How to spot assumption drift early.
Assumption drift is the most expensive kind of drift, and the easiest to miss. It happens when the inputs your plan was built on stop matching reality. The model keeps doing math on inputs that are no longer accurate, the variance compounds quietly, and finance finds out three or four weeks later in the close.
This post is a practical guide to catching assumption drift early, whether or not you are running Peaky. The principles work in any planning tool. The cadence is what matters.
Which assumptions drift first
Not all assumptions drift at the same speed. Some are stable for quarters. Some can drift inside a week. Knowing which ones move first is half of the discipline.
Five drivers tend to drift first at scaling companies.
CAC. Customer acquisition cost moves the moment a campaign launches, a channel saturates, or a market segment shifts. If you are running paid acquisition, CAC can drift twenty percent inside a single week of a new campaign. Watch it weekly.
Pipeline conversion rate. The percentage of pipeline that closes within a given time horizon. This drifts whenever your sales motion changes. New AE ramping. New segment. New pricing tier. Compensation change. Pipeline conversion is the assumption most teams check least often and need to check most.
Churn. The slow killer. Churn rarely moves dramatically in a single month, but it can creep half a point a month for three months in a row. The cumulative impact is large. Cumulative impact is the thing the variance column never shows because each individual month was within tolerance.
Sales cycle length. The number of days between first contact and closed deal. This drifts whenever you move upmarket or downmarket, whenever competitor activity intensifies, whenever a new objection enters the conversation. It tends to drift up before pipeline conversion drifts down, which makes it a useful leading indicator.
Hiring velocity. The pace at which new hires actually join. The plan usually assumes the hires that were budgeted will start on the budgeted dates. Reality almost never delivers that. Either hires get pulled forward or delayed. Either way, the burn diverges from the plan within a quarter.
If you only watch five drivers, watch those.
How to build a leading-indicator list
Once you know which drivers drift first, the next move is to identify a leading indicator for each one. A leading indicator is a signal that moves before the driver itself moves.
For CAC, the leading indicator is daily ad spend by channel and impression volume. If impression volume drops while spend stays constant, CAC will drift up before the end of the month. If a new campaign launches, the spend pattern itself is the early signal.
For pipeline conversion, the leading indicators are average deal size, stage-to-stage conversion rate, and AE productivity. A drop in stage-2-to-stage-3 conversion is visible weeks before the final pipeline conversion variance shows up.
For churn, the leading indicators are usage signals: login frequency, feature adoption, support ticket trend. These move before customers actually churn.
For sales cycle, the leading indicator is average days in each stage. If deals start sitting longer in stage 4, the cycle will lengthen before the next quarterly average is in.
For hiring velocity, the leading indicator is offer acceptance rate and time-to-start. A drop in offer acceptance means hires will be late or replaced.
Build the leading-indicator list once. Use it every week.
What weekly discipline looks like
Once a week, look at the leading indicators. Compare them to the assumption your plan is using. Flag anything outside a defined tolerance.
The discipline is more important than the tool. You can do this in a spreadsheet. You can do this with Looker. You can do this with Peaky. The cadence is what catches the drift early.
Most finance teams have either no weekly discipline at all or a weekly discipline that does not connect leading indicators back to plan assumptions. The work the weekly discipline does is to close the loop between operational reality and planning input.
The tolerance question
The hardest part of catching assumption drift early is deciding what counts as "drift" versus what counts as "noise".
Set a default tolerance per driver, based on how much variance the business can absorb before the impact is meaningful. For most Series A to C companies, the defaults look like this.
CAC: 15% from baseline. Below that is noise. Above it is a campaign conversation.
Pipeline conversion: 5 percentage points. Smaller moves are within historical volatility. Larger moves indicate something structural.
Churn: 50 basis points. Smaller moves are common. Larger moves require a CS intervention.
Sales cycle: 15% on the average. Less than that is normal noise. More than that is a sign of friction.
Hiring velocity: any miss on the planned start date by more than two weeks.
These are starting points. Tune them based on your own historical volatility. The point is to have them set in advance, so that the decision about whether to act is not made in the moment of seeing the data.
What to do when you see drift
When a leading indicator moves outside tolerance, three things should happen.
First, raise the signal to the driver owner. The CRO owns pipeline. The CMO owns CAC. The Head of CS owns churn. The Head of People owns hiring velocity. Each driver needs an owner who is on the hook for the signal.
Second, decide whether the assumption needs to update. Sometimes the right answer is to update the plan to match reality. Sometimes the right answer is to change the underlying reality so the plan can hold. Both are valid. The discipline is to make the choice deliberately rather than letting the plan drift quietly.
Third, communicate the change. The plan now expects something different. The board pack, the operating review, the investor update should all reflect the new assumption. Misalignment between the plan version finance is using and the plan version operations is using is its own kind of drift.
What not to do
A few common mistakes are worth avoiding.
Do not chase every signal. The point of setting tolerances in advance is to filter noise. If you re-evaluate every small move, the team loses faith in the tolerance system and starts treating every signal as actionable, which is exhausting and counterproductive.
Do not move the plan every time the data wobbles. If CAC drifts 12% one week and back to baseline the next, the plan does not need to change. The right move is to flag the wobble, watch for repeat, and update only when the move sustains.
Do not pretend assumption drift is a finance problem alone. The driver owners need to be in the conversation. Finance can identify the drift. Only the function owner can decide what to do about it.
What Peaky adds
Most of this discipline can be done manually. It is hard but possible.
What Peaky adds is the automation of the comparison. Instead of a finance team checking five drivers every week against leading indicators in five separate systems, the comparison runs continuously in the background. When something moves outside tolerance, a signal appears with context, impact, and a recommended adjustment.
The discipline does not change. The work behind the discipline does.
If you want to see what continuous assumption-drift detection looks like on your own data, you can start a Peaky trial at peaky.ai. Most teams have their first signal within a day of connecting.
The financial infrastructure for scaling without surprises.


