When Smart Bidding spikes your CPC and it starts to feel like a scam
You flip a Search campaign to Maximize Conversions or tCPA. For a few days nothing dramatic happens. Then one conversion lands, and the next clicks jump to a CPC that looks nothing like your recent average. Spend piles onto a single keyword. A few of those clicks never show a useful search term. The Learning phase status keeps blinking while your gut says the Auction is being gamed.
That reaction makes sense when you have been living on Manual CPC for years. Smart Bidding is an auction participant with a wide bid range, not a polite assistant that stays near your old Manual CPC. When conversion data is thin at the campaign level, the model reaches for expensive inventory because it has not yet learned which cheaper queries actually close. The spike is usually a starved model overbidding, not a secret fee that Google only charges after you convert.
Treat Smart Bidding like a bidder that has to earn a floor of proof
Manual CPC lets you set the price of each click on purpose. Smart Bidding sets prices from patterns in the conversion data that campaign has already produced. If that campaign is still under roughly twenty to thirty conversions in the last thirty days, those patterns are guesses. Maximize Conversions will still enter the Auction. It will just do it with a noisy map of what a win looks like.
That is why brand terms and non-brand terms deserve different treatment. On brand, Manual CPC with a high impression share goal usually keeps CPC stable because you already know the economics. On non-brand, you can pre-train with Manual CPC, or with Maximize Clicks aimed at a cheap micro-conversion that marks real intent, until the campaign holds enough conversion data to justify a switch. Skipping that floor and jumping straight into tCPA is how accounts teach themselves that automation equals pain.
Portfolio bidding helps when one campaign alone cannot hit the floor. Pool related campaigns under one strategy so the model sees a denser stream of outcomes. Add a Max CPC cap when runaway click prices are the failure mode you fear most. Caps trade some volume for sanity. That trade is often worth it on small budgets while the Learning phase is still messy.
What to do when CPC jumps after a conversion
Do not rip the strategy off the same afternoon. First check whether the campaign ever cleared the data floor before you switched. Second, open the keyword and search terms view and pause bleed terms that are eating budget without conversions. Smart Bidding will keep adjusting bids on keywords that stay eligible, and it will not permanently retire a keyword that keeps winning auctions while failing to convert. You still have to manage that inventory.
Third, look at Max CPC and portfolio settings. If Maximize Conversions can bid without an upper bound, a single high-signal Auction can absorb a shocking share of the daily budget. A portfolio Max CPC will not make the model brilliant overnight, but it stops one click from rewriting your week. Fourth, confirm that the conversion you are optimizing toward is the outcome the business cares about. If the tag fires on every soft form fill, Smart Bidding will hunt more of those fills at whatever CPC the Auction demands.
If you are mid-switch and the CPC chart already looks hostile, use the Bid Strategy Switch Checklist. It is a readiness table, day-one target math, and a week-by-week recovery chart for moving from Manual CPC into Maximize Conversions, tCPA, or Target ROAS without treating every spike as proof that automation is a scam.
A switch order that keeps control without freezing growth
Run this sequence before you decide Smart Bidding is broken forever:
1. Earn the floor on Manual CPC first
Stay on Manual CPC until the campaign, not the whole account, is producing about twenty to thirty conversions across thirty days. If true leads are rare, add a Maximize Clicks phase on a micro-conversion that correlates with sales conversations, then graduate once that signal is steady.
2. Cap the Auction while the model learns
Move into Portfolio bidding with a Max CPC that sits above your healthy Manual CPC average but below the nightmare spikes you refuse to fund. You should expect a Learning phase after the switch, so do not stack five other account changes in the same week.
3. Keep brand on Manual CPC when Smart Bidding inflates it
If branded CPC climbs for no competitive reason after automation, pull brand back to Manual CPC and leave non-brand on the automated strategy. That split is normal, not a moral failure.
4. Pause bleed keywords on a schedule
Once a week, cut keywords and search terms that spend hard with no conversions. Smart Bidding will keep testing them if you leave them eligible. Your job is still to shrink the junk the model is allowed to bid on.
5. Diagnose spikes as data or constraint failures
Ask whether conversion data was too thin, whether Max CPC was missing, whether the conversion action is too soft, or whether one keyword is absorbing the budget. Those are fixable constraints that you can check in an afternoon. Flipping back to Manual CPC in panic, then flipping again next month, keeps resetting the Learning phase and teaches you nothing.
What a healthy switch looks like afterward
Healthy Smart Bidding does not mean every CPC matches your old Manual CPC average. It means average CPC and CPA settle inside a band the business can live with, spend spreads across converting queries instead of one runaway keyword, and the Learning phase eventually clears without you rebuilding the campaign every Friday. You still review search terms, you still protect brand where automation inflates CPC, and you still refuse to optimize toward junk conversion data.
If the spikes continue after the floor, the cap, and the keyword hygiene are in place, then revisit the strategy. Until then, treat Smart Bidding as an aggressive bidder that overpays when you starve it. Feed it campaign-level conversion data, constrain the Auction with Max CPC when you need a guardrail, and switch only when the checklist says the account is ready.