A Projected Number Is Not a Reported Result
You are reading a case study and a number stops you. A dollar figure per customer, stated as an annual value, in a sentence that sounds like a report of something that happened. Ask one question before filing it away. Has that money arrived, or has it been calculated to arrive? Those are different claims, and the difference is usually visible in the sentence itself.
The tense is a fact you can check
Some phrases mark a number as forward-looking however confidently the sentence is built. Is projected to reach. Is estimated at. Is expected to generate. Modeled. Annualized run rate of. If retention holds at. Each describes arithmetic performed on an assumption, not a figure counted off a ledger. The assumption might be excellent. It is still untested.
The contrasting set is blunter. Generated. Resulted in. Brought in. Closed at. Booked. These point at something that landed somewhere countable: an invoice, a bank record, a closed-won field in a CRM. A reader can ask to see it. With a projection there is no record to ask for.
Here it gets harder, because grammar and substance can point in opposite directions. Take this invented line: “the campaign delivered a $50,000 lifetime value.” That is past tense, and delivered is a realized verb, but the number is still a forecast, because lifetime value describes revenue that has not all arrived. The verb tells you how the writer framed the claim. The metric tells you whether it can be about the past at all. When they disagree, believe the metric. That is one narrow application of how to read a case study without being misled.
Why a lifetime value figure is a projection by definition
A lifetime value number, however labelled, covers a window extending past the data collected. It answers what a customer will be worth, pricing in purchases not yet made and retention not yet demonstrated. Revenue collected from a cohort to date is an observation. Extend it forward and it becomes an estimate with error bars nobody drew.
Platforms are explicit about this in their own documentation. Google’s page on GA4 predictive metrics defines “Predicted revenue” as “The revenue expected from all purchase key events within the next 28 days from a user who was active in the last 28 days.” The same page says Google Analytics “automatically enriches your data by bringing Google machine-learning expertise to bear on your dataset to predict the future behavior of your users.” That language names a forward window, names a mechanism, and calls the output a prediction, not a total. It is cited only to show what “predicted” revenue means technically, not to suggest any case study in this library uses that platform.
A hypothetical case study line, read closely
Consider this sentence, invented for this article. It describes no real company or campaign, and its figures are not reported results.
“The campaign delivered a projected annual value of $42 per customer.”
Delivered is doing work it has not earned, because what was delivered is the projection, not the $42. An annual per-customer value assumes a repeat-purchase rate nobody can have observed for a year unless the campaign ran a year. If it launched thirty days ago, most of that $42 is forecast wearing the clothes of a result.
The honest version of the identical claim, also hypothetical: “Based on the first 30 days of purchases, we modeled a $42 annual value per customer if repeat-purchase behavior holds steady. Revenue actually booked from this cohort so far is $11.”
That rewrite is not weaker, just accurately scoped. It names the observed data, names the assumption, and puts the realized number beside the modeled one, so a reader can judge the extrapolation. The first version prevented that. Whether either figure could be checked is a separate question, covered in whether a number can be backed up at all.
Four questions for any headline number
- Has the time window this number covers actually finished? If any part is still ahead, part of the number is a forecast.
- Does the verb describe something that happened, or something calculated to happen? Then check whether the metric agrees with the verb.
- What assumption would have to hold for this number to arrive as stated? Usually retention or repeat rate. If you cannot name it, it was not disclosed.
- Is there a smaller, already-realized number underneath, and is it shown? The observed base is the more honest figure, and its absence is informative.
A disclosed projection is not the problem
Projecting a result is legitimate and often unavoidable. A final lifetime value figure cannot exist while the customer is still a customer. Sizing an opportunity and deciding whether to keep spending both require estimating forward. Sceptical reading is not the same as bad faith.
The line falls somewhere specific. The failure is not that a projection exists. It is a projection presented without its assumptions and without its tense, so a modeled figure reads as money already banked. A projection that shows its working is a contribution. One dressed as a receipt is not, which is the instinct behind the number no case study can show you.
Projected language next to realized language
| Projected phrasing | Realized counterpart |
|---|---|
| Projected annual value | Revenue collected to date |
| Expected to generate | Generated |
| Modeled lifetime value | Total revenue from this cohort so far |
| Annualized run rate | Revenue booked this quarter |
| Estimated impact | Measured impact |
| Pipeline influenced | Deals closed |
Read a few for yourself
This reflex builds fastest on real published work, not examples written to make a point. Browse the case study library and run the four questions against the headline figures there, remembering that each number belongs to the agency that reported it. If you publish results, submit your own.
FAQ
Is a projected number automatically untrustworthy?
No. Projections are a normal part of reporting, particularly for a metric covering a window that has not finished. The problem is not that a projection exists. It is one phrased so it reads as an already-banked result, with no assumption named and no realized figure beside it.
What specific words should make me stop and check a number?
Projected to, estimated at, expected to generate, modeled, annualized, and conditionals such as “if retention holds.” Those are the easy catches. The harder discipline is remembering that past-tense grammar does not guarantee the number was observed. If the metric describes a future window, it is a forecast whatever verb is attached.
Sources
- Google Analytics Help: [GA4] Predictive metrics, fetched 2026-09-06.
