Push Notification ROI Calculator

How this calculator works

The calculator above models a push program as a simple funnel: people who can receive a message, people who actually receive it, people who engage with it, people who convert, and what each conversion is worth. Multiply those together, subtract what the channel costs you, and you have an estimate of monthly return. Every input is one you can pull from your own analytics or campaign history — nothing here depends on industry benchmarks.

Treat the output as an estimate, not a forecast. The formula is deliberately linear and transparent so you can see exactly which assumption is doing the work. If the result looks too good, one of the inputs is optimistic, and this page is designed to help you find which one.

The same funnel applies whether the message arrives as an app push, an SMS, a web push, or a wallet pass update on the lock screen. What changes between channels is the cost structure and how each stage is measured — both covered below.

The inputs, defined

Each field in the calculator corresponds to one of these. Read the definition before you type a number — most inflated ROI estimates come from using a nearby-but-wrong metric.

The formula, step by step

This is the exact arithmetic the calculator runs. The example column uses invented round numbers purely to show the mechanics — it is not a benchmark and your figures will differ.

Where to get honest numbers

The formula is only as good as its inputs. Work through these before trusting the output.

1. 1

   Pull the last 90 days, not your best campaign

   Average your funnel rates across every send in a recent window. Your flash-sale blowout is real, but it is not your Tuesday. If you only have a few campaigns, say so when you present the number.

2. 2

   Match the CTR definition to the channel

   App push measures opens, email measures clicks, wallet passes surface on the lock screen where a message can influence a visit without a tap ever being recorded. Enter the rate your channel actually reports, and interpret the output with that definition in mind.

3. 3

   Use margin, not revenue, for value per action

   A 40-dollar order with 60 percent cost of goods and a shipped-free promise is not worth 40 dollars. Finance will make this adjustment for you if you do not make it first.

4. 4

   Pick one attribution window and state it

   A conversion within 24 hours of a send is a very different claim than one within 7 days. Either window can be defensible — mixing them, or leaving the window unstated, is not.

5. 5

   Discount for incrementality

   The clean way to measure it is a holdout: exclude a random slice of the audience from a campaign and compare conversion between groups. Until you have run one, pick a deliberately modest factor and label it as an assumption.

6. 6

   Run three scenarios

   Enter pessimistic, expected, and optimistic values and record all three outputs. The spread between them is the real finding — it tells you how much confidence the estimate deserves.

Reading the result

Break-even first. Divide your monthly channel cost by the incremental value of one conversion (value per action times incrementality factor). That is the number of caused conversions per month the program needs to pay for itself. For channels with low flat costs and no per-message fee, this threshold is often startlingly small — which is exactly why it is worth computing rather than asserting.

The formula is linear, so errors pass straight through: if your CTR estimate is off by half, the output is off by half. The input you are least certain of sets the error bar on the whole result. For most teams that is the incrementality factor, followed by conversion rate — audience size and cost are usually known precisely, so agonize over the rates, not the counts.

Then read your three scenarios together. If even the pessimistic case clears cost, the program justifies itself and the conversation moves to how to grow the audience. If only the optimistic case clears cost, you are betting on best-case behavior — run a holdout or improve a funnel stage before scaling spend. And since each funnel stage is a multiplier, the calculator also tells you where work pays off: compare what a plausible improvement in delivery, CTR, or conversion each does to the output, and invest in the stage with the most realistic headroom.

Sanity checks before you quote this number

Run down this list before the estimate goes into a deck, a budget request, or a client report. Every item is a way real programs quietly overstate ROI.

• Inputs come from your own data

  Every rate was pulled from your analytics or campaign history, not from a vendor case study or an industry roundup.

• CTR and conversion share a cohort

  Both rates were measured on the same audience over the same period, so multiplying them is legitimate.

• No double-counted conversions

  A purchase attributed to the push is not also being claimed by email, paid social, or SMS in the same report.

• Value per action is net of discounts

  If the campaign carried a coupon, the redemption value reflects the discounted price and the margin, not list price.

• Incrementality is below 100 percent

  Some attributed converters would have bought anyway. If your factor is 1.0, the estimate is a ceiling, and you should say so.

• The send cadence is survivable

  Sends per month sits at a level your opt-out and uninstall data supports. Funnel rates measured at 4 sends do not hold at 20.

• Audience size is current actives

  Lists decay. The number reflects today's reachable audience, not the historical total of everyone who ever opted in.

• Labor is in the cost line

  Hours spent writing, designing, and scheduling campaigns are part of monthly channel cost, not free.

• Per-message fees are modeled where they exist

  For SMS and similar channels, cost grows with audience and cadence — check the ROI still holds at the audience size you are projecting, not just today's.

• The attribution window is stated

  The report says which window was used, and every campaign in the average used the same one.

• The result is presented as a range

  Pessimistic and optimistic scenarios appear alongside the expected case. A single point estimate overstates your certainty.

• The window is not just peak season

  If your 90 days covered a holiday spike, either widen the window or label the estimate as seasonal.

• No silent cannibalization

  The estimate does not assume another channel goes dark. If push replaces sends you were making elsewhere, count the incremental lift, not the transferred total.