Media Monitoring Taxonomy Drift: When Volume Falls Quietly

A media monitoring taxonomy degrades the moment an institution renames what you are tracking, and it fails without an error: mentions keep happening, the query stops matching, and the volume line falls as though the conversation stopped. Philippine examples sit in the budget cycle (SAGIP to SAGP) and in the statute book (NEDA to DEPDev). Treat taxonomy review as a dated monthly practice with a named owner.

Every monitoring setup rests on a quiet assumption: that if it stopped working, somebody would get an error. Keyword lists do not fail that way. They degrade the day an institution renames the thing you are watching, and the first symptom reads like good news, a volume line coming gently down.

A media monitoring taxonomy is the set of strings standing between public conversation and the chart you brief from. Rename a fund mid-appropriations, or rename a department by statute, and the query stops matching a conversation that carries on without it. Both happened here inside eighteen months.

One Renamed Fund, Two Acronyms, One Blind Chart

A House appropriations subcommittee can do this to a multibillion-peso programme in one session. Reporting on the 2026 national budget, GMA News recorded the cut and the rename together: SAGIP, the Strengthening Assistance for Government Infrastructure and Social Programs, was cut from P80.86 billion to P45 billion for foreign-assisted projects, and the remainder "will be now called the Strengthening Assistance for Government Programs (SAGP)." One fund, two acronyms, one appropriations season.

The same mechanism surfaced again at the end of September, in a different kind of renaming. On 29 September 2026 the US president released an executive order directing executive departments and agencies to use "super intelligence," or SI, rather than AI in their official communications, in place of the technologies and systems covered by the existing AI definition in US law. Poynter's fact check names the collision: the industry already uses "superintelligence" for a system exceeding human cognition in nearly all fields that does not yet exist, traceable to philosopher Nick Bostrom's 1998 framing, while practitioners today more commonly say AGI. Poynter reports little to no change in how AI industry leaders refer to the technology since the directive.

One term, two meanings, depending on the speaker. Syracuse University's Baobao Zhang, who studies the politics of AI, told Poynter that "it's really important that we get our definition right with any attempts at regulating this very fast-moving technology," and that systems falling outside a definition can still cause serious harm. Definitional drift is a governance problem before it is a style quibble, and it compounds the shift in how monitoring is being rebuilt for AI-era discovery.

Silent Failure: No Error, No Alert, Just a Falling Line

A renaming splits one topic into two unlinked strings. Mentions keep happening, the query stops matching them, and the dashboard reports a decline that is an instrumentation artefact, not a change in public conversation. Nothing is built to notice, because nothing failed. Think of a rain gauge somebody moved under the eaves: still reporting on schedule, no longer measuring the weather.

Official usage and media usage also separate for far longer than anyone expects, and the gap is measurable. Nieman Lab's analysis of Nexis newspaper data found US newspapers using "Gulf of Mexico" 1.48 times as often as "Gulf of America" in February 2025, the first full month after the renaming, 6.27 times as often by August 2025 and 8.234 times by September. Counting only stories that used one name or the other, excluding stories about the renaming itself, the old name appeared 10.28 times more often that September. That is one author's database query, reported again by MediaPost, so read it as a single newsroom analysis rather than an industry-wide finding.

The Philippine case points the other way. The Manila Times reported Senate Finance chair Sherwin Gatchalian saying the lump-sum fund was taken out of the enacted budget, "tinanggal na nga natin sa 2026 budget," and that January 2026 report still spells the programme out under its original long name, months after the House renamed it. A list cut over to SAGP alone would have undercounted coverage that was still happening. A list left on SAGIP would have missed the new string.

The second-order damage reaches the principal. Sentiment baselines and share-of-voice comparisons computed across the break are wrong in both directions: the post-rename period looks quieter than it was, and the recovery after the fix reads as a spike somebody will attribute to a campaign. Automated sentiment across English, Filipino and Taglish is directional rather than settled, so an unmarked break compounds a reading that was already approximate. Any series offered as board-ready measurement should carry that break on its face.

How Many Strings Does One Agency Actually Need?

Sometimes the rename arrives with a statute and a commencement date. GMA News reported Republic Act 12145, signed 10 April 2025, transforming the National Economic and Development Authority into the Department of Economy, Planning, and Development, effective 15 days after publication. NEDA had been the standing keyword for decades of economic coverage. One entity, two strings, and a statutory date that belongs on the chart.

Two names for one subject can also carry very different measured volumes with no formal rename at all. A peer-reviewed comparison of Google relative search volume published in PLOS ONE in 2015 put "global warming" at a mean of 25.33 plus or minus 2.05 (n = 8,464) against 7.97 plus or minus 0.74 (n = 8,283) for "climate change," roughly a 3.2 times gap for one underlying topic, converging to near-identical volumes by 2014. The authors concluded that "the public press determines how a term is used." Those figures are a decade old and describe a usage shift, not an official renaming, so treat them as foundational context rather than current measurement.

For a Philippine office the working unit is a set, not a pair: an English name, a Filipino rendering, an acronym, a hashtag spelling, the short form a radio anchor actually says, and Bisaya or Ilocano variants where regional coverage carries the story. No sourced figure exists for how local coverage divides across those registers, so enumerate the set per programme instead of estimating the split, and keep provincial and vernacular sources in the same view as the broadsheets so a missing variant shows up as a thin region rather than a quiet month.

Put Media Monitoring Taxonomy Review on the Monthly Calendar

Let's turn that into something you can put in a workplan, with the authority stated plainly: no published guidance from a measurement standards body on keyword or query governance surfaced this cycle, so what follows is Media Meter's practice aligned with monitoring-industry guidance rather than a certified standard.

Two vendors publish usable guidance, disclosure attached. CARMA, which sells monitoring and evaluation and so describes a category it competes in, writes in its note on monitoring setup that "periodically checking in on your monitoring keywords and phrases is an important step," warns that "terminology can shift rapidly" when the topics are in the news, and asks directly whether organisations in your field have been renamed or rebranded since the last review. Fullintel, also a monitoring vendor, publishes a cadence in its guide to running media monitoring: "Keywords aren't 'set and forget,'" with a monthly review of problematic terms, a quarterly comprehensive review with stakeholders and an annual overhaul, plus a change-control rule, "implement a buddy system where a second team member reviews all keyword changes before they go live."

Five artefacts turn that into keyword list upkeep an office can audit:

• A term-change log with two dates. When the official change took effect, and when coverage actually followed it.
• Alias mapping, held as one topic. Old string to new, every register enumerated, so the series stays continuous.
• A visible re-baseline marker at the exact point on the chart where the list changed.
• A named analyst owner for the monthly review, with a second reviewer before any change goes live.
• A line in the daily summary naming which terms changed this cycle.

Give the review a defined input. The Presidential Communications Office is the body whose memorandum circulars prescribe mandated official wording and content across national agencies, GOCCs and state universities, which makes its published issuances list the natural first stop each month, alongside your own department's memoranda. Read that list for what it is: the circulars posted there prescribe ceremonial content, not terminology, including Memorandum Circular No. 117 of 29 April 2026 on official music videos at flag ceremonies and Memorandum Circular No. 52 of 9 June 2024 on the Bagong Pilipinas Hymn and Pledge. The input is the stream of issuances, not those two documents.

What to Hand the Secretary When the Term Changes Mid-Cycle

One page, answering the question before it is asked: the old term and the new one, the effective date, the date coverage followed, what the chart does at the break, and what the volume would have been on either list alone. Then the sentence a mention count can never produce, that the decline between these two weeks is instrumentation, with the corrected series beside it.

That page is also the honest answer to the procurement question, because it is a deliverable rather than a feature. No Philippine agency or LGU was found to have published its own clipping taxonomy or monitoring methodology, so there is no local precedent to copy and nothing to point at in a terms of reference. Whoever writes taxonomy upkeep into theirs first sets the bar for the next office, which makes accreditation and procurement readiness worth settling before a rename, not during one. For what a maintained taxonomy looks like on a briefing page, start with the sector brief library.

Key takeaways

• Treat a falling volume line as a question about the query first and public conversation second.
• Log both dates for every rename, the official effectivity and the date coverage followed. The Philippine record shows those diverging by months.
• Enumerate the full alias set per programme: English name, Filipino rendering, acronym, hashtag spelling, broadcast short form, regional variants.
• Mark the re-baseline on the chart and brief the break, so nobody attributes the recovery to a campaign.
• Schedule the taxonomy review monthly with a named owner and a second reviewer, with new PCO circulars and agency memoranda as the standing input.

Frequently asked questions

What is a media monitoring taxonomy, and why does it drift?

It is the structured set of keywords, topics and aliases deciding what your monitoring returns. It drifts when an institution renames a programme, an agency or a concept: the old string stops matching new coverage, the new string has no history, and the split happens without any error in the pipeline.

How often should a government media monitoring keyword list be reviewed?

Monthly, with quarterly and annual passes layered on top. That cadence comes from monitoring-industry guidance rather than a measurement standard. Fullintel, a monitoring vendor, publishes a monthly adjustment, a quarterly stakeholder review and an annual overhaul, and CARMA, which sells monitoring and evaluation, names renaming and rebranding as explicit review triggers.

How do you tell an instrumentation drop from a real fall in coverage?

Check whether the subject was renamed in the period, then re-run the old and new terms separately over the same window. If the combined series holds while either string alone declines, the drop is the query. The SAGIP case shows the pattern: coverage kept using the original long name months after the rename.

Can you still compare sentiment and share of voice across a rename?

Only with the break marked and the alias mapping applied, and even then read the result as directional. Automated sentiment across English, Filipino and Taglish is approximate at best, not settled, so a baseline computed across an unmarked term change compounds an estimate that was already soft in both directions.

About Media Meter

Media Meter supports government communicators with accredited, procurement-ready media monitoring: daily briefings, sentiment tracking and issue monitoring across national, provincial and vernacular media. The part that keeps a chart honest through a rename is the analyst maintaining the taxonomy, not the tool storing it.