still a guess?