Peec and Scrunch document changes to historical results
Peec separates a dashboard display name from the tracked name and aliases used to match brand mentions. Its documentation says saving brand-setting changes triggers reprocessing of existing chats. A cosmetic label and a matching rule therefore deserve different treatment in a report's change log. Peec competitor configuration
Scrunch documents several distinct effects. Changing case-sensitive matching for an existing brand, or changing excluded names, re-evaluates historical responses. Excluding a website instead reclassifies historical citations as third party without re-evaluating the stored responses; the total citation count stays the same while the ownership split changes. Those are different changes to different measures. Scrunch Configuration API
These disclosures support a procurement test, not a product ranking. They do not establish that either vendor preserves every earlier classification, provides a side-by-side historical view, or includes a particular export in every plan. Ask for those capabilities to be demonstrated and included in the proposal if your reporting depends on them.
A fixed set of answers makes the difference visible
Consider this invented example. It is arithmetic for a demonstration, not a result from a product test. A fictional company, Northstar Analytics, also trades as Northstar Labs. Six saved answers have the following contents. None changes between the two classifications.
Scroll horizontally to compare every column. Every column remains available.
| Saved answer | Text or source recorded | Original rule | Corrected rule |
|---|---|---|---|
| A | Names Northstar Analytics | Brand mention | Brand mention |
| B | Names Northstar Labs | Unmatched name | Brand mention |
| C | Names Northstar Labs | Unmatched name | Brand mention |
| D | Names an unrelated Northstar Travel | Not this brand | Not this brand |
| E | Names another supplier | Not this brand | Not this brand |
| F | Links to the company's site without naming it in the answer | Owned-source citation only | Owned-source citation only |
Under the stated rules, the number of answers with a brand mention rises from 1 of 6 to 3 of 6. The collection contains zero new answers. Answer F remains a citation-only case; a URL does not become a textual mention because the company owns it. None of the mention counts establishes that the answer recommended the company.
This is the central visual to ask for in a demo: the same six answer IDs, the old classification and the new classification. A graph showing a higher historical score without those rows cannot explain the correction.
Ask the vendor to demonstrate the correction on saved data
Use a disposable demonstration project or an authorized copy of existing data. Have the vendor explain whether the operation makes new provider requests or incurs charges before changing anything. A sales demonstration should not silently revise your production baseline.
First, save the answer text, question, collection date, engine or product surface, cited URLs and existing entity rules. Keep the exact answer IDs. Also save the vendor's metric definition: the denominator might be answers, questions or another population, and different tools need not calculate it alike.
Next, make one agreed change. Add a genuine alias, exclude a namesake or correct which business owns a cited domain. Ask the vendor to show which records were reclassified and why. An unseen name should be inspectable before it is merged into a tracked competitor; a retailer, product and parent company should not be combined merely because they occur together.
Finally, compare the before and after exports. Check that the saved answer text and collection dates remain the same, identify any newly collected responses separately, and ask whether the original classification can still be reproduced. If the system replaces its old classification, a frozen export may be your only record of what a client actually received.
For a common word or an acquired brand, inspect false matches as well as recovered matches. More matches are not automatically more accurate. Decide the entity boundary first, then apply it consistently to your business and its competitors.
Keep the original report and the corrected comparison
Two views answer different business questions. The original report records what the team reported under the rules in force at the time. A recomputed report applies a consistent corrected definition across the saved history. Both can be useful, but they should carry their rule version and preparation date.
If a new competitor is added, also check whether the metric's comparison set changes. A share calculated against tracked competitors can move when the competitor set changes; an answer-level mention percentage need not. Require the vendor's actual formula before interpreting either movement.
This is separate from changing the questions being measured. Our question-set trend guide covers that problem. Here, the question and saved answer can stay fixed while their classification changes.
The purchase is easier to defend when a vendor can show the affected answer rows, the matching rule and the resulting metric. If it cannot, record that as a reporting limitation before relying on a historical chart in a client or board update.