What you’ll learn
- What is paid media forecasting?
- Why paid forecasts miss
- Plan for the learning phase
- The forecast formula, step by step
- Four models that survive real auctions
- Use the platform forecasts, then check them
A home-cleaning company in Singapore plans S$3,000 a month on Google Ads. The founder's spreadsheet says 30 jobs. Week one brings two enquiries, and by week three she is ready to switch the campaign off. Nothing was wrong with the campaign. The forecast took the steady-state average and applied it from day one, and it never priced in the weeks when the bidding system is still learning.
Paid media forecasts go wrong in a handful of predictable places. This guide covers where they break, the step-by-step formula we use with clients, a worked example in Singapore dollars, and the local calendar effects most templates leave out.
What is paid media forecasting?
Paid media forecasting is estimating what an ad budget will deliver before you spend it: how many people see the ads, how many click and enquire, and what that returns in sales. A useful forecast gives a range of outcomes, states its assumptions, and allows for the learning period at launch.
Sources: Neil Patel, paid media forecasting, Meta Business Help, about the learning phase, Google Ads Help, about Performance Planner.
Why paid forecasts miss
When a forecast misses, find which input drifted before you build the next one, or the same miss repeats. Most misses come from five places.
Cost per click rises
CPCs come out of an auction, so they move when competitors move. NP Digital's campaign data puts CPC inflation behind 54% of paid forecast failures, the single biggest cause. Singapore makes this sharper because many categories have only a handful of serious bidders. One new competitor with a bigger budget can lift your CPC for weeks.
Conversion rates swing with the market
Your landing page converts at a different rate when buyers are confident than when they're cautious, whatever the traffic quality. A flat conversion rate assumption is optimistic in a slow quarter and conservative in a boom.
Creative wears out
The same audience sees the same ad, click-through rates fall, and the effective cost per click climbs. If your forecast has no refresh cadence built in, it drifts more optimistic every week the campaign runs.
"Creative decay is not an edge case. It is a predictable curve."
Neil Patel, paid media forecasting
Automated bidding makes the calls
Smart Bidding on Google and Advantage+ on Meta optimise toward signals you don't fully control. Many teams plan to rescue a weak month by changing bids, but bid changes now move results less than they used to, because the algorithm makes most of the decisions.
The same buyer is counted twice
If you run Google and Meta at the same people, both platforms may claim the same enquiry. Add the two platforms' reported conversions together and your forecast assumes more customers than you'll actually get.
Need help with marketing? DMA builds and runs campaigns that grow Singapore businesses.
Free strategy call ›Plan for the learning phase
New campaigns usually perform worse at the start, and that's normal rather than a sign of failure. Google says a bid strategy can take a few conversion cycles, typically one or two, to calibrate, and that creating a strategy, changing its settings or adding and removing campaigns, ad groups or keywords puts it back into learning. Meta's guidance is that an ad set needs about 50 optimisation events in seven days to leave its learning phase. Big edits to budget, audience or creative restart it.
For a small Singapore account, 50 conversions a week on one ad set is a lot. Many local campaigns never get there on a purchase goal, which is why they stay in "learning limited" and cost more than the forecast assumed. Two practical fixes: optimise for an earlier step that happens more often, such as a WhatsApp click or form start, or put fewer ad sets on the same budget.
Build the forecast week by week. Show weeks one to three as the weak period and agree in advance that nobody pauses the campaign before week four unless tracking is broken. A campaign paused in week two never gets the data it needed to improve.
The forecast formula, step by step
Work through three questions in order, and let each answer feed the next: how many people will you reach, how many will act, and what will the business get back.
- Reach: budget divided by cost per click gives clicks. For awareness campaigns, budget divided by CPM and multiplied by 1,000 gives impressions.
- Efficiency: clicks multiplied by your landing page conversion rate gives enquiries. Then multiply by the share that become real quotes or qualified leads.
- Profit: qualified leads multiplied by your close rate gives sales, and sales multiplied by average order value gives revenue. Subtract ad spend and cost of delivery to see what you actually keep.
Pull your CPC from Google's Keyword Planner or your own account history. Global benchmark figures lean on US accounts and can be far off what a Singapore auction charges. Our Google Ads cost guide for Singapore explains what moves local CPCs, and the Google Ads benchmarks post gives starting ranges by industry if you have no history yet.
A worked example in Singapore dollars
Back to the cleaning company. These are illustrative inputs, not market rates: S$3,000 a month on search, an average job worth S$400, 40% of enquiries turning into quotes and half of quotes closing. Here's the same budget under three sets of assumptions.
| Line | Conservative | Expected | Aggressive |
|---|---|---|---|
| Monthly budget | S$3,000 | S$3,000 | S$3,000 |
| Cost per click | S$3.00 | S$2.50 | S$2.30 |
| Clicks | 1,000 | 1,200 | 1,304 |
| Enquiry rate | 5% | 6% | 7% |
| Enquiries | 50 | 72 | 91 |
| Quotes (40%) | 20 | 29 | 37 |
| Jobs won (50%) | 10 | 14 | 18 |
| Revenue at S$400 a job | S$4,000 | S$5,760 | S$7,300 |
| Ad cost per job | S$300 | S$208 | S$164 |
| Return on ad spend | 1.3x | 1.9x | 2.4x |
Two things jump out. The founder's 30 jobs was never on the table at this budget, even in the aggressive case. And the gap between conservative and aggressive is almost double, from S$4,000 to S$7,300, on the same spend. That range is the honest forecast. A single number would have hidden it.
Four models that survive real auctions
A spreadsheet like the one above is enough for most SMEs. Once you spend across several channels, four refinements make forecasts hold up better.
- Cohort forecasting groups customers by the month you acquired them, whenever they actually paid. It stops a recent campaign looking better than it is just because its customers haven't churned yet.
- Blended cost per acquisition divides total ad spend across all channels by total new customers. It cancels out the double counting when Google and Meta both claim the same sale.
- Incrementality adjustments reduce the conversions you credit to ads by the share that would have happened anyway. Branded search and retargeting are where this matters most.
- Spend elasticity maps how returns fall as you add budget. Doubling spend rarely doubles leads, because the extra money buys pricier clicks from less likely buyers.
If you're unsure how much of your Google traffic is really driven by ads, our guide to combining SEO and PPC covers how to test branded search without losing sales.
Use the platform forecasts, then check them
Google's Performance Planner simulates the relevant ad auctions from the last 7 to 10 days, including seasonality, competitor activity and landing pages, and shows forecast conversions and cost per acquisition at different budgets. Meta shows estimated audience size and results when you set up an ad set. Both are useful inputs. Neither knows your close rate, your margin or your quiet months.
Treat platform forecasts as the reach and efficiency steps of your own model, then apply your own sales numbers. When the platform's forecast and yours disagree by more than a third, find out why before you commit the budget.
Adjust for Singapore's calendar
Templates written for the US plan around Black Friday and Thanksgiving. A Singapore forecast needs its own calendar, and the dates of several festivals move every year. Use the same weeks from last year's account data where you have it.
| Period | When | What often changes | Forecast adjustment to test |
|---|---|---|---|
| Chinese New Year | January or February | Gift and food searches peak, many B2B and service searches drop for a week | Lower service lead volume in festival week; higher CPC on gift terms beforehand |
| Hari Raya Puasa | Moves each year | Demand for clothing, food and travel rises before the holiday | Shift retail budget into the two weeks before |
| June school holidays | Late May to June | Travel and family activity searches rise; tuition and enrichment shift | Model enrichment and travel separately |
| 9.9 to 12.12 sale days | September to December | Marketplace and e-commerce ad competition stacks up | Raise CPC assumptions for retail around each date |
| Black Friday and Cyber Monday | Late November | Online retail competition peaks | Use the conservative CPC case for retail |
| Year-end | December | Many B2B buyers on leave; consumer spending high | Cut B2B lead targets; keep consumer budgets |
Forecasting with a small budget
Most Singapore SMEs don't have the conversion volume that big-brand forecasting assumes. With fewer than 30 conversions a month, a single week's results tell you very little, so forecast monthly ranges and judge them over a quarter.
Track lead quality alongside volume. A form on a cheap campaign can fill up with time-wasters. Record which leads became quotes, in a simple sheet or your CRM, and fix tracking before you forecast if you can't see that yet. Our guide to measuring traffic you can't see in GA4 covers the self-reported "how did you hear about us" field, which helps here too.
Present the forecast so it gets believed
Management doesn't need a perfect forecast. It needs to know the assumptions and what happens if they're wrong. State four of them in writing every time: the CPC you assumed, the conversion rate, how often creative will be refreshed and how long you expect the learning period to last.
Show the three scenarios side by side, and report in money the business already tracks, such as revenue, cost per customer and payback time, rather than clicks and ROAS alone. Then review the forecast against actuals every month with the same short agenda: what missed, which input drifted and what you'll change. After three months you'll have a model built on your own numbers instead of borrowed ones.
Frequently asked questions
How do you forecast paid media results?
Divide budget by cost per click to get clicks, multiply by your conversion rate to get enquiries, then by your close rate and order value to get revenue. Run it three times with conservative, expected and aggressive inputs, and show weeks one to three as a learning period.
Why are my Google Ads results worse than forecast in the first weeks?
Usually the learning period. Google says a new or changed bid strategy can take one to two conversion cycles to calibrate. Meta needs about 50 optimisation events in seven days per ad set. Expect weaker results in that window.
How accurate is Google Ads Performance Planner?
It's a good starting point because it simulates the last 7 to 10 days of real auctions. It doesn't know your close rate, margins or seasonal dips, so use it for the clicks and conversions part of your forecast and add your own sales numbers on top.
What is the biggest reason ad forecasts fail?
Rising cost per click. NP Digital's data puts CPC inflation behind 54% of paid forecast failures. Build a conservative case with CPC about 20% higher than today, and check it against last year's prices for the same weeks.
How much budget do I need for a reliable forecast?
Enough to produce steady conversion data. With fewer than 30 conversions a month, forecast monthly ranges and judge results over a quarter, or optimise for a more frequent action such as a WhatsApp click so the bidding system has enough signals.
Get a forecast you can plan around
We build three-scenario forecasts from your own account data, set the tracking that makes them checkable and review them with you every month. If you're planning a Google or Meta budget for next quarter, talk to our SEM team.
