Contents
The most dangerous SEO proposal is not necessarily the expensive one. It may be the forty-page deck that never tells you which page will change in month one.
I once opened a proposal with “page one” in giant green type. Forty pages later, it still had not named a URL it would change in month one. That was the day I stopped confusing confidence with competence.
The first question I ask when a proposal sounds too polished
Case studies matter, but one ranking cannot reveal competition, demand, commercial value, elapsed time or the resources behind it. Nor does it prove that the consultant can diagnose your website. Crop a screenshot tightly enough and almost any story becomes plausible. A sharp observation, perhaps, but a useful one.
Ask a better question: How will you find the most important problem on my site and decide what comes first? A worthwhile SEO expert connects technology, content, search intent, authority, data and commercial goals, then either implements the work or makes responsibility unmistakably clear.
Hiring an SEO is a little like choosing a contractor. Portfolio photographs are relevant, but you also want to know whether they measured the room, understand the structure, handle variations properly and return when something leaks. “Trust me, it will look amazing” is atmosphere, not a method.
I once opened a proposal with “page one” in giant green type. Forty pages later, it still had not named a URL it would change in month one. That was the day I stopped confusing confidence with competence.
Silence in the first meeting is fine. A consultant who says “I need to inspect that” can be safer than one who has a perfect answer before asking for access.
02
Technical SEO is where a nice pitch meets the plumbing
Excellent content has no stable foundation when Google cannot reliably crawl, render, index or interpret it. A consultant need not be a full-stack engineer. They must, however, identify a technical fault, explain the likely consequence, communicate with developers and verify the repair. That final verb is doing quite a lot of work.
- Agreement between robots.txt, sitemaps, noindex and canonical signals
- Redirects, 404s, migrations, protocol variants and duplicate URLs
- JavaScript that conceals important copy or links
- Core Web Vitals, mobile rendering and template-level experience
- Architecture, pagination, faceted navigation, internal links and orphan pages
- Hreflang, regional URLs and canonicals on multilingual sites
An ecommerce filter can manufacture tens of thousands of URLs and muddle crawling. A migration without complete redirects can disconnect entrances that took years to earn. Four extra blog posts do not repair either failure. If a technical audit is merely a tool export with a new logo, you bought printing—not diagnosis.
Here is the unglamorous test: ask them to open one real URL and explain what Google, a browser and a customer each see. If the explanation jumps straight to a paid tool score, I get nervous. Plumbing first, paint later.
I have seen a whole Hong Kong language folder blocked by one line in robots.txt. No amount of fresh blog copy was going to rescue that week.
03
One red warning is not a diagnosis
Do not count red warnings. Choose one issue the consultant has raised and ask which URLs are affected, how it was confirmed, what it changes for crawling or customers, who fixes it, how the repair is tested and what happens if it waits.
Real diagnosis has scope, evidence and priority, and it admits uncertainty. “Twelve thousand parameter URLs are being discovered, but Search Console sampling is incomplete; I want to inspect logs and index coverage before choosing a canonical or crawling response” is more credible than “duplicate content is critically damaging your site.”
If every 404 supposedly requires a replacement page, every image above 100 KB is declared fatal, or a Lighthouse score is presented as a ranking score, pause. Technical SEO is not a parking-ticket book. Impact and sequence are the actual craft.
A red triangle in an audit is a clue, not a verdict. Pick one warning and keep asking “how do you know?” until the answer reaches a page, a log, a test, or an honest “we do not know yet.”
04
Before keywords, find the job of the page
Keyword research is not a spreadsheet downloaded and sorted by volume. The consultant must decide whether the searcher wants to learn, compare, buy or find a local service, then build or improve the appropriate destination. Get this wrong and beautiful writing merely pushes hard against the wrong door.
| Decision | A useful answer |
|---|---|
| Search intent | Informational, comparative, transactional or local? |
| Results format | Do articles, services, products, lists, videos or maps dominate? |
| Commercial value | Is this demand plausibly near an enquiry or revenue? |
| Existing assets | Improve an old page, consolidate overlap or create something new? |
A service query answered by a beginner’s article, or a pricing query met with a thousand words defining SEO, reflects an intent failure rather than a minor copy problem. During the meeting, provide two or three genuine queries and ask which page should target each one—and why. Weak thinking appears rather quickly.
Try three real searches from your own phone—not a conference-room keyword list. “SEO consultant Hong Kong”, “SEO price” and “Shopify migration” are different little creatures, even when a spreadsheet puts them in one column.
The expert should notice the result-page shape, not merely recite search volume. Sometimes the correct recommendation is to improve an old page and leave a new article unwritten.
05
A report should make somebody do something
Many monthly SEO reports contain twenty charts and no next move. Reporting should become a decision instrument: what changed, plausible reasons, the action to take, and the evidence that would show whether it worked.
Review the title, snippet, brand recognition and promise against the query.
Inspect intent, trust, calls to action, speed, forms and tracking.
Compare evidence, internal links, technical friction and competitor authority.
Tools describe what happened. Expertise explains why it may have happened and what to do. Correlation still needs restraint. A rise can follow new content, but it can also reflect a competitor’s loss, seasonality, a system update or a measurement change. Mature consultants do not autograph every upward curve.
Request an anonymised report. Look for notes, page-level actions, owners and deadlines rather than only site-wide averages. Average position is particularly mischievous: branded queries can improve while valuable non-brand visibility declines, and the average may keep smiling.
I use Search Console to understand queries and pages, GA4 to understand sessions and behaviour, and the CRM to find out whether somebody became a useful lead. They overlap, but they are not triplets. A report that mixes them into one heroic number is doing theatre.
Give me one red movement and one green movement. I want to hear what the expert would do on Monday morning.
06
An industry logo is not a transferable skill
Ecommerce, B2B services, local businesses, education, finance, multilingual publishing and international stores have different failure modes. If every client receives four posts and a bag of links each month, the strategy is thin. Packages simplify proposals; they should not replace judgement.
| Setting | Typical problems | A revealing question |
|---|---|---|
| Ecommerce | Category intent, facets, unavailable products, schema and internal linking at scale | How are demand and links retained when a product is discontinued? |
| B2B services | Long decisions, cases, trust, content and lead quality | How do you distinguish more traffic from better enquiries? |
| Hong Kong local | Traditional Chinese, English, local intent, Maps and entity consistency | How do search habits differ by language? |
| International | Hreflang, canonicals, regional stores, translation and fulfilment differences | Which URL is global, and which is local? |
Direct industry experience helps but is not mandatory. The stronger signal is whether the consultant can transfer experience from a similar structure and business model while identifying your exceptions. A thoughtful outsider can be more dependable than somebody waving a familiar logo and avoiding the process.
A consultant can know the furniture trade and still be lost inside a B2B sales cycle. Conversely, someone from SaaS may understand the shape of your problem better than a specialist who has only repeated the same package. Transferable thinking is the tell.
07
Ask what did not work
A move from position 90 to 8 sounds tremendous. Was it a branded term? How much demand existed? Did the domain already have authority? How many developers, writers and public-relations resources were involved? A destination without a baseline is a holiday photograph; it does not tell you how to travel there.
A useful case identifies the site type, initial problem, objective, actions, period, constraints and outcome. Better still, it includes one attempt that did not work and explains the correction. A person who has apparently won every experiment may simply have removed the other slides.
Ask about attribution. Did the client’s team create the content? Was there a major redesign? Did television advertising lift brand searches at the same time? The SEO may still have made an essential contribution, but one consultant did not personally manufacture the weather.
On my own tracking I have watched terms such as “SEO”, “SEO Consultant”, “SEO Expert” and “SEO顧問” sit at very different baselines—59, 11, 23 and 15 in one snapshot. None of those numbers, alone, is a trophy. The useful question is what page, intent and business action sat behind each one.
Ask for the awkward slide: the experiment that stalled, the link that disappeared, the page that attracted traffic but no enquiries. I trust a scar more than a spotless brochure.
08
The price line is not the scope
A higher price is not proof of effectiveness, and a low one is not automatically economical. Compare scope, depth, content and technical support, analysis, outreach and—most importantly—who performs the work. The smallest type in a proposal can matter more than its cover.
| Model | Suitable use | Main caution |
|---|---|---|
| Monthly retainer | Building organic demand over time | Outputs, strategic revision and execution scope |
| Audit or project | Migration, redesign or a defined problem | Who implements and verifies? |
| Consulting | An existing content and development team | Does the internal team have capacity? |
| Performance fee | A narrowly and fairly defined target | Query quality, attribution and short-term tactics |
| Revenue share | Transactions with robust tracking | Baseline, refunds, brand demand and data access |
Basic support in Hong Kong may begin around HK$5,000 per month, but site size, competition, languages, content volume, technical repairs and outreach can move the price sharply. Treat that as context, not an official tariff. Ask how much is strategy, how much is implementation and what happens outside scope.
An inexpensive service may be honest but advisory; trouble begins when you thought somebody was cooking and actually bought the menu. An expensive agency may also delegate daily work to junior staff. Price must be read beside delivery, review and responsibility.
For context, a Hong Kong engagement around HK$5,000 a month can be perfectly reasonable or wildly insufficient, depending on whether it includes strategy, writing, development and outreach. It is not a market tariff. It is a starting question.
If you are unsure, a small paid test—perhaps around HK$500 for a tightly scoped diagnosis—can reveal far more than a free ninety-page deck. Cheap is not always honest, but vague is never a bargain.
09
Who will actually touch the website?
SEO can involve research, writing, design, development, analytics and promotion. One person need not perform every task. Outsourcing is not inherently suspicious. Unclear ownership and absent quality control are.
Identify the daily contact, strategist, writer, technical implementer and link-building source. If staff change, who preserves context? Can the client inspect drafts, technical tickets and live links? How long do approvals take? These sound administrative, yet they determine whether strategy reaches the website.
I would rather hear, “A partner writer prepares this section from my brief and I review every draft,” than “our professional team handles everything.” Which team? What review? This is not an interrogation. It is turning on the light before working together.
Ask the question bluntly: “Will I speak to the person doing the work?” I have no objection to a partner writer or developer; I object to the mysterious plural—“we”—when nobody can say who owns the next action.
At a small consultancy, one person may wear three hats. That is fine, as long as the hats are named and the capacity is real.
10
If the report never changes a decision, it is wallpaper
Rankings are useful, but they cannot prove performance alone. A usable monthly report connects search demand, landing-page movement, technical health, completed work and business outcomes, then establishes the next priorities.
- Search: impressions, clicks, CTR, queries, ranking distribution and brand/non-brand separation.
- Pages: destinations that rose, stalled or lost visibility, with likely reasons.
- Technical: indexing, crawling, canonicals, redirects, Web Vitals and release risks.
- Work completed: the URLs edited, content published and links earned or placed.
- Commercial: forms, calls, WhatsApp, qualified leads, sales and attributable revenue.
- Decisions: items waiting for client approval, development help or deliberate postponement.
Bad news belongs in the report. Hiding a decline inside a site-wide average wastes the repair window. A consultant should not appear only beside green arrows; the red side is precisely where judgement is most valuable.
A useful report can be slightly boring. It names the page, the signal, the likely cause and the next owner. Give me that over a dashboard full of confetti and no sentence beginning “therefore”.
11
Links: show me the page, not the domain score
If outreach is included, ask which types of sites and pages will link, who writes the surrounding content, whether placement is disclosed, whether it remains visible, and how topical fit and risk are assessed. “One hundred high-DA backlinks” leaves almost every important question unanswered.
Page-level traffic, subject, editorial quality, outbound-link pattern and placement often matter more than domain rating in isolation. A new guest-post URL having zero page-level authority is ordinary—it has just been created. A site on which most articles appear to sell positions is another issue entirely.
Ask how lost links are handled and whether reports supply live URLs. A social profile, directory citation, syndicated release and genuinely edited feature are not the same object. Count them, certainly, but separate the types. One hundred grains of rice and one hundred banknotes are both “100” in a spreadsheet.
When I check a proposed link, I want a live, crawlable, indexed and relevant page—not a domain-rating souvenir. I also look at anchor variety; repeating the same exact-match phrase until it squeaks is not a strategy.
Ask to see one link the expert declined. The refusal often tells you more than the win.
12
Own the keys; agree the goodbye
A long-term relationship needs written boundaries. The agreement should cover minimum term, notice, deliverables, revision limits, link ownership, additional fees, reporting rhythm and a reasonable statement that rankings are not guaranteed.
Search Console, GA4, Tag Manager, Business Profile, CMS and advertising accounts should ideally belong to the client, with appropriate access granted to the consultant. Do not anchor the company to a supplier’s private email. On exit, the keyword map, content, technical records, link list and necessary access should be handed over; retention and deletion also need an agreement.
If full server access is requested, ask whether it is necessary and whether a lower permission can achieve the task. Convenience does not automatically justify unlimited access. This is not the glamorous paragraph. It is, inconveniently, the paragraph everybody wishes they had discussed once a relationship goes wrong.
Do not make exit sound dramatic. People change suppliers, jobs and priorities; that is ordinary life. The unordinary part is discovering that the domain, Search Console and years of content were tied to a former supplier’s private email.
Put the goodbye in writing while everyone is still being polite.
13
Red flags often whisper
- Guaranteed number-one rankings: nobody controls Google, competitors and demand.
- A complete methodological black box: clients should know what changed, the content direction and the types of links involved.
- Link volume without source quality: page relevance, real audience, editorial pattern and risk are avoided.
- Ranking screenshots alone: no indexing, traffic, technical, conversion or next-step analysis.
- The identical package for everybody: no prioritisation by site, resources or objective.
- Every decline blamed on an update: no timeline, affected-page pattern or alternative explanation.
- Pressure to sign a long contract immediately: no sample, ownership or exit detail.
One concern is not always a conviction. A new independent consultant may have fewer public cases but a rigorous method; a large agency may respond slowly because its process is layered. Look for the pattern: the more precise your question becomes, the foggier the answer gets. That is rarely reassuring.
The loud red flags are easy. The quiet one is a proposal that becomes less specific every time you ask a sensible follow-up. Another is the sales call where nobody asks about margins, capacity or what a qualified enquiry looks like.
14
Choose the shape of the team, not the biggest team
Before choosing a supplier type, inspect what the company already has. An organisation with writers, developers and analysts may need a senior consultant who sets priorities. A business with nobody available to implement changes can receive a flawless advisory report and let it retire quietly in Google Drive.
| Model | Typical advantage | Typical constraint |
|---|---|---|
| Independent consultant | Direct communication, lighter overhead and a short distance from strategy to work | Capacity, cover during absence and limited design or engineering resources |
| SEO agency | Multiple disciplines, greater throughput and replacement capacity | Layers of handover, senior involvement and rigid packages |
| In-house hire | Brand knowledge, faster internal follow-up and retained knowledge | Recruitment cost; one hire rarely covers every discipline |
| Hybrid | Internal ownership with external technical, editorial or outreach depth | Unclear boundaries can create a queue in both directions |
No model wins by nature. A small service company may move faster with one experienced consultant editing core pages directly. A large retailer may require engineering, editorial operations, analytics and digital PR at once, which quickly exceeds one person’s ceiling. The relevant question is not the thickness of a company directory; it is whether the available capability matches today’s constraint.
Do not ask only how many employees an agency has. Ask which people will enter your account, what each contributes and who makes the final judgement. A ten-person team represented solely by an account executive who forwards a monthly PDF still feels, from the client’s chair, like one person.
I once saw a ten-person agency behave like one tired account manager because every decision travelled through the same inbox. Headcount is not delivery. Ask who can edit, who can approve and who can cover a sick day.
15
What should the first 90 days leave behind?
“SEO takes time” is true and occasionally used as a blanket. Waiting is not the only possible activity. Diagnosis, baselines, page changes and tracking can be visible before rankings move decisively.
- Days 1–30: confirm access, conversions, technical risks, competition, query intent and priority pages. Protect existing value before expanding.
- Days 31–60: resolve high-impact blockers, complete briefs or edits for core pages, and establish internal-linking, content and outreach workflows.
- Days 61–90: publish or release the first changes, verify indexing and measurement, observe early query and page signals, then adjust the next batch.
The order should vary. A migration may devote most of the quarter to URL mapping, redirects and quality assurance. A modest local service site can often improve service pages, local entities and conversion tracking sooner. A sensible plan explains the difference rather than printing the months as an immutable calendar.
Ask for first-quarter work as “deliverable + owner + dependency + validation.” “Complete technical audit” is insufficient unless somebody is assigned to repair findings, verify them and record deliberate deferrals. Even before a ranking takes flight, you can see whether the programme is accumulating useful change or orbiting the same deck.
By day 90 I want a cleaner baseline, a few shipped changes, a tracking sanity check and a reasoned next batch. I do not need a victory parade. I need evidence that the programme can move without being pushed down the MTR stairs every week.
16
Your CMS has opinions
WordPress, Shopify, Magento, headless systems and custom platforms all permit title edits; their practical constraints differ enormously. Plugins overwrite canonicals, themes duplicate headings, faceted navigation generates URLs and JavaScript delays important content. A strategy detached from the CMS becomes theoretical with impressive speed.
Ask which platforms the consultant has operated, then request one platform-specific failure and the method used to verify its repair. Shopify collection and product canonical behaviour, Magento filters, WordPress taxonomies and multilingual plugins each have distinct temperaments. Concrete experience is harder to imitate than “we support all platforms.”
Establish who may change production. Mature workflows use backups, staging, change records, quality assurance and rollback—especially for migrations, URL rules, robots directives and canonicals. A consultant who can identify faults but cannot work with development planning may leave every correction in the permanent state of “recommended.”
They do not need to write every line of code. A useful development ticket describes the present behaviour, desired behaviour, affected templates, acceptance test and risk. If the developer can act without guessing, the recommendation is genuinely implementable.
Every CMS has a tiny trapdoor. Shopify has its collection logic, WordPress has taxonomies and plugins, Elementor has templates that can quietly multiply headings. Ask for one trapdoor the expert has met in real life.
“We support every platform” is a broad sentence. Show me the ticket.
17
A bilingual site is two markets, not two copies
Hong Kong websites commonly publish Traditional Chinese and English, sometimes with Simplified Chinese as well. They can describe one service without serving identical demand. English searches may use formal service names; Chinese users mix written Chinese, Cantonese expressions and English abbreviations. Location, price and trust cues shift too.
A capable consultant checks hreflang, canonicals, language switching, URL structure and page correspondence, but technical correctness is only the floor. The harder decision is which pages require localisation, which need accurate translation and which language lacks enough demand to justify a page. Mechanical duplication creates maintenance and a collection of pages with no search role.
Give the candidate a set of English and Chinese queries. Ask whether they belong to the same page type and whether titles and examples should differ. “Add hreflang” solves only half the problem. In a smaller market, moving one step away from intent can remove a substantial share of useful demand.
Local SEO is not repeating “Hong Kong” ten times. Name, address, telephone, service area, Business Profile, reviews, map category, landing pages and actual operations should agree. A service-area business should not invent an address to chase map visibility. Clever for a moment; troublesome for much longer.
Traditional Chinese and English are not merely two labels on a language switcher. A Hong Kong reader may change language, device and search phrasing halfway through a purchase. The expert should care about URL pairing, self-canonicals, hreflang, local entities and whether both pages deserve to exist.
18
Who is allowed to kill a bad draft?
Many retainers promise a monthly article count without explaining topic selection, subject-matter input, fact checking, revision allowance or rejection. Output is easy to count, so it becomes a substitute for value. An article delivered punctually to the wrong intent remains on the wrong platform.
Request a content-brief example. A useful brief covers reader stage, page purpose, required questions, acceptable sources, internal links, call to action, brand boundaries and claims that require care—not just a primary keyword and word count. Finance, medical and legal topics need appropriate review.
AI can assist research, organisation and drafting. The consultant should still locate human judgement: verification, first-hand experience, brand voice, examples, privacy and final accountability. “Our AI is advanced, so drafts need no editing” does not describe efficiency. It converts risk into a speed claim.
Ask about maintenance too. Old prices, discontinued products, changed rules and broken links do not remain correct because a page once ranked. A content library needs updating, consolidation and, occasionally, removal. The person willing to prevent an unnecessary article can be more valuable than the one forever promising four more.
AI is useful as a fast junior assistant—good at organising notes, occasionally cheeky with facts, never the managing editor. Ask who checks claims, who interviews the subject expert and who can say “not this week”. That last person protects your brand.
19
If the lead tracking is leaky, every argument is fake
Before work begins, confirm that Search Console, GA4, Tag Manager, forms, calls, WhatsApp, ecommerce transactions and the CRM connect at a useful basic level. Otherwise, after three months, rankings may be the only evidence available while “did we get customers?” becomes a contest of memory.
An SEO need not be a complete analytics engineer, but should define conversions, test events and notice obvious leakage. Thank-you pages that count again on refresh, identical events for every WhatsApp button and daily internal tests are small holes that turn a report into a bucket.
Attribution also requires humility. A buyer may discover a guide organically, click a branded advertisement two weeks later and convert through WhatsApp. Last-click reporting awards the final step. At minimum, compare first touch, assisted paths, brand/non-brand demand and lead quality, while accepting that no model reconstructs a human decision perfectly.
Ask, “If organic sessions rise 40 per cent and CRM leads do not, what do you inspect first?” A strong answer works through tracking, intent, landing pages, location, device, form friction and sales follow-up. It does not automatically prescribe another three months of waiting.
There is no shame in saying the measurement is unfinished. There is shame in celebrating a 40 per cent session lift when the form event fires twice and the sales team records nothing. An expert should be able to find the leak without blaming the weather.
20
The retainer is only the cash part
An SEO quote lists supplier fees. The actual programme also consumes interviews, editorial approval, legal review, design, engineering, translation, photography, CRM maintenance and management time. These are not reasons to avoid the work; they are reasons to see it before the proposal becomes a schedule.
Ask for client dependencies: subject-matter hours each month, content approver, development cycle, image owner and compliance sign-off. If the company cannot provide weekly expert interviews, design a more realistic collection method instead of allowing every delay to become “pending client.”
Testing and opportunity costs count too. Changing fifty titles on one day is fast but makes messaging effects difficult to separate. Continually producing new URLs may divide existing authority. The cost is not merely cash; it includes engineering capacity and the comparison you can no longer run.
A mature proposal names exclusions. Are photography, publication fees, paid tools, extra languages and substantial development separate? Ambiguity does not remove cost. It postpones the conversation until everybody is a little less patient.
The invoice is only the visible cash. Your team will also spend interview hours, approval time, developer time and a little patience. If the proposal needs six internal signatures for one paragraph, the promised publishing rhythm is fiction.
21
References: ask the awkward question
For a substantial engagement, request one or two reference conversations where possible. Not every case can be public, and confidentiality is legitimate. Ask specific questions: were deliverables timely, were problems disclosed early, could recommendations be implemented, did reports improve decisions, and what was unexpectedly difficult?
“Were you satisfied?” invites a courteous yes. “If you hired them again, which contract term would you change?” tends to produce information. Notice how the candidate protects client data, too. Someone who opens another company’s complete analytics account to impress you may one day use yours in the same performance.
| Criterion | Example weight | Evidence to seek |
|---|---|---|
| Diagnosis and priority | 25% | Connects issue, impact and sequence |
| Execution | 20% | Clear ownership, process, CMS familiarity and QA |
| Intent and commercial understanding | 20% | Queries, pages and revenue paths agree |
| Reporting and communication | 15% | Data becomes action; bad news remains visible |
| Transparency and risk | 10% | Links, accounts, contract and limits are explicit |
| Price and fit | 10% | Scope is sufficient and dependencies are feasible |
Adjust the weighting to the situation, but write it before the final pitch. It stops meeting atmosphere from becoming the whole decision. Charisma is welcome; it simply should not receive 80 per cent. Put comparable evidence and one principal risk beside each candidate. The choice often becomes surprisingly plain.
Do not ask a former client only “were you happy?” Most people say yes and return to their calendar. Ask, “What would you change in the agreement if we hired them again?” That question has elbows.
22
A proposal is useful when it admits what it cannot do
If three candidates receive three different briefs, their quotations cannot be compared fairly. Before requesting proposals, list the site, languages, markets, principal products, current visibility, known technical constraints, internal capacity, previous SEO and genuine commercial objective. Imperfect information is acceptable. A shared map is the point.
Request current assumptions, priority opportunities, first-phase scope, deliverables, client dependencies, measurement, risk and explicit exclusions. Expectations should concern controllable work and reasonable indicators, not a promise that a query will occupy a fixed position on a fixed date. Organic search is not a reserved advertising slot.
Read the exclusions carefully. Does on-page work omit development? Does an article fee cover interviews, upload, imagery and three languages? Are publication fees included in outreach or invoiced for every placement? Work left unwritten has a curious habit of becoming the other party’s responsibility.
Compare setup, retainer, content, engineering, publication and tool fees in separate columns, then calculate a six- or twelve-month total. A cheap first month followed by an extra fee for every action is not comparable with a higher retainer that includes implementation. The proposal’s job is to expose the trade-off, not conceal it beneath one bold number.
Give each candidate the same messy brief: languages, market, CMS, previous work, budget and the one outcome that matters. Then compare assumptions and exclusions in separate columns. The cheapest quote often becomes less cheap when you add the work it quietly left on your desk.
23
The WhatsApp question usually beats another meeting
More meetings are not automatically better. What matters is whether a question reaches its owner when necessary and whether decisions survive the call. A weekly hour in which nobody approves the new title is simply a group observation of the same traffic jam.
Establish a modest operating rhythm: everyday channel, incident escalation, editorial approver, technical sign-off, reporting date and quarterly reprioritisation. Give meaningful actions an owner and a date. “We will follow up” is otherwise an extraordinarily courteous form of disappearance.
Define consultant autonomy. May they update meta descriptions and internal links directly? Must claims, legal copy, URLs and templates receive approval? Sending every reversible edit to senior management slows the programme; changing production without review creates another kind of problem. Mature teams tier approval by impact and reversibility.
A valuable consultant translates complexity into a choice: “Option A is faster but leaves duplicate URLs; option B takes two more weeks and removes the template cause. I recommend B because 8,000 pages are involved.” Jargon deposited without a decision is not sophistication. Expertise need not make the client feel excluded.
Agree where small questions live—email, Slack, WhatsApp, whatever your team actually opens. At 11:47 p.m. a client may not need a seminar; they need to know whether a broken form is being handled. A response rhythm is part of the service, not office decoration.
24
Traffic fell. Now what?
Visibility moves. Ranking systems, competitors, seasonality, deployments, tracking failures, results-page formats and stale content can all contribute. The useful skill is not predicting every update. It is narrowing causes with discipline when something changes.
- Validate the data: do Search Console, analytics and rank tracking agree, and is the site functioning?
- Find the pattern: which date, directory, template, language, device and query group moved?
- Align the timeline: were there releases, migrations, robots changes, new templates, content edits or unusual links?
- Inspect external context: did competitors, demand, result features or confirmed updates change?
- Stabilise before rebuilding: repair known faults and avoid rewriting the whole site in panic.
Ask for a past loss and how it was investigated. A persuasive answer includes a mistaken hypothesis and how it was eliminated, not only “we refreshed content and recovered.” Prescribing more posts and links for every condition resembles one medicine for every patient. It occasionally works; it is not diagnosis.
Incident communication matters as much as analysis. Management needs impact, confirmed facts, open questions, temporary measures and the next update time. A chart delivered after two silent weeks may technically be a report. As collaboration, it arrives rather late.
When a ranking drops, I want a timeline before a theory: release, template, robots, seasonality, competitor, tracking. “Google update” may eventually be part of the answer, but it is a terrible opening sentence.
The best incident note I have read contained a wrong hypothesis. That detail made the rest believable.
25
Imagine it failed—then tell me why
Proposals naturally describe the success path. Before deciding, ask each candidate for the three most likely reasons the programme might disappoint after six months. The question brings competition, engineering capacity, editorial approval, authority, measurement and the product itself into the room—not only what the consultant controls.
A credible answer may be uncomfortable: “Your current approval takes six weeks, so the proposed output is impossible,” or “Search demand in this niche imposes a ceiling on leads.” Discomfort is not pessimism. Discovering a constraint before signing is cheaper than renaming it an unexpected event afterwards.
Then conduct a small pre-mortem. Assume the engagement failed, and record the cause, early indicator and prevention. Insufficient engineering may first appear as untouched tickets in month one; declining outreach quality as increasingly irrelevant publications; poor lead quality as traffic concentrated on beginner queries. Monitored risk becomes more useful than a generic disclaimer.
Finally ask an ordinary question: can our company implement this person’s recommendations? The cleverest strategy that cannot be executed retains a commercial value of zero. Capability, budget, speed and internal reality must fit. Otherwise the proposal improves beautifully and the website does not. End of story.
Ask the candidate to disappoint you on purpose: “What could make this plan fail?” If the answer is only “the market”, the pre-mortem has not started. Good experts name the dependency that keeps them awake.
26
Let the expert interview you
The client should not ask every question. If a consultant recommends monthly articles and links before asking about the business model, margin, market and internal resources, they do not yet know where useful demand should go. A rapid answer can simply mean the package was written before the call.
A serious candidate will usually investigate questions such as:
- Which products, services and regions should grow—and which should not?
- What counts as a qualified enquiry, and how long and valuable is a typical sale?
- Which channels have worked, and do offline activity, advertising or seasonality influence brand search?
- Who maintains the website, how quickly can technical changes ship, and are staging and backups available?
- Who supplies subject expertise, approves content and reviews regulatory or brand risk?
- Is the immediate objective new demand, conversion, regional expansion or protection of existing visibility?
They should also request evidence rather than accept “our SEO is poor” as a complete diagnosis. Search Console reveals queries and pages; analytics and CRM data describe sessions and leads; the CMS and a crawl expose structure. Access can be granted in stages and according to need. There is no reason to hand over the entire server password during the first video call.
Useful questions uncover constraints. What is the budget? How long will management tolerate uncertainty? If content requires six levels of approval, should proposed output be reduced? If the most profitable service has no clear differentiation, should the offer and conversion argument improve before more traffic arrives? SEO cannot perform a spell on a product nobody wants, however tactfully the sentence is written.
Watch whether the answers alter the proposal. Some suppliers ask ten questions and return with the standard package. Another may ask four, then connect each answer to a priority page, measurement choice and dependency. Question count is not the competition. Continuity between discovery and recommendation is.
There should also be room for disagreement. A capable expert may challenge the preferred keyword because the results show a different intent, advise preserving an unfashionable old page because it earns links, or recommend paid search for an urgent launch while SEO develops. Agreement feels smooth in a sales call; useful advice occasionally introduces friction.
Finally, an expert may decline the engagement. If budget, timetable, risk or expectations are fundamentally incompatible, proposing a smaller audit or a later start can be more responsible than forcing every prospect into a retainer. The willingness to say “this is not the right project yet” is itself evidence of judgement.
Notice whether your answers change the proposal. If they ask ten questions and return the same package they sell everyone, the interview was decoration. If four answers move the page priorities, measurement and timeline, somebody was listening.
And yes, a good expert may tell you to wait. That is not a rejection; sometimes it is the first useful piece of advice.
27
Try a small paid test before a long marriage
- How will you identify the first problem worth solving?
- What is the largest search opportunity and risk in this market?
- Which pages would change in the first three months, and why?
- Which tasks do you perform, and which go to colleagues or partners?
- How do you separate service, article, category and comparison intent?
- How do you protect current visibility during a redesign or migration?
- How will the report connect search, traffic and conversion?
- Where do links come from, and how is page quality evaluated?
- Can I see an anonymised report and one complete case?
- If expected signals do not appear, how will you re-diagnose?
- Which people, approvals and development resources must we provide?
- How are accounts, content and data handed over on exit?
If uncertainty remains, commission a paid audit, keyword map or single-page project rather than demanding free speculative labour. Test the quality of thought, communication, punctuality and usefulness of the output. A small project cannot simulate a year, but it contains more nutrition than another recital of “we are experts.”
A good SEO expert makes the reason for each step visible. The choice is not who sounds most confident, uses the most presentation animation or quotes the lowest fee. Choose the person who diagnoses accurately, makes trade-offs, executes and connects search work to commercial goals. The engagement should leave less fog—not another monthly report nobody can interpret.
A small paid test cannot predict a year. It can show whether the person thinks clearly, writes down assumptions, meets a deadline and tells you what remains unknown. I would rather buy one honest week than marry a beautiful promise for twelve months.