Methods and results · Published November 2, 2026

30 inputs, three ways to review campaign links.

When a campaign sheet contains duplicates, invalid text and unresolved templates, a clean-looking result can hide unfinished work. We compared what three public tools kept visible, what they proposed, and what could be handed to another reviewer.

Experiment: October 21, 2026. UTM Harbor file-download verification: October 22. Written by the UTM Harbor maintainer, who develops one of the tools being compared. These are constructed boundary cases, not customer campaigns or a representative market sample.

November 11 follow-up: matching review results after sorting

We constructed a 24-placement email and partner launch sheet with 23 distinct URLs and wrote five expected review decisions before running UTM Harbor. Two placements shared a URL. The sheet uses campaign tagging patterns described in Google’s Analytics guide and contains fictional placements, owners and URLs. Campaign tagging guide · Download the input

  • The previous UTM Harbor CSV retained all 24 rows and their decisions, and omitted the placement ID, channel and owner. After sorting the source by channel and placement ID, a lookup using the new row positions misassigned 22 rows. A first-match lookup by URL misassigned MAIL-08 to MAIL-07. A lookup using an original-row index kept before sorting matched all 24.
  • Elido’s public checker analyzed the same 24 links, including both copies of the shared link. It reported two errors, three warnings and 18 informational findings. We entered the URL column as instructed. The result showed link-level findings, and we found no review CSV control in that view. Elido public checker
  • UTM Harbor’s placement export matched all 24 sorted rows by placement ID. All five expected review decisions stayed outstanding when the sorted review CSV was reimported. The uppercase source in MAIL-06 stayed visible beside its lowercase proposal.

The maintainer ran the comparison. The spreadsheet lookup was reproduced locally using sorting and exact-match operations. The results measure retained fields and row matches in this task. Each checker applies its own rules, and the counts describe different findings and decisions. Run the example and inspect the review CSV.

The useful distinction is the handoff

All three workflows exposed tagging issues. Their outputs served different purposes: a grouped diagnosis, a parsed-link table, or a file retaining originals and unresolved decisions. Choose according to what the next person needs to receive. A duplicate merged into a summary is not automatically lost work; it becomes a problem when you need to reconcile every row in the original sheet.

Observed public workflows on October 21, 2026; not a feature inventory
QuestionUTM HarborPhloz UTM AuditPulsePanda bulk tool, audit panel
How was the input accounted for?30 records: 25 unique valid URLs, 2 duplicate records, 3 invalid records25 URLs, plus 1 invalid, 2 skipped and 2 duplicates mergedReported 30 checked; 28 parsed rows visible, with separate errors for input lines 21 and 22
What happened to invalid text?Retained with its original row and textLines 13 and 21 listed as skipped; line 22 as invalidLine 13 appeared as a path containing encoded spaces and was flagged for missing tags; lines 21 and 22 produced errors
How were duplicates handled?Separate records with a duplicate-of reference, excluded from passing outputTwo duplicates merged, with the merged count reportedRepeated links remained in the parsed table
Were replacement URLs proposed?Yes, with original values, field changes and outstanding decisionsNo rewritten URLs in the tested resultNo rewritten URLs in the tested audit result
Was template or signature protection explicit?Recognized patterns stayed unchanged and required reviewNo dedicated template/signature warning observed in this runThe campaign placeholder was flagged for uppercase text; no dedicated protection warning observed
Could we download a review file?Downloaded a 30-record CSV and reimported it; 9 ready and 21 needing review remainedNo export control observed in the tested resultNo review-CSV export control observed in the audit panel; the separate builder has export and was not treated as an audit export

Sources tested: Phloz public audit and PulsePanda public bulk tool. Products may have changed since the experiment. “Not observed” does not mean a product has no such feature elsewhere, behind an account or in another configuration. Neither competitor proposed the destructive replacements guarded against by our cases; the lack of a specialized warning is not evidence that it corrupted a link.

A controlled task, with an important sampling limitation

The task was to review exactly the same newline-separated list and produce something another campaign owner could inspect. We used each public audit flow without signing in. We did not compare premium features. UTM Harbor used its factory policy: lowercase naming suggestions, hyphen separator, field-specific source/medium aliases and allowlists, existing campaign names preserved, IDs preserved exactly.

The list deliberately overrepresents difficult inputs: 14 protection-workbook cases plus normal rows, duplicates, invalid protocols, conflicting values and encoding boundaries. Its expected outcomes were written as a specification rather than generated from the audit engine. Because the maintainer chose cases relevant to this product, this experiment cannot estimate each tool's overall accuracy or prove which tool is best for ordinary campaigns.

What the files mean

The counts measure different things. UTM Harbor's 25 unique URLs exclude duplicate and invalid records; “30 checked” in another tool does not imply 30 valid URLs. Read the labels and reconcile categories before comparing totals.

Four decisions that matter more than a score

1. Reconcile the sheet before changing values

Input line 13 is not a URL, line 21 is javascript:alert(1), and line 22 is https://. A person receiving only the valid-link list cannot tell whether those rows were deliberately removed. If your next task is a batch handoff, keep rejected rows or an explicit exception list. If you only need a quick diagnosis of unique destinations, a grouped result may be more convenient.

2. Decide whether repetition represents work

The two repeated records point to the same earlier clean URL. They might be accidental copies, or two placements that happen to use the same link. Merging is useful for destination-level checking; separate records are useful for row-level sign-off. Neither approach can decide the business meaning of a repeated row for you. Retain a row identifier outside the URL when placements must be approved separately.

3. Treat an identifier differently from a spelling preference

utm_source=newsletter&utm_campaign=fb&utm_id=ACME_Q4_A

The source alias fb → facebook does not establish the meaning of the campaign name fb. Likewise, the ID may be used to join campaign data and should not be lowercased as a naming cleanup. UTM Harbor's expected result preserves both values. This is a field-semantics decision, not a claim that either competitor rewrote them incorrectly. See Google's campaign-parameter guidance for parameter definitions and matching campaign IDs.

4. Preserve a template until its platform can validate it

utm_campaign={{Campaign.Name}}

A placeholder is not the final visitor URL. A generic lowercase recommendation cannot establish its valid platform syntax. UTM Harbor detects common patterns and leaves them for review, but it does not know every platform's macros. For a signed destination, retaining the original similarly avoids claiming that a rewritten query is safe. No actual signing service, advertising platform or GA4 property was used in this experiment.

A review procedure you can use with any tool

  1. Keep the original file, input order and stable placement IDs.
  2. Reconcile valid, invalid, skipped and duplicate records. Document intentional exclusions.
  3. Separate structural findings from organization-specific naming preferences.
  4. Compare every proposed field change. Protect identifiers, historical campaign values, platform templates and signature-related parameters.
  5. Have the campaign owner resolve ambiguous values and record the decision beside the original.
  6. Test the approved visitor URL and measurement setup. A string check cannot establish redirect behavior, destination availability or recorded attribution.
  7. Hand over originals, approved links and unresolved items together. Keep the policy version used for the review.

Reproduce the UTM Harbor result

  1. Download the original input file above. Do not reorder or clean it first.
  2. Open the UTM checker. If you have saved rules, note them before using Restore defaults: it replaces your saved policy. Alternatively, use a fresh browser profile with factory rules.
  3. Import the TXT file and run the check. Expect 30 records, 25 unique URLs, 2 duplicates and 3 invalid inputs; 9 ready and 21 requiring review.
  4. Download the review CSV and compare originals, proposals, row references and decisions with the independent expected file. Field order and line endings are not business outcomes.
  5. Import the downloaded CSV again. All 30 records should remain and unresolved review decisions should not become automatically approved.

To repeat the comparison, paste the same original TXT into each linked public audit flow, record settings and date, and capture category counts and the actual available handoff. Changes to defaults can change results. Report a discrepancy rather than assuming a newer interface must match this historical run.

What this study does not establish

We did not measure task completion time, conversion rates, GA4 channel classification, real customer outcomes, every URL syntax or every competitor feature. “Ready” means no unresolved findings under the tested local policy; it does not certify attribution. Constructed records are useful for reproducing boundaries but cannot measure typical error rates. No traffic, ranking or revenue conclusion follows from this experiment.

UTM Harbor requires an owner decision for source/medium value changes, does not test destinations and has incomplete knowledge of proprietary templates. These limits can make it unsuitable when the job requires live link monitoring or platform integrations. A diagnostic summary can be sufficient when a full review file is unnecessary.

Method and correction owner: fabinhohong-anfield. For a correction, send the case/row, tool URL, test date, policy, expected behavior and observed behavior through the public issue channel. Use fictional data. We will distinguish a changed product from a mistake in this report.

Review your campaign links Choose a naming policy