Keyword Research Without Lying to Yourself

Keyword research goes wrong in a specific way: people collect the words they *wish* customers typed, build pages for them, and then can't understand why nothing ranks. The tool told them 5,000 searches a month. The tool wasn't lying, exactly. The person just stopped asking the only question that matters — would *my* store, today, actually win this? This is part two of the honest SEO playbook. Today: free brainstorming, the platform's AI research (with its estimates treated as estimates), clustering by intent, and tracking that doesn't flatter you. ## Free brainstorming that beats most paid tools Before you touch any keyword tool, spend an hour where the real language lives: - **Autocomplete.** Type your head term ("berbere spice"), read the suggestions, write down every one you could actually serve. Then do it for your top 5 products. - **"People also ask" boxes.** These are questions with buyer intent behind them: "Is berbere spicier than cayenne?" "What is berbere made of?" Each one is a potential article or a product-page section. - **Competitor product titles.** The sellers ranking above you on marketplaces have already done research. Their title patterns are keyword lists in disguise. - **Your own inbox.** This is the goldmine. Every message a customer sent — "does this scarf fit over hijab," "how long does delivery to Adama take" — is a search phrase written by a real buyer. Ten customer emails will give you keywords no tool surfaces, because no tool has your customers. Do this before any tool. Tools rank and estimate; only this tells you what your buyers *sound like*. ## Use the AI research — without worshipping it BirrStore's platform includes **AI keyword research** built for store owners. Type a seed term, get back related keywords with search volume figures. Use it — it saves hours. But read the labels. **The volumes are modeled estimates, and the platform says so right on them.** That's not a disclaimer to ignore; it's the correct description of what any keyword volume is. Nobody — not Google, not any tool — publishes real search counts. Every number anywhere is a model. The honest difference is that this one tells you the truth about itself. So use the numbers the right way: - **For relative comparison:** "berbere spice" at 1,900/mo modeled is clearly bigger than "authentic Ethiopian berbere" at 320/mo modeled. That ordering is real and useful. - **Not as promises:** a modeled 1,900/mo does not mean 1,900 people will find *your* page, or that you can beat everyone ranking for it today. > An estimate is a compass, not a map. It tells you direction — never how far, and never what's in the way. The platform carries the estimate label through to reports and keyword exports too, so you don't accidentally present a modeled number to a partner or in a plan as if it were measured fact. Keep the label when you share. "Estimated" is not weakness; it's what makes the rest of your plan credible. ## Cluster by intent, not by volume Here's where most research dies: a flat list of 200 keywords, sorted by volume, worked top-down. Wrong shape. Sort by **what the searcher is trying to do**: - **Buy-now intent** — "buy berbere online," "berbere 250g price." These people want a product page. Map them to products. - **Compare intent** — "best berbere brand," "berbere vs cayenne." These want a category page or a comparison article. Map them to categories and a few honest comparison posts. - **Learn intent** — "what is berbere made of," "how to store berbere." These want an article. They're earlier in the journey, they share more, and they become buyers weeks later. Map them to the blog. Three buckets, every keyword assigned to a page you will actually create. Any keyword you can't honestly attach to a page you're willing to build gets deleted. That deletion is the whole discipline — a list of 40 keywords mapped to real pages beats a list of 400 keywords mapped to nothing. ## Track honestly When you start tracking rankings, three rules: - **Check weekly, not daily.** Rankings wobble day to day for reasons that mean nothing. Daily checking is anxiety with a dashboard. - **Expect the staircase, not the rocket.** Small stores move in steps: nowhere, then page 3, then page 2, then the bottom of page 1 for specific phrases. Weeks, not days. - **Keep the estimate labels in any export.** When you report "we target keywords worth 4,200/mo," the honest version is "estimated 4,200/mo, modeled." The version of you in six months will be grateful the numbers came with their honesty attached. ## Do this now - **Harvest 20 phrases** from autocomplete, people-also-ask, competitor titles, and your own customer inbox - **Run your top 3 seed terms through the platform's AI keyword research** and note that the volumes are modeled estimates - **Sort every keyword into buy / compare / learn** and map each cluster to a product, category, or article - **Delete every keyword you can't attach to a page you'll actually build** - **Pick your 5 starter keywords** — where demand is real and competition looks beatable - **Set a weekly ranking check** on your calendar, and leave the daily dashboard alone Part three is where research becomes repair: **running the site audit and fixing what it finds in the right order** — broken pages first, missing titles second, thin content third, alt text fourth, speed last. And why that order specifically. The audit will hand you 200 problems; you'll fix the six that matter this week.