Conversion window: match the sales cycle or keep the 30-day default?

Say you run lead gen for a B2B software firm where the median time from first click to signed contract is about 45 days. Google's default click-through window is 30 days, and the conversion lag report will show plenty of conversions landing after it closes. The usual advice is to leave it at 30 so Smart Bidding gets fast feedback. I think that's backwards. If the window cuts off a big share of your real conversions, the algorithm is learning from a partial picture, and it'll favour keywords that close quickly over the ones that actually bring pipeline. I'd extend the window to cover most of your lag, then check the lag chart to confirm the extra days are real and not noise. Has anyone seen the learning period get noticeably longer after making that change?

Mostly agree, with one mechanical detail: the click-through window is set per conversion action, not per account, and it can go up to 90 days. You don't have to pick one number for everything. If the signed-contract action gets 90 days and a demo-booked action keeps 30, Smart Bidding can use the fast signal while the contract data still counts for the full cycle. Just be clear about which action the campaign optimises to, since that's the one whose window matters for bidding.

On learning periods, a longer window won't stretch them by itself. Learning is driven more by how many conversions come in and how much you changed. What a longer window does is push each conversion later into the data, so the algorithm waits longer to see whether a bid change worked. On a 45-day cycle that delay is a real cost, but it's a trade-off, not a reason to skip the change.

The conversion window is set per action, but whether an action drives bidding is a separate setting. Only primary actions feed Smart Bidding, so a 90-day contract action marked secondary will show up in reporting but won't change any bids. The trade-off Theo describes only applies if the long action is primary. On a 45-day cycle, I'd keep demo-booked as the primary action for bidding, then import signed contracts as offline conversions with their value, so they still get credited to the original click.

Sasha's split makes sense, but it hinges on how many demo-booked conversions you get per month. If demos are plentiful, optimising to them keeps the learning phase healthy. If they're thin, a 30-day primary action can starve Smart Bidding, and the offline contract data may be the better primary signal even with the delay.

For me it comes down to how well demo bookings predict signed contracts across your keywords, more than how many you get. If demos from one query close at twice the rate of another, optimising to demos will keep rewarding the cheaper, weaker leads, and more volume won't fix that.

Capturing the gclid on the demo form is what makes this checkable. If the lead record doesn't store the click ID at submission, you can't tie a signed contract back to the search term that brought the lead in, so per-query close rates stay a guess. Most CRMs can take a hidden field filled from the URL, so it's usually a small change, but it has to be live before the contracts start landing. Once it is, a simple export of demos and contracts grouped by search term will show whether that cheap query is actually weak or just slow to close. With a 45-day cycle, you want that data before you decide which action becomes primary.

The gclid is what makes this checkable, but it only works if your demo form can store it. Some hosted forms and CRMs strip URL parameters or can't hold a custom field, and fixing that can take a developer and a while.

If yours can't, enhanced conversions for leads is the fallback. You send hashed email or phone with the conversion and Google tries to match it to the original click. Match rates swing a lot by audience, and they depend on how many leads use the same email as their Google account, so you'll get a partial picture rather than a clean per-query one. Still better than guessing, but a low match rate shouldn't be what decides anything on a cheap query. And if the match rate is poor for one query and fine for another, that gap deserves a look before you decide which action goes primary.