Chapter 4 — Quiz

Twenty-four self-check questions: multiple choice and short answer. Answer everything before opening the key at the bottom. Aim to explain, not just recognize — the short-answer items are where the understanding lives.

Multiple choice

1. "Things, not strings" refers to Google's shift toward: - A) counting keywords more accurately - B) understanding real-world entities and their relationships - C) ranking by page length - D) matching exact phrases only

2. The Knowledge Graph is best described as: - A) a list of the top 200 ranking factors - B) a database of entities and the relationships between them - C) Google's paid-advertising system - D) a tool for measuring page speed

3. A knowledge panel is: - A) a ranking factor you can buy - B) the visible, on-SERP readout of the Knowledge Graph for a recognized entity - C) the same thing as a featured snippet - D) a Search Console report

4. In TF-IDF, a term appearing in nearly every document in the collection gets: - A) the highest weight - B) roughly zero inverse-document-frequency weight - C) a negative ranking penalty - D) doubled term frequency

5. The BM25 idea that repeating a term many more times adds less and less value is called: - A) length normalization - B) inverse document frequency - C) term-frequency saturation - D) co-occurrence

6. BM25's length normalization exists so that: - A) longer pages always rank higher - B) shorter pages are always penalized - C) a longer document doesn't out-rank a focused one merely for containing a term more times - D) pages must hit an exact word count

7. "LSI keywords," as sold by many SEO tools, are: - A) a confirmed Google ranking factor - B) a made-up term; LSI is a real but web-unsuited IR technique Google says it doesn't use - C) the official name for schema properties - D) required for a knowledge panel

8. Co-occurrence helps Google because: - A) it counts keyword density - B) related concepts reliably appear together across the web, so covering them signals genuine depth - C) it measures backlinks - D) it replaces indexing

9. Topical authority is: - A) a public score Google publishes for every site - B) the same as third-party Domain Authority - C) subject-specific recognition earned by covering a topic broadly and deeply (a strong working model, not a confirmed score) - D) guaranteed by adding schema

10. BERT (a Chapter 2 system used here) most improved Google's handling of: - A) image file sizes - B) words in context, including prepositions and word order - C) backlink counting - D) crawl budget

11. The sameAs property is used to: - A) mark two pages as duplicates for canonicalization - B) link your entity to authoritative references describing the same entity - C) request faster crawling - D) set a page's title tag

12. Which is the least ambiguous way to tell Google which entity a page is about? - A) surrounding context (prose) - B) co-occurrence across the web - C) structured data (JSON-LD) that states it outright - D) the page's word count

13. A page can rank for a query it never literally contains because Google: - A) ignores the query text entirely - B) resolves the query to its entity and intent and matches meaning, not just strings - C) always shows the longest page - D) randomizes results

14. Disambiguation matters most when: - A) your brand name collides with another entity - B) your page loads slowly - C) you have too many backlinks - D) your sitemap is missing

15. Which statement about the claim "Google ranks by TF-IDF/BM25" is most honest? - A) It is confirmed by Google - B) The formulas are real IR mathematics, but Google has never said it ranks by them; at most something in BM25's spirit is one component among many - C) The formulas are fake - D) BM25 is Google's complete ranking algorithm

Short answer

16. Give three failure modes of pure string matching that "understanding things" fixes.

17. Explain, in two sentences, why keyword stuffing "stopped working" using a concept from §4.4.

18. What does a knowledge panel tell you about how well Google understands you as an entity, and why can't you control it directly?

19. Distinguish breadth from depth in topical coverage, and explain why you need both.

20. A page comprehensively covers "how a water heater works" but does not rank for "emergency water heater repair near me." Name the concept (from Chapter 3) that explains why, and say whether more topical depth fixes it.

21. Why is co-occurrence not an instruction to paste a list of "related words" onto your page?

22. Name the three sources (from §4.2) Google leans on to corroborate entity facts, and explain why "corroboration, not declaration" is the operative principle.

23. At which stage of the Chapter 1 pipeline do entities and topical coverage help — and why can't they rescue a page that fails an earlier stage?

24. Your client wants a "guaranteed knowledge panel." Give the honest response and one legitimate action that actually improves the odds.

Answer key **1.** B. **2.** B. **3.** B. **4.** B. **5.** C. **6.** C. **7.** B. **8.** B. **9.** C. **10.** B. **11.** B. **12.** C. **13.** B. **14.** A. **15.** B. **16.** (1) *Synonyms* — "fix my heater" vs. a "furnace repair" page; the strings differ but the need is the same. (2) *Ambiguity* — one string ("jaguar," "apple") points at several entities. (3) *Conversational/complex queries* — natural questions contain few "keywords" but a clear meaning. Understanding entities and meaning handles all three. **17.** BM25's *term-frequency saturation* means each additional use of a word adds less and less, so hammering a phrase past a few natural uses barely helps; its *length normalization* penalizes padding a page just to fit more instances. The math itself removes the payoff stuffing once had. **18.** A panel is a readout of the facts Google is confident it knows about you as a thing; a rich panel means Google understands you well, a thin or absent one means it is still guessing. You can't control it directly because it is assembled from many corroborating sources — you influence it by being consistently and accurately described everywhere (and, if you're the subject, by claiming it and *suggesting* corrections). **19.** *Breadth* is covering the whole topic — every meaningful subtopic and question; *depth* is covering each of those thoroughly and genuinely. Breadth without depth is a pile of thin pages (which hurts you); depth without breadth is one good page with no reinforcing coverage. Topical authority needs both. **20.** *Intent mismatch* (Chapter 3): the searcher wants to hire someone now (transactional/local), and the page is informational. More depth does not fix a wrong page *type* — the SERP will favor local/service results, so the explainer competes for the wrong slot. **21.** Because co-occurrence is a pattern Google *learned from millions of genuinely expert documents* — the related concepts appear on good pages because those pages actually explain the subject. Bolting a word list on reverses the causation; it adds no genuine depth and Google's systems increasingly tell the two apart. **22.** Wikipedia and Wikidata, structured data published across the web, and Google's own crawl (plus other authoritative/licensed databases). "Corroboration, not declaration" holds because Google believes an entity's facts when *independent* sources agree — you can't simply assert facts on your own site (or fabricate a Wikipedia entry) and expect them accepted. **23.** The *ranking* stage — relevance and topical coverage are how a page wins the query-time contest. They can't help a page that fails discovery, crawling, rendering, or indexing, because an un-indexed page isn't in the contest at all. Diagnosis before treatment (Chapter 1). **24.** Honest response: no one can guarantee a knowledge panel any more than a #1 ranking — panels appear for entities Google is confident it understands, and Google decides. A legitimate action: make your entity consistently and accurately described across your site and authoritative profiles (and, for a local business, a complete, verified Google Business Profile), so Google has corroborated facts to build from.