Push Platform Comparison Scorecard

How to use this scorecard

This is the working version of the comparison tables in the guide: nine criteria, a weighting method, and a blank matrix you fill in for your own shortlist. Use it directly on this page or print it and score with a pen. Weighting takes minutes; honest scoring takes a demo and a trial.

The scorecard is deliberately blank. Platforms change their features and pricing quarter to quarter, and your weights change the answer more than any published verdict would. A team shipping order alerts and a team running retention campaigns can score the same platform two points apart on the same row, and both be right.

The method is simple: define what matters (criteria), decide how much each thing matters (weights totaling 100), score each platform 1 to 5 from evidence you actually saw, and multiply. The highest weighted total wins, unless a heavily weighted criterion scores so low it becomes a disqualifier.

The nine criteria

Score each criterion from 1 (weak) to 5 (excellent). The two outer columns describe what the extremes look like so two people scoring independently land close together.

The scoring method

1. 1

   Shortlist three platforms

   Use the category tables in the guide to pick candidates. Beyond three, scoring quality collapses because nobody trials four platforms honestly.

2. 2

   Set your weights

   Pick the closest column from the suggested-weights table below or write your own. Weights must total 100 so every point given to one criterion is visibly taken from another.

3. 3

   Score 1 to 5 from evidence

   Score only what you saw in a demo, a trial, or a documentation page you can cite. Write the evidence in the notes column next to the score; a score with no evidence is a guess.

4. 4

   Compute weighted totals

   Multiply each score by its weight and sum per platform. The maximum possible total is 500.

5. 5

   Apply the disqualifier rule

   Any criterion weighted 10 or more that scored a 1 or 2 is a disqualifier regardless of the total. Totals rank platforms; disqualifiers remove them.

6. 6

   Re-score after a trial

   Run a two-week trial with the leader and re-score delivery, developer experience, and support from lived evidence. Scores nearly always move once real traffic flows.

Suggested weights by program type

Three starting points, each totaling 100. Adjust them to your program, but keep the total at 100 so the trade-offs stay explicit.

The scorecard

Fill in your weights, then score each platform 1 to 5 per row. Multiply score by weight, sum the column, and note the evidence behind every score.

Sanity checks before you decide

Run these before treating the winning total as a decision. Each one catches a way scorecards quietly go wrong.

• Every score has cited evidence

  If the notes column is empty, the score came from a pitch deck or a hunch. Re-score or mark it unknown.

• Pricing modeled at twice your volume

  Model both subscriber growth and send growth. Per-MAU and per-send pricing diverge sharply as a program succeeds.

• Overage behavior confirmed in writing

  Ask exactly what happens at the limit: throttled sends, an automatic bill, or a blocked account. Get the answer in email, not on a call.

• Test sends reached your own devices

  Verify on iOS and Android, including how the message renders on a locked screen, before scoring delivery above a 3.

• Export path exists for everything

  Subscribers, tokens, and event history should export without a support ticket. If leaving is hard, the renewal negotiation is already lost.

• Migration cost estimated honestly

  App push tokens usually move between platforms; web push subscriptions tied to a vendor's service worker and keys often cannot. Know which situation you are in.

• Frequency caps and quiet hours actually configured

  Set them up in the trial rather than trusting the docs. Where the setting lives tells you how the platform thinks about subscriber fatigue.

• DPA and data residency obtained in writing

  If compliance matters to your program, a signed DPA and a clear answer on where data lives are prerequisites, not nice-to-haves.

• Support tested with a real ticket

  Send a genuine technical question during the trial and time the answer. Trial-period support is the best it will ever be.

• Contract terms read before the demo glow fades

  Check term length, the auto-renew notice window, and any price-increase clause. These outlast every feature on the scorecard.

• Unlimited claims checked for fair-use caps

  Unlimited sends and unlimited contacts usually carry a fair-use clause. Find it and read it against your peak month.

• A second stakeholder scored independently

  Have someone else score the same platforms without seeing your sheet. Large gaps between scorers reveal criteria you defined too vaguely.