Why few conversions are a bidding problem
Every Smart Bidding strategy works the same way at heart: for each auction it predicts the chance of a conversion, and its value, from the signals available, and bids accordingly. The predictions come from what the account has recorded. Google's conversion tracking documentation describes conversions as the actions a business values after someone interacts with an ad, recorded by the account's tracking, and those recorded actions are the whole of what the strategy learns from.
The Smart Bidding strategies guide states the consequence: a model with many recent conversions predicts better than one with few, and a campaign whose conversions are rare and irregular will hold a target loosely. Its status column can show a limited status for low conversion data, and the guide's table lists the responses: broaden the conversion action, pool campaigns, or run without a target longer.
Few conversions are not the same as a small budget. A kitchen installer, a commercial cleaner or a stove fitter can spend a healthy sum and still record a handful of enquiries a month, because each one is valuable and slow to arrive. The same thinness affects the advertiser as much as the model: a cost per acquisition (CPACost of each conversion.) built on a few conversions swings widely from month to month by chance, so any judgement made on it is as shaky as the strategy's. The guide to reading small numbers explains why rates built on small counts jump around.
| Option | What it does for the data | When it fits | What it costs |
|---|---|---|---|
| Consolidate campaigns | Pools conversions that were split for no control reason | Several campaigns share a goal, budget logic and target | Some reporting granularity |
| Bid on a more frequent action | Gives the strategy more events to learn from | A step before the sale predicts the sale well | Risk of optimising to the wrong step |
| Portfolio strategy | Pools data across campaigns kept separate for other reasons | Campaigns share a goal and a margin | One shared target for all of them |
| Maximise conversions without a target | Uses a thin signal without asking it to hold an average | Reliable tracking, few conversions | Less control over cost per conversion |
| Manual or capped click bidding | Does not rely on conversion learning at all | Too few or untrustworthy conversions | No auction-time prediction |
Consolidate before you automate
The first remedy costs nothing but tidiness. Many low-volume accounts are low-volume per campaign rather than in total, because the conversions are split across campaigns created by theme, by location or by habit. Each campaign's strategy then learns from its own small share.
The account structure guide sets the rule: under Smart Bidding, campaigns should be as few as the control needs allow, and what remains of granular manual-era structures is fragmentation, with ad groups and campaigns each holding a handful of conversions a month and too thin to tell whether a change helped. Its test for a campaign split is a control need: its own budget, its own target or strategy, its own location or language settings, a different conversion goal, or reporting only the campaign level can give. A split by keyword theme alone belongs at ad group level. The account audit counts campaigns with too few conversions for their bid strategy to learn, which is the list to start from.
Consolidation is not one campaign for everything. Keep separate campaigns where the economics differ, such as a high-margin service beside a low-margin one, or where a budget must be guaranteed. Merge the rest, keeping the themes as ad groups so that ads and landing pages stay relevant. The academy exercise on steering bidding through a sale and an outage for a garden building maker practises deciding which thin quotation campaigns to merge and which action the strategy should bid towards.
A more frequent action as the bidding signal
When even the consolidated campaign records too few conversions, the next lever is what counts as one. The Smart Bidding guide names the option: optimise towards a more frequent action that correlates with the real goal. For a lead business, that usually means a step earlier in the sales process than the sale itself.
The Google Ads setup for lead generation gives the practical version. Closed deals are the truest signal but often too rare to bid on, so import an earlier stage, such as a qualified or contacted lead, as the primary action, and keep closed deals as a secondary action for reporting until volume grows. The offline conversion import entry explains how outcomes recorded outside the website are uploaded against the original click, which is what lets the real outcome stay in view.
The danger is choosing a step that is frequent but uninformative. The measurement hub warns that a primary action with a high conversion rate and little value will be bought in quantity; the strategy treats whatever it is told to count as worth chasing. Three checks keep the proxy honest. Pick the step closest to the sale that still occurs often enough to help. Check, from the sales records, how often that step becomes a customer, and whether the rate is similar across campaigns and keywords; a step that converts to sales well from one campaign and badly from another will steer money the wrong way. And keep reporting the real outcome beside it, so a drift in quality shows before it costs a quarter.
Pooling campaigns in a portfolio
Some campaigns have to stay separate, for budget, location or reporting reasons, yet share a goal. A portfolio bid strategy applies one strategy and one target across them. The Smart Bidding guide describes the pooled data as the main benefit: several campaigns with modest conversion volume each learn better as one portfolio than as several separate strategies.
The price is the shared target. The portfolio allocates spend across its campaigns to hit the target overall, so one campaign can run above it while another runs below. That is right when the campaigns are economically equivalent and wrong when they are not. Group campaigns into a portfolio when they share a goal and a margin; keep them apart when they do not, and accept thinner data as the cost of a correct target. Bid limits on a portfolio are documented and generally discouraged, for the reasons the Smart Bidding guide gives.
Strategies that need less data
The choice of Bid StrategyThe rule an ad platform follows to set the bid in each auction. matters more when data is thin, because strategies differ in what they ask of it.
Maximise conversions without a target asks least of a conversion-based strategy: spend the budget to get as many conversions as it can, without holding an average. The comparison of maximise conversions and target CPA notes that it works with thin history, while a target needs enough volume to hold an average reliably, and recommends maximise conversions, or pooling campaigns, before adding a target when conversions are very few.
Manual CPC, or maximise clicks with a maximum cost per click, does not rely on conversion learning at all. The comparison of manual and automated bidding lists accounts with only a few conversions in a typical month among the cases for manual bidding or an automated clicks goal with caps, and keeps manual control for conversion tracking that is broken or under review. The Smart Bidding guide describes maximise clicks with a maximum CPC as a common bridge to Smart Bidding, because it gathers data faster than manual bidding on a new campaign.
The order that usually works is: consolidate, choose the bidding signal, then pick the least demanding strategy the data supports, and move up only when the evidence does. A product on its launch day is the extreme case, with no conversions of its own at all, and the guide to paid search for a product launch applies these strategies to the launch window and says when a target can be added.
Targets on thin data
A target is a promise the strategy tries to keep on average, and on thin data the average is hard to keep and hard to read. The Smart Bidding guide's approach applies with extra force. Run without a target first, let the campaign show what it achieves at the budget, then add a target near the observed figure and tighten it in steps. Starting with an aggressive target produces very little spend, because the model finds few auctions it believes will meet it.
Thin data changes three details. Compute the observed figure over a longer window, so it rests on enough conversions to be stable. Set the first target generously, near or slightly above the observed cost, and keep it within the economic ceiling the CPA calculator gives from customer value and conversion rate. And hold it: the Smart Bidding guide lists large target changes among the triggers for a learning period, and on thin data each change takes longer to read. A target that is changed every few weeks in a low-volume campaign is never actually tested.
Judging results over longer windows
The advertiser's patience matters as much as the strategy's. Low-volume campaigns need longer windows for every judgement: whether a strategy works, whether a target is right, whether consolidation helped. Weeks are too short; in most low-volume accounts the honest unit is a month at least, and often a quarter.
Two effects make short windows worse. Conversion LagThe time between an ad interaction and the conversion it leads to, which leaves the most recent reporting days incomplete. is usually long in these accounts, because valuable purchases are considered slowly and imported outcomes arrive weeks after the click, so recent periods are always incomplete. And Sample SizeThe number of users or sessions an experiment needs to detect a given effect reliably. is small, so a difference between two periods can be chance. Read closed periods, compare like with like, including the same season a year earlier where demand is seasonal, and judge changes at account level rather than campaign by campaign, since consolidation moves conversions between campaigns. When a strategy change does seem to have made things worse, the results dropped after a bid strategy change diagnosis lists too few conversions for the strategy to learn from among its causes and sets out how to tell it from the others. The bidding hub sets out how targets derive from margin, which is the standard every reading is judged against.
A stove installer with a few leads a month (illustrative)
A stove installer gives its bidding more to learn from
Cindermint Stoves, an invented installer of wood-burning stoves, runs three search campaigns, one each for wood burner installation, multi-fuel stoves and log burner fitting, each on target CPA with its own budget. Together they record a few quote requests a month, and all three show a limited status for low conversion data. The site also offers a free home survey, booked online; the business's records show that most installations start with a survey, and that about one survey in five becomes a sale.
The plan has three parts. The three campaigns become one, with the themes kept as ad groups. Survey bookings and quote requests both become primary conversion actions, counted once per click, and installations sold are imported from the job records as a secondary action. The merged campaign runs on maximise conversions without a target for the first quarter.
| Measure | Quarter before | Quarter after |
|---|---|---|
| Campaigns | 3 | 1 |
| Spend | £4,200 | £4,300 |
| Primary conversions the strategy saw | 11 quote requests | 13 quote requests and 41 survey bookings |
| Installations sold, from job records | 4 | 7 |
| Cost per installation sold | £1,050 | £614 |
Because stove demand rises into autumn, the team also checks the same quarter a year earlier, when the old structure sold five installations. The improvement holds against both baselines, so the next quarter adds a target CPA of £85 per primary conversion, near the £80 the campaign achieved, and leaves it alone.
Common mistakes
- Automating before consolidating. Several thin campaigns learn less than one campaign with their combined data.
- Bidding on a frequent action that does not predict sales. The strategy buys cheap, useless conversions.
- Dropping the real outcome from view. Import it and report it beside the proxy.
- Pooling unlike campaigns in a portfolio. One shared target, wrong for every member.
- An aggressive target on thin data. The campaign spends little and teaches nothing.
- Changing the target every few weeks. On low volume, no change is ever given time to be read.
- Judging on a week. Conversion lag and small counts make short windows noise.
