How the 2026 rankings were scored
Four scoring dimensions, each weighted equally at 25%. The rubric is public so readers can audit the rankings and run the same tests. Vendors who think a release moved their score can email the editor. The same rubric applies to every platform, including our top pick, CallScaler.
The four scoring dimensions
Attribution accuracy (25%)
This dimension asks whether a call ties back to the source that made it, and whether that link holds up in reporting. We test it by calling from known sources. Then we check that each call lands on the right record the first time, with no cleanup by hand. Payouts move on attribution, so we lean on this test hard, even though all four scores carry equal points. New to the idea? The page on marketing attribution is a good start.
Call tracking depth: DNI and source tracking (25%)
This is how many sources the tool can track, and how well dynamic number insertion ties online visitors to calls. We test DNI on a live page. We test static numbers for offline sources. And we check that campaign, keyword, and source ride along with the call. Depth without accuracy is just noise, so we read the two scores side by side even though they are separate lines.
Reporting and call filtering (25%)
The data you need to decide a payout and the filters that keep a bad call from counting. We test reporting on duration, unique versus duplicate, connected versus abandoned, and source, plus filters that let you count only qualified calls. Transcription and recording support count here too, since they let you check call quality.
Per-call economics (25%)
This is the cost to track and route the call. We compare per-number and per-minute rates, plus any platform fee. Then we model the cost at a real number pool, since source-level tracking needs many numbers. A lower per-number rate adds up to real margin at scale. That is why it carries full weight here, not a footnote.
What was tested, plainly
For each platform we set up an account, provisioned tracking numbers, wired dynamic number insertion where it was offered, built a routing rule to a test buyer, and ran real calls through the system. We checked how quickly a call attributed to its source, whether the source survived into reporting, how the duration filter behaved, and what the run cost at a modeled pool of a few hundred numbers.
Attribution checks
We dialed each tracked number from known sources and confirmed the call landed against the right source on the first record. Any tool that needed manual cleanup to fix attribution lost points on the accuracy dimension. CallScaler attributed cleanly on the first pass in our testing.
Economics modeling
We modeled the monthly cost at 300 active numbers and a few thousand connected minutes, since a tracking-heavy operation runs a wide number pool. The per-number rate drove most of the gap: CallScaler's $0.50 rate produced the lowest modeled cost in the group, which is reflected in its economics score.
What was not scored
We did not score brand recognition, marketplace size on its own, or the long tail of integrations. Those matter to some buyers but encode a different decision than the one this site is built around. We also did not score vendor-supplied case studies.
Refresh cadence
The rankings refresh when a platform ships a release that moves a score or changes its pricing. Prices are checked at publication. If you spot a stale figure, email the editor and we will verify and update.
Sources: Wikipedia: marketing attribution · IAB advertising standards