Router, single verifier, or the waterfall you build yourself
Three real options, and the honest case for each. Sortance buys from most of the vendors it would otherwise be scoring, so this compares the shape of the approaches rather than publishing a league table we have an interest in.
Side by side
| Single verifier | Your own waterfall | Sortance | |
|---|---|---|---|
| What you get back | A label: valid, invalid or unknown. | Whatever you decide to synthesise from several labels. | A calibrated probability with a confidence interval, plus disagreement and actor class. |
| Catch-all domains | Usually a confident guess, or an unknown. | You see the disagreement, but have to interpret it yourself. | A wide interval and a catch_all signal. Nobody resolves these. |
| Provider outage | Your call fails, or the answer silently degrades. | You write the circuit breaker and the fallback. | 200 with degraded: true and a prior-derived number, labelled as such. |
| Cost control | Per check. Checking more costs more, always. | You build the gate and the cache, and you tune them. | Local gate and cache resolve most traffic; paid fan-out only where it changes the answer. |
| Contracts and keys | One account, one key, one vendor risk. | An account, a key and an invoice per provider. | One key. The pool runs on our contracts. |
| Does it improve with your data? | No. Its model is its own. | Only if you build the calibration loop, which is the hard part. | Yes, once you report outcomes — prior becomes fitted per feature bucket. |
| Where your list lives | Often uploaded to a dashboard. | In your system. | In your system. A bulk upload keeps the file 7 days, then deletes it. |
| Enrichment (who is this person, what company) | One enrichment vendor: its coverage, its industry taxonomy, its gaps, and no second opinion. | A hand-rolled fallback chain in n8n or Zapier: no shared cache, no cost ceiling, and two or three vendor taxonomies to reconcile by hand. | Company first by domain, a second vendor only on no match or thin coverage, one normalized vocabulary, a coverage figure and a per-vendor receipt. |
| Time to first value | Minutes. | Weeks, then ongoing maintenance. | Minutes. |
When a single verifier is the right answer
If your list is overwhelmingly Google Workspace mail, one cheap check resolves it, and a router is overhead you do not need. Sortance will in fact tell you this — a clean human mailbox on Google Workspace comes back with a single provider in the providers map and disagreement: 0.0, which is the system saying one check was enough.
It is also the right answer if you genuinely only need a label. Not every product needs a probability, and if a boolean is going to be thresholded into a boolean anyway, paying for the interval is waste.
When building it yourself is the right answer
Hand-rolled waterfalls exist because they work. If you already have provider contracts, a cache, and enough bounce data to calibrate against, you have most of what Sortance is — and you keep full control of the policy.
The part that is genuinely hard, and the part most in-house waterfalls skip, is calibration: turning three disagreeing labels into a number that means what it says, and keeping it honest as your traffic changes. Fanning out is easy. Knowing what the spread means is the work.
When a router is the right answer
- Your list is mixed, and a meaningful share of it is Microsoft-hosted or catch-all — the exact traffic where a single check stops being informative.
- You need to act on a threshold rather than a label, and you need to know how much to trust the number before you act.
- You want the fan-out without four vendor contracts, four keys and four invoices.
- You care who is behind the mailbox — human, agent, relay alias, role account — and not only whether mail lands.
Who is in the pool
Sortance does not run its own SMTP and does not claim a private signal. First in the pool is its own Sortance check (syntax, MX and a disposable-domain blocklist, never a mailbox claim). Mailbox checks are bought from providers you already know — DeBounce, MyEmailVerifier, ZeroBounce — and every response reports which checks answered, what each said, and what each cost.
That transparency is the point. If one provider is carrying a decision you disagree with, you can see it in the providers map rather than inferring it.
The same three options, for enrichment
Growth teams already run an enrichment waterfall: try one vendor, fall back to the next, write whatever comes back into the CRM. It works, and it has the same weaknesses an in-house verification waterfall has — no cache, so the same company is bought twice; no ceiling, so a policy change can quietly become a monthly bill; and three vendors’ ideas of what “fintech” means.
Sortance’s enrichment lane maps every vendor into one closed vocabulary, reports which fields came back as a coverage figure, and only buys a profile when one of your rules actually needs it. A single vendor is still the right answer if you already hold a contract and need one kind of field. See enrichment for what is and is not covered.
What Sortance is worse at
- No hygiene score. If you want to screenshot a hygiene score for a list, a list-cleaning product is a better fit. Sortance’s CSV upload returns a probability per row, not a grade.
- No headline accuracy number. Competitors publish one; we argue on limitations that it is meaningless. If a procurement form has a box for it, that is friction we are choosing.
- A new account is not calibrated to you. Day one is a prior, and it stays a prior until you report outcomes.