Agencies
Forecasting results for clients without promising what you do not control
The client wants a number and you control neither the auction nor their sales team. How to build a forecast with a range, stated assumptions and a review date.
Also available in: Português · Español
"How many leads will we get next month?"
The wrong answer is a single number. The other wrong answer is "it depends". Between the two there is a useful forecast: a range, stated assumptions and a review date.
Why the single number is a trap
You do not control the auction, seasonality, what a competitor will do, or how fast the client's sales team answers. Promising 200 leads is betting on four variables, of which you command one.
And the cost of being wrong is asymmetric. Delivering 240 earns no proportional praise; delivering 160 earns a difficult meeting. The single number is a bet where you win little and lose a lot.
The structure of an honest forecast
NOVEMBER FORECAST · ACME
Spend $40,000
Leads 190 to 240
CPL $167 to $210
Estimated sales 17 to 26
Estimated revenue $119,000 to $182,000
ASSUMPTIONS
· CPL over the last 90 days: $188 (ranging $160 to $215)
· Current close rate: 9.2% — held
· Lead response time under 1 hour
· No material change in the auction or the offer
WHAT WOULD BREAK THIS FORECAST
· Response time above 4 hours (−40% on sales, per history)
· The main creative saturating with no replacement
· A new competitor entering the auction hard
REVIEW: the 14th, with two weeks of data
Three parts make this forecast work.
The range. It comes from the real variation in your history, not from a comfortable guess. If CPL swung between $160 and $215 over the last 90 days, that is the range — inventing a narrower one to look confident is promising what your own history contradicts.
The assumptions. They move into the conversation the things that are not yours. Response time on that list is not an accusation: it is a statement that the forecast depends on it.
The review. A set date to reassess. A forecast is not a promise, it is a hypothesis with a deadline.
How to arrive at the range
With history (3 months or more)
1. Average CPL over the last 90 days and the real variation in the period
2. Lead range = budget ÷ CPL range
3. Close rate over the last 90 days, from the CRM
4. Sales range = lead range × close rate range
5. Adjust for seasonality if the same month last year was atypical
Step 3 is the one most people skip, and it is what separates a lead forecast from a results forecast.
With no history
Wide ranges, declared as wide:
The first month is learning. The realistic range is broad — 120 to 220 leads — and it will narrow in week 3, when we have our own data. Until then I am using references from similar accounts, which are good for planning and not for holding anyone to.
That honesty up front avoids the entire month-2 confrontation.
Deviations and how to explain them
When results fall outside the range, a well-built forecast already contains the diagnosis:
| what happened | what it means |
|---|---|
| CPL in range, leads in range, sales below | the close-rate assumption failed — sales or quality |
| CPL above range | auction, saturation or an audience change |
| Leads in range, revenue below | product mix changed, lower ticket |
| Everything above | great, and worth understanding why before assuming it repeats |
The first row is the most common and the most valuable: it shows, using the number you handed over at the start of the month, that media delivered and the deviation sits after it.
That is not dodging responsibility — it is locating the problem in the right place, which is what allows fixing it.
The forecasts you should not make
A sales guarantee. Half the path to the sale is not yours. Guaranteeing sales means taking on the result of the client's sales team, and that bet is lost often.
A 12-month forecast with monthly numbers. Nobody forecasts September's auction in January. Do a year in quarterly ranges, or do not do it.
A forecast that ignores capacity. Promising 400 leads to a team that handles 150 a month is promising a problem. Check capacity before the number.
A forecast without sales in the room. The close rate is an assumption of your forecast and someone else's work. If they did not agree to the number, you are forecasting another team's performance alone.
The review ritual
Put the review date in the forecast itself, and keep it.
On the day, three lines are enough:
The 14th: 97 leads (projecting 205, in range). CPL $179, in range. Close rate at 7.1% against the 9.2% forecast — response time rose to 3h. We have adjusted the sales forecast to 15–20 and have already spoken to the sales team.
A forecast revised mid-course is a management tool. A forecast remembered only at month end is a document for assigning blame.
For the report structure that closes the forecast loop, see the report clients actually read. For the assumption that most often breaks a forecast, see lead response time.
Frequently asked questions
Should I guarantee a number of leads or sales?
Never guarantee sales, because half the path to them is not yours. Work with a lead range and a sales estimate that depends on stated assumptions.
How do I forecast for a new account with no history?
Use wide ranges, declare a learning month, and review in week 4. A precise forecast with no history is a guess dressed as method.
What do I do when the forecast does not hold?
Show which assumption failed. If CPL stayed in range and sales did not come, the deviation is in sales conversion, and the forecast itself demonstrates that.
Does a range not look insecure to the client?
It looks less insecure than a single number that misses. A range with explicit assumptions is what an honest professional delivers, and clients notice the difference.