Email Service Providers: How to Choose the Right ESP
Nine questions to ask before signing, including the three about leaving that nobody asks until they need to.
On this page
The short version
- Nine questions, and the three about leaving are the ones nobody asks until they are leaving.
- Feature comparison tables are the least useful input — almost every platform has almost every feature, and the differences are in how well each works at your size.
- Ask for the answers in writing during the sales conversation. What a salesperson confirms is a different artefact from what a marketing page implies.
Platform comparisons tend to be feature grids, which is the least discriminating way to choose. Every serious platform sends email, segments lists, runs automations and reports on them. The differences are in the details of how, at what scale, and what happens when you want to leave.
The nine questions below are the ones whose answers actually vary, and several of them are difficult to find on any marketing page.
Nine questions
| Question | Why it matters | Good answer looks like |
|---|---|---|
| Can I authenticate with my own domain? | DKIM alignment and reputation accrue to you | Domain-aligned DKIM and a custom return-path |
| Shared or dedicated IPs, and how are pools managed? | Your placement depends on your neighbours | Pools segmented by sender behaviour |
| What does pricing count — contacts or sends? | Contact-based pricing charges for dead addresses | Clearly stated, with the next tier visible |
| When are automation rules evaluated? | Schedule-time evaluation mails people who have left the segment | At send time |
| What is the real send speed at my volume? | A slow send undermines any time-sensitive campaign | A stated rate, not "fast" |
| Which webhooks and events are exposed? | Everything downstream depends on this | Documented, with retries |
| Can I export subscribers with consent records? | Consent data cannot be reconstructed | Full export including source and timestamp |
| Can I export engagement history and suppression? | Segments start empty without it | Both, in a usable format |
| What happens to my data if I stop paying? | Access can be cut before you have exported | A stated grace period |
The evaluation question that catches people
When automation rules are evaluated — at the moment a send is scheduled, or at the moment it goes out — is a technical detail with a very visible consequence.
Schedule-time evaluation means a campaign built on Monday for Friday sends to Monday's segment membership. Anyone who unsubscribed, bought, or crossed a threshold in between is treated as though they had not. That is how a customer receives a promotion for something they purchased on Wednesday.
It is rarely documented prominently. Test it during a trial: build a segment, schedule a send, change a subscriber so they no longer match, and see whether they receive it.
What to test during a trial
Not the features, which will work. Test the things that are hard to reverse.
Set up domain-aligned authentication end to end and confirm from a received message. Import a realistic sample and check what happens to the fields you care about. Build one automation with an exit condition and verify the exit fires. Export everything and look at what came out.
That last one is the most informative hour you will spend, because it tells you what leaving costs before you have anything to lose.
What deliverability claims are worth
Every platform advertises high deliverability, usually as a percentage in the high nineties. That figure is almost always acceptance by the receiving server, which counts a message filed straight into spam as delivered.
It is also mostly about your list and your sending behaviour rather than the platform. A good platform gives you the tools — authentication, suppression, monitoring — and cannot compensate for a poorly maintained list.
The useful questions are about infrastructure rather than outcomes: how pools are segmented, whether senders with poor practices are removed, and what support exists when placement drops.
Before signing
- Domain-aligned DKIM and a custom return-path are supported
- Pricing basis and the next tier boundary are known
- Automation evaluation timing has been tested, not assumed
- Webhook coverage is documented
- A full export including consent records has been run
- Engagement history and suppression export in a usable form
- The non-payment grace period is stated in writing
- Send speed at your volume is a number, not an adjective
Frequently asked questions
Is a cheaper platform worth switching to?
Only after costing the migration honestly — export, re-warming, template and automation rebuilds, and the deliverability dip. Savings are annual; migration cost is immediate and often exceeds a year of the difference.
Should I choose based on integrations?
Check the specific integration you need rather than the count. A directory of five hundred integrations tells you nothing about whether the one with your commerce platform syncs the fields you need.
How much does support matter?
More than it seems until placement drops on a Friday. Ask what support tier includes deliverability help specifically — many exclude it from lower plans, which is where it is most needed.