A stateable constraint is defined as a limit an advertiser can express as a single number that holds true across the whole account. Manual bidding vs cost caps vs bid caps is really a question about whether you have one: a cost cap constrains an average, a bid cap constrains every auction, and manual bidding constrains nothing except what you remember to adjust. Around 70% of carts are abandoned on average, so the conversion volume a cap needs in order to work is scarcer than traffic figures imply [Baymard Institute, 2025]. This guide covers what each strategy actually constrains, why a cap stops spending, and how to pick the number.
- Set the cap from contribution margin, never from your trailing cost per acquisition.
- Use a cost cap when an average holds and a bid cap when no single result may exceed a ceiling.
- Read underspend as the system obeying a cap below market, not as a fault.
- Move a cap in two or three deliberate steps, leaving each long enough to clear learning.
- Avoid capped strategies in a low-volume account, which cannot feed them.
- Keep manual bidding for a hard ceiling or a deliberately thin slice of cheap inventory.
Omniconvert has measured store performance across the CROBenchmark dataset of 7,000+ websites in 15+ industries, against 248+ audit criteria, over 13 years in eCommerce [CROBenchmark Report 2026, Omniconvert]. The pattern relevant here is that accounts in trouble with bid strategy are usually not using the wrong one. They are using a number nobody derived, taken from last quarter's average, and then treating the resulting underdelivery as a platform problem.
Manual bidding vs cost caps vs bid caps: what each one constrains
The distinction matters because it decides what underdelivery means. Under a cost cap, expensive auctions are permitted provided cheap ones balance them, so the campaign can keep buying through a costly week. Under a bid cap, anything above the number is simply never bid, so a costly week means a quiet week.
Manual bidding sits outside both. You name a bid and the system spends against it, with no commitment about what a result costs. It was the standard arrangement for years and it is now the option that asks most of you and offers least in return, because you are setting a number against an auction you cannot observe.
Picking between them is therefore a question about your own business rather than about the platform. The useful form of that question is: can you state a single number that is true across the account? If you can, a cap will enforce it. If you cannot, a cap will enforce a number you made up. The campaign-level version of this decision is in ABO vs CBO.
| Dimension | Manual bidding | Cost cap | Bid cap |
|---|---|---|---|
| What it constrains | Your bid | The average | Every auction |
| Expensive auctions | Possible | Allowed, if balanced | Never entered |
| Volume available | Moderate | Higher | Lowest |
| Behaviour in a costly week | Overspends | Keeps buying | Goes quiet |
| Conversion volume needed | Low | High | High |
| Where the number comes from | Guesswork | Contribution margin | Hard ceiling |
| Best case for it | Thin cheap inventory | Most scaling accounts | An unbreachable limit |
Why a cap stops spending, and why that is correct
This is the single most common misreading in paid social, and it produces a predictable sequence: the cap is set, delivery falls, the account is declared broken, and somebody reverts to lowest cost in a hurry. The cap was working. It was set below what the auction costs.
Meta's own campaign-budget guidance has long made the same point about consolidation and learning: a strategy needs enough events inside one structure to find a stable answer [Meta, 2026]. A tight cap spread across many small ad sets starves each one twice over, once on signal and once on eligible inventory.
So the diagnostic question when a capped campaign goes quiet is not what is wrong with the platform. It is whether you would rather have the number or the volume, because at that moment you cannot have both. That is a commercial decision and it belongs with whoever owns the margin.
The way teams get this wrong is by nudging the cap. Delivery softens, somebody raises the cap slightly on Tuesday, it softens again, they raise it again on Thursday, and the account spends the next three weeks permanently in learning and never stabilising. The chart looks like a bidding problem and it is an operator problem. The fix is unglamorous: decide the ceiling from margin, move toward it in two or three steps, and leave each step alone for long enough that the result means something.
Where the cap number actually comes from
Derive it once and write it down. Take the gross margin on a representative order, subtract real delivery cost, the blended discount rate and the expected return rate, and what remains is the most a conversion can cost before the order stops being worth having. That figure is defensible in a meeting, which your trailing average is not.
Two refinements are worth the effort. First, work it out for a first order and for a repeat order separately, because the two can afford very different numbers and treating them as one is how acquisition gets underfunded. The retention economics are well established: a 5% retention lift can raise profits by 25% to 95% [Harvard Business Review, 2014].
Second, check the figure against your category rather than only against your own history. A free leaderboard score benchmarks your store against real competitors in your category and country across six dimensions: Creative and Ads, Reviews and UGC, AI Visibility, Agentic Commerce, Competitor Synthesis and CRO. AI Visibility and Agentic Commerce are the two nobody else scores. For what rivals are running at what apparent commitment, start with Meta's own Ad Library.
AliveCor used Omniconvert to run a structured A/B testing programme and achieved +21% conversion rate, +5% revenue per visitor, and 94% statistical relevance across their experiments. [Omniconvert, AliveCor case study] That is an on-site testing result rather than a bidding one, and the transferable discipline is the statistical relevance: a cap change read before learning has cleared is not a result, whichever way it points.
Which strategy suits which account
Low weekly conversion volume rules out both capped strategies, whatever the pricing theory says. The system needs enough events to find the distribution it is capping, and without them a cap produces erratic delivery that is indistinguishable from a bad audience. Lowest cost, consolidated into fewer campaigns, is the honest position until volume exists.
With volume, the choice follows the constraint. An average you can live with points at a cost cap, which is most scaling accounts. A limit no single conversion may breach, usually because of a contractual or inventory commitment, points at a bid cap and you accept the lost volume as the price of certainty.
Manual bidding survives in a narrow band: a thin slice of deliberately cheap inventory where you accept low volume, or a situation where you are testing a market with no intention of scaling it. Treating it as the default in 2026 means competing against automated bidding with less information than it has. The structural question that sits above all of this is in broad targeting vs interest stacks in 2026, and what makes a concept worth bidding on at all is in hook, angle, CTA.
Downstream of the bid sits the question of whether the order was worth winning. Omniconvert Commander (formerly Nexus) is an AI for eCommerce growth engine that unifies commerce data, prioritises experiments by True Profit, and generates campaigns and creative you approve before they go live.
Frequently asked questions
What is the difference between a cost cap and a bid cap?
A cost cap constrains the average cost per result across the campaign, so individual auctions may cost more or less as long as the average holds. A bid cap constrains every single auction, so nothing above the number is ever bid. The first buys flexibility, the second buys certainty, and the second delivers less volume for it.
Is manual bidding still worth using?
Rarely as a default, and genuinely in two cases: a hard commercial ceiling you cannot exceed on any single conversion, and an account where you are deliberately buying a thin slice of cheap inventory and accept low volume. Outside those, you are competing against automated bidding with less information than it has.
Why has my cost cap stopped spending?
Almost always because the cap is below what the auction currently costs for your objective. A cap is a constraint rather than a target, so if nothing can be bought at that price the campaign buys nothing. Underspend is the system obeying you, not failing, and the fix is to decide whether the number or the volume matters more.
Should I raise a cap gradually or in one step?
In a few deliberate steps, with enough time between them to clear learning. Small frequent adjustments restart learning repeatedly and produce an account that never stabilises, which then gets diagnosed as a bidding problem. Decide the ceiling, move toward it in two or three moves, and leave each one alone long enough to read.
Which strategy suits a low-volume account?
Usually none of the capped ones. A cap needs enough weekly conversions for the system to find the distribution it is capping, and a low-volume account cannot supply that. Lowest cost with a sensible budget, consolidated into fewer campaigns, is the honest starting point until volume exists.
How do I know the cap number to use?
From your own contribution margin, not from your current cost per acquisition. Work out what a conversion can cost before the order stops being worth having, after delivery, discount and returns. That figure is a constraint you can defend; your trailing average is just a description of what happened.
Derive the number before you choose the strategy
Work out, once and in writing, the most a conversion can cost before the order stops being worth having. Gross margin on a representative order, less real delivery cost, less blended discount, less expected returns. Do it separately for a first order and a repeat order.
Those two figures tell you which strategy you are entitled to. If they sit comfortably above what the auction currently charges, a cost cap will hold and you can scale into it. If they sit below, no bid strategy will fix that, and the work belongs on margin, delivery cost or conversion rate rather than in the bidding panel. The accounts that keep changing strategy are usually the ones that never did this arithmetic.