Bid Simulators in Google Ads 2026: Forecast Budget and Target Changes Before You Save
Every budget increase and every loosened target is a bet placed blind. Raise tCPA by 30% and you might get 40% more conversions — or the same conversions at 30% more cost. Add budget and you either capture auctions you were missing or pour money into auctions the algorithm was deliberately skipping. Both questions have a built-in answer in the interface: bid simulators, along with their budget and target counterparts.
It is a badly underused tool — most accounts either never open it or read it as a promise rather than a model. Here is what it actually computes, where it is systematically wrong, why it keeps vanishing from the interface, and how to turn its curve into a scaling decision.
What simulators actually compute
A simulator is a replay of auctions that already happened. Google takes the last 7 days of your campaign’s auction data and recalculates what would have happened at a different bid, target or budget. The output is a curve with estimated impressions, clicks, cost, conversions and conversion value at several values of that parameter.
The operative phrase is “would have happened”. This is not a forecast of the future; it is a reconstruction of the past under different inputs. Everything useful and everything limiting flows from that.
There are three distinct simulators and they should not be confused:
| Type | Variable | Where it lives | What it answers |
|---|---|---|---|
| Bid simulator | Maximum CPC | Max. CPC column on ad group and keyword pages | Manual and semi-manual bidding |
| Target simulator | Target CPA or target ROAS | Bid strategy column at campaign level | Smart Bidding: the volume-versus-efficiency trade-off |
| Budget simulator | Daily budget | Budget column on the Campaigns page | Whether you are budget-constrained and what the next slice of volume costs |
Shopping campaigns get an additional simulator on the product groups page. Supported campaign types include Search, Display, Shopping, App, Performance Max, Demand Gen and Hotel. The notable exception is Travel campaigns, where simulators are not available.
Reading the curve: three points, not one
The classic mistake is glancing at “+20% budget”, seeing more conversions, and calling it a decision. The curve is useful in full, because it shows the shape of the return, not a single delta.
- The “you are here” point. The simulator marks your current value. Everything else is read relative to it.
- The slope to the right. How steeply conversions grow as the parameter rises. A gentle slope means cheap growth and under-investment. A flat section means you have hit the ceiling and extra money buys nothing.
- The slope to the left. How much volume you lose by tightening. It is common to discover that cutting tCPA by 15% costs 40% of conversions — which changes the decision entirely.
Do the arithmetic on marginal values, not averages. Example: at a $100 budget you get 20 conversions, so $5 each. The simulator says $130 yields 24. The new average is $5.42, which looks fine. But the marginal cost of those four extra conversions is $30 / 4 = $7.50. That is the number the decision rests on: if your unit economics tolerate $7.50 per lead, scale; if your ceiling is $6, do not. How to derive that ceiling is covered in our piece on unit economics and LTV in media buying.
A simulator answers “what does the next conversion cost”, not “what is my average CPA”. Every scaling decision is made on the first number.
Why the simulator is missing from your interface
“I don’t have that icon” is the most common complaint, and the causes are a short, fixable list:
| Cause | What to do |
|---|---|
| Not enough historical data — new campaign or very low auction volume | Accumulate 7+ days of steady delivery; at very low volume it will not appear at all |
| The campaign uses a shared budget | Move to an individual budget if you need the forecast — shared budgets are not supported |
| The daily budget was hit during the last 7 days | Relieve the budget constraint first — auction data is truncated, so there is nothing to model |
| An experiment is running on the campaign | Wait for the experiment to finish |
| Product groups are subdivided by Item ID | Restructure the grouping — item-level simulation is not offered |
| Unsupported campaign type (Travel) | Use other estimation methods: experiments, geo tests |
The “budget was hit” case deserves emphasis. It is not a technicality — it is a finding. If the campaign spent its full daily cap for seven straight days, you have no idea how many auctions you never saw, and there is no data to simulate. Fix the budget constraint first, discuss targets second. Reverse that order and you are tuning a strategy that was never allowed to show its ceiling. Cross-campaign allocation is covered in our Performance Planner guide.
Where bid simulators are systematically wrong
1. They assume the world stood still
The model replays last week. If competitors raised bids, a season started, or your assortment changed, last week’s auctions no longer describe tomorrow’s. In volatile niches and ahead of peaks, treat the output as an order of magnitude only. To check whether auction pressure actually shifted, use impression share and auction insights.
2. They know nothing about the learning period
A simulator shows the steady state “after”, as if the strategy operated at the new target from day one. In reality there are 5–7 days of Smart Bidding learning in between, during which results are worse than either scenario. A 40% target jump will almost certainly produce a dip the curve never mentions.
3. They treat conversion lag differently by campaign type
An important technical detail: Performance Max and App simulators include expected future conversions, while simulators for other campaign types count only conversions recorded within the simulated period. With a long sales cycle, a Search campaign simulator will systematically understate results. By how much depends on your conversion window and conversion lag.
4. They vary one parameter at a time
The budget curve is drawn at your current target; the target curve is drawn at your current budget. Raise both at once and the two forecasts do not add up — they interfere. Change one variable, then measure.
5. They are blind to lead quality
Simulators count conversions exactly as your account counts them. If half your leads are junk, the curve will faithfully predict more junk. Before scaling, confirm the strategy is learning on the right signal — see primary and secondary conversions.
A working decision procedure
Step 1. Identify your binding constraint
Open the budget simulator. A steep right-hand slope plus regular budget capping means you are budget-constrained, and target talk is premature. A near-flat curve means money is not the limit — the constraint is your target, audience or creative.
Step 2. Compute the marginal cost
Take two adjacent points and calculate (cost₂ − cost₁) / (conversions₂ − conversions₁). Compare against your margin ceiling. Everything below the ceiling is growth headroom.
Step 3. Move in steps, not leaps
A workable rule of thumb: 10–20% per change, then a 7–14 day observation window (longer with a long sales cycle). Big jumps deliver a double hit — learning reset plus entry into unfamiliar auctions. Simulators are also more reliable near your current value: the further the point, the wider its real confidence interval.
Step 4. Record the change
One change at a time, with the date and the expected effect written down. Two weeks later you can compare forecast against actual and, incidentally, learn how accurate simulators are in your specific niche. To find and confirm the edit afterwards, use Google Ads change history.
Step 5. Prove it with an experiment when the stakes are high
A simulator is an estimate, not evidence. For decisions involving serious money, run a split experiment or a geo test — those produce causal answers rather than a replay of history.
Simulators and Performance Max
In PMax the target simulator matters more than anywhere else, because targets and budget are close to the only levers you hold. Two specifics:
- The PMax simulator incorporates expected future conversions, so it suffers less from lag than a Search simulator does. For exactly that reason its numbers are not directly comparable with Search figures — different methodology.
- tROAS curves in PMax are often steeper than expected: a 10–15% target reduction opens noticeably more auctions. That also means traffic composition shifts, so check the segment mix of conversions and not just the count.
Practical sequence: read the target simulator to learn the shape, move the target once by 10–15%, observe for a week, then iterate. Campaign-level fundamentals are in the complete Performance Max guide.
Simulator versus Performance Planner
| Simulators | Performance Planner | |
|---|---|---|
| Horizon | Replay of the last 7 days | Forecast for a future period (month, quarter) |
| Scope | One campaign, ad group or keyword | A portfolio of campaigns and the split between them |
| Seasonality | Not modelled | Uses historical seasonal patterns |
| Best question | “What happens if I raise the target 15% today?” | “How do I split next quarter’s budget across six campaigns?” |
Both are built on historical data and neither knows about your upcoming promotion or your competitors’ plans. Simulators win on tactical steps, the planner wins on budgeting. If a peak is coming, layer seasonality adjustments on top of both — neither tool knows your sale is happening.
Five situations where the simulator changes the decision
1. “The client wants more leads on the same budget”
Open the target simulator. If loosening tCPA by 15% adds 8% volume and loosening by 30% adds only 12%, the curve is effectively flat: there are no cheap incremental auctions and you would simply overpay for the same traffic. The conversation moves from “raise the target” to “expand geos, audiences or inventory”. This is one of the rare cases where a standard report actively protects the budget from a pointless edit.
2. “The campaign hits its daily cap every day”
Either the budget simulator is missing (data truncated by the cap) or it shows a steep right-hand slope. Both mean you are under-invested. Compute marginal cost at the first step up, and if it fits your margin, raise the budget 20% and observe for a week. Critically, do not move the target in the same week — otherwise you will never know which lever worked.
3. “We need to cut spend 25% with minimal damage”
The left half of the curve is the most underused planning tool in the platform. Run the budget simulator across all campaigns and find where cutting costs the fewest conversions. It frequently turns out that the campaign to cut is not the one with the worst CPA but the one with the flattest left-hand slope — where losing a quarter of the budget costs only single-digit volume.
4. “We want a higher tROAS to protect margin”
The simulator quantifies how much revenue you trade for that margin. A typical outcome: moving tROAS from 400% to 500% improves margin but removes a third of revenue. At that point it stops being an advertising decision and becomes a business-priority one — but it should be argued with numbers rather than instincts.
5. “We are launching something new and do not know the starting target”
Here the simulator is useless — there is no auction history yet. Use a different approach: start from actual conversion cost in comparable campaigns, let 2–3 weeks of data accumulate, and only then open the simulator. Reading a simulator on a brand-new campaign is reading an empty curve.
Putting simulators into a weekly routine
A one-off visit teaches you little. Value accumulates when you compare forecast to outcome and build a correction factor for your own niche. The routine costs 20–30 minutes a week:
- Monday, 10 minutes. Walk your top five campaigns by spend, note which ones have no simulator, and record why (budget-capped / experiment running / low volume). The budget-capped list is your scaling shortlist.
- Marginal cost pass. For each candidate, calculate the cost of the next slice of conversions and compare it to your margin ceiling. Keep that ceiling written down, or every decision becomes a judgement call.
- One change per campaign per cycle. A 10–20% step, with the date and the expected number logged.
- Reconcile after 7–14 days. Forecast versus actual. A gap larger than 1.5x is a prompt to check for parallel edits and shifts in auction pressure.
- Recalibrate quarterly. If the simulator is consistently 20–30% optimistic in your vertical, apply that correction as a standing rule. Far more useful than arguing with the tool.
This turns the simulator from a decorative chart into the basis for money decisions — and gives you a clean line to use with a client or a finance lead: not “let’s try raising the budget”, but “the next 30 conversions cost $7.50 each, our ceiling is $9, so we scale”.
Common mistakes
- Reading the curve as a guarantee. It is a model built on week-old data, not a commitment from Google.
- Thinking in averages. An average CPA hides the fact that the last conversions cost twice what the first ones did.
- Jumping to the far end of the curve. The further from your current value, the less reliable the estimate.
- Changing target and budget together. The forecasts do not add, and the result cannot be decomposed later.
- Ignoring “no data available”. A missing simulator is a diagnosis — budget-capped, experiment running, volume too low — not a UI glitch.
- Scaling on a bad signal. If the strategy learns on junk conversions, the simulator will accurately predict more junk.
FAQ
What data do Google Ads bid simulators use?
Auction data from your campaign over the last 7 days. The tool replays those auctions at a different bid, target or budget and reports what the outcome would have been.
Why is there no simulator icon in my account?
The usual causes: insufficient historical data, a shared budget, the daily budget being hit in the last 7 days, an active experiment, product groups subdivided by Item ID, or an unsupported campaign type.
How much should I trust the forecast?
As an order-of-magnitude guide for small steps. Accuracy is highest close to your current value and in stable niches; on large jumps or in seasonal verticals the deviation can be several-fold.
Do simulators work with Smart Bidding?
Yes. Target-based strategies get a target simulator showing the relationship between tCPA/tROAS, conversion volume and cost. It is the main way to evaluate the efficiency-versus-scale trade-off before touching a target.
Is there a simulator for Performance Max?
Yes, PMax is supported. Its distinguishing feature is that estimates include expected future conversions, whereas most other campaign types only count conversions recorded within the simulated window.
Does the simulator account for the learning period?
No. It shows an assumed steady state, not the transition. Budget 5–7 days of unstable metrics after any significant target or strategy change.
How big a target change is safe in one move?
A practical guideline is 10–20% per step with a 7–14 day observation window. This is not a platform rule; it is a way of avoiding a learning reset and an unfamiliar-auction shift at the same time.
Why is my budget simulator almost flat?
Because there is nowhere left to spend: you are already capturing nearly all available auctions at the current target. Look for growth in a looser target, wider audiences, more geos or more inventory — not in budget.
Can I use a simulator to justify cutting budget?
Yes, and it is the most underused application. The left side of the curve shows what you lose by cutting. It is common to find that a 20% budget reduction costs only 5% of conversions — a strong argument for reallocating that money.
How is a simulator different from Performance Planner?
A simulator replays the last 7 days for one entity and is built for tactical steps. Performance Planner forecasts a future period across a set of campaigns and is built for budgeting. Only the planner models seasonality.
Are simulators available in the API?
Simulator data can be pulled through the API for your own calculations, which is convenient if you keep reporting in BigQuery and want marginal cost per conversion computed automatically rather than by hand.
What if the actual result does not match the forecast?
Check three things: whether learning is still in progress, whether auction pressure changed, and whether you altered something else at the same time. If all three are clean, record a correction factor for your niche and apply it to future estimates.
The short version
Bid, budget and target simulators let you see the shape of the return before spending money to discover it. Read them by marginal cost rather than average, move in 10–20% steps, and remember the model is blind to the learning period, to seasonality and to the quality of your leads. A missing simulator is information in itself — usually that you are budget-capped and the target conversation is premature.
Related reading: tracing a performance drop through change history when a change goes wrong, and expanding reach with custom segments when the budget curve has flattened out. Stable billing and access for scaling live under agency Google Ads accounts; the rest of the stack is in PPC Rebels services.
