I did keyword research for my savings app. Then I actually looked at the search results.

In March I found a keyword with 500,000 monthly searches and almost no competition. Google Keyword Planner scored "application budget" at a competition index of 10 out of 100 in the French market. My app is a budgeting-adjacent savings tool with a French locale. I remember writing, in my own research notes, that the French-speaking market appeared underserved and that this could be the fastest path to organic traffic.

Three months later I removed that keyword from every title and heading on my site. Not because I couldn't rank for it eventually, but because I finally looked at the actual search results page and understood that no web app will ever rank for it. The traffic was real. It just was never addressed to me.

This is the story of how I found that out, what it cost, and what I target now instead. If you run a web app and you have ever picked keywords from a planner tool without opening the results page, this will probably sting a little. It should. It stung me.

The app, briefly

Akluma is a free savings tracker. You set a goal, it builds a saving schedule, and it keeps the plan honest when life changes it. It is a web app (a PWA, technically), built solo with Laravel, and it speaks three languages: English, Turkish, and French. Each locale has its own URLs, its own copy, its own SEO surface.

The relevant part for this story: it is a web app. You use it in a browser. There was no app store listing when this research happened. Hold that thought, because the entire lesson lives inside it.

At the start of this story, my Google Search Console looked like the "before" photo in an ad: 2 pages indexed, 0 backlinks, and every single query that showed my site was someone typing "akluma". People who already knew the name. Nobody else.

The research that felt like winning

I did the keyword research properly, or so I thought. Google Keyword Planner, seed keywords per language, full CSV exports, filtering out competitor brands and irrelevant niches. The numbers came back looking like this:

• English: "saving apps" at 50,000 searches per month, competition index 3. "savings tracker" at 5,000 per month, index 19.
• French: "application budget" at 500,000 per month, index 10. "épargne" also at 500,000.
• Turkish: "para biriktirme" (saving money) at 500 per month, index 5.

Those French numbers did not look real. Half a million monthly searches with almost nobody bidding against them. I concluded the Francophone market was underserved by English-first competitors, picked my targets, and rewrote the site around them.

One thing I did get right at this stage: I did not translate the copy between languages. Each locale got its own keyword targets and its own writing. The English page targeted "savings tracker", the French page targeted "application budget", the Turkish page targeted "para biriktirme". Transcreation instead of translation. That decision survived everything that came after, even though most of the specific keywords did not.

It half-worked, and the half that worked fooled me

The content overhaul had visible effects within days. Indexed pages went from 2 to 10. When I shipped a new About page in three languages, Google indexed all three variants the same day they deployed, which for a tiny site is remarkably fast. The first three backlinks appeared. Pages I had not even touched got pulled into the index along with the ones I rewrote.

So the machinery was working. Google was crawling, reading, indexing, and clearly paying more attention than before.

And yet: zero non-brand queries. Weeks after the rewrite, the only searches that surfaced my site were still "akluma" and one heroic misspelling. Google had seen my "application budget" page and had decided, silently, that it was not what those 500,000 people wanted.

The standard advice at this point is patience. New content takes months to rank. That advice is true and it is also a wonderful place for a wrong strategy to hide, because "wait longer" and "this will never work" look identical from the inside for the first few months.

The question that broke it open

What saved me was a question I now ask about every research tool: what does this tool not see?

Google Keyword Planner is a Google Ads tool. It reports what people type into google.com. It knows nothing about what people type into the Play Store or App Store search bars. And for app-shaped products, that blind spot is not a detail: roughly 65 percent of app discovery happens through in-store search. Web search is about a quarter of the pie. My entire research covered the small slice and was structurally silent about the rest.

That was blind spot number one, and it explained what I should do next (app store research, which I did later, more on that below). But it did not yet explain the dead French keyword. That took blind spot number two:

Volume is not intent. A planner tool tells you how many people type a phrase. It tells you nothing about what Google has decided those people want. The only place that decision is visible is the results page itself.

So I did the least sophisticated SEO research imaginable. I opened an incognito window and searched for my own keywords.

Twelve searches, one survivor

The method, so you can reproduce it: incognito window, locale forced through URL parameters. gl=us&hl=en for English results as an American user would see them, gl=fr&hl=fr for France, gl=tr&hl=tr for Turkey. For each target keyword, catalogue page 1: who ranks, which SERP features appear, what the related searches suggest.

I verified 12 keywords this way. Here is what page 1 actually looked like.

"saving apps" (English, 50K/month). An AI Overview recommending specific native apps by name (Acorns, Rocket Money, YNAB). A direct Apple App Store listing. A Google Play editorial result. Then "best apps" listicles from finance sites. Every single result assumes you want to install something on your phone. A web app has no shot here, at any level of content quality, because the query does not mean what its words say. "Saving apps" means "help me choose a native app to download".

"application budget" (French, 500K/month). My goldmine. This one was worse, and Google made the intent so explicit it was almost funny. At the top of the page, Google renders app-category filter chips: "Sans banque", "Familial", "Enveloppe", "Couple". Google has built a native-app comparison UI directly into the results page. The first result is a Play Store listing with 482,203 reviews. Below that, "meilleures applications" listicles, a direct App Store listing, and a "people also ask" box where every question is about apps. Half a million searches a month, and effectively all of them are people choosing which app to install from a store I was not in.

"épargne" (French, 500K/month). Wrong in a completely different way. Page 1 is the French government (economie.gouv.fr), Banque de France, Wikipedia, and a map of local Caisse d'Épargne bank branches. People searching this want definitions, interest rates, and savings accounts. Not a tool. And even if the intent fit, the competition is literally the French state.

The Turkish cluster ("para biriktirme" and variants). Informational intent. Page 1 is owned by the content-marketing blogs of major Turkish banks: Yapı Kredi, Akbank, Garanti BBVA. People want methods and strategies for saving money, and the banks write those articles at industrial scale. An app product page does not belong in that conversation.

"savings tracker" (English, 5K/month). The one survivor. Page 1 is a genuine mix: Excel spreadsheet tutorials on YouTube, printable templates, a government savings calculator (Investor.gov), one bank's web tool, and yes, some app store listings, but no single format dominates. When spreadsheets and printables rank, the door is open for a web tool, because you can genuinely be better than a spreadsheet. The related searches even included "savings tracker online", which is about as direct a web-tool intent signal as exists.

Final score: 12 keywords verified, 1 reachable. The keyword strategy I had built the site around was about 92 percent fiction.

The pivot, in three directions

English: shrink the target and stop lying to myself. Every keyword containing "app" came out of the titles and H1s. The homepage H1 went from "The Saving App..." to "The Savings Tracker...", and the long-tail "savings tracker online" got worked into the copy. A 5K keyword I can actually contest beats a 50K keyword I cannot.

French: redo the research, intent-first this time. Since every original French target was dead, I picked five new candidates and ran them through the same SERP verification before writing a single word of copy. The results were a nice tour of the ways a keyword can fail:

• "suivi épargne en ligne" (online savings tracking): winner. The SERP has the same mixed DIY profile as English "savings tracker": Excel templates, spreadsheet tutorials, web calculators, exactly one app listing. Reachable.
• "calculateur épargne" (savings calculator): 10 out of 10 results are web tools, several from banks and the French market regulator. Zero app listings. A strong secondary.
• "objectif épargne" (savings goal): dead, for a reason no tool would ever surface. "Objectif Épargne" turns out to be a branded Swiss Life insurance product, so page 1 is contract PDFs and subscription forms.
• "gérer son épargne en ligne" (manage your savings online): dead. "En ligne" reads as online banking in this context, so the SERP is banks and account-comparison sites.

The French app name had literally been "Akluma — Application Budget". It is now "Akluma — Suivi d'Épargne". I had named my product after the keyword it could never win.

The store question: one codebase, two keyword universes. Remember the 65 percent of discovery that Keyword Planner cannot see? I eventually researched that side too (AppTweak plus manual Play Store autocomplete), ahead of shipping the app to the Play Store as a TWA. The French results are my favorite finding of the entire project. "Application budget", 500K a month on Google, is also wrong for the Play Store: it is the expense-tracker category, dominated by an established app, with the highest ranking difficulty I saw anywhere. The right Play Store keyword for a French savings app is "tirelire" (piggy bank), which gets maybe 50 searches a month on Google and is therefore invisible to SEO research, but on the Play Store has zero competition and matches the product exactly. Bonus: Quebec searches "tirelire" at four times France's volume, still with zero competition.

Same product. Same language. Same codebase. The web search channel and the app store channel want completely different words.

Where it stands

The corrected copy shipped to production on July 14, 2026, so I will not pretend to show you a rankings graph. What I can verify today: the indexing surge was real but not all of it stuck (2 pages before the rewrite, 10 at the peak, 6 as I write this), the first backlinks exist, and the site no longer targets keywords it cannot win. Whether "savings tracker" and "suivi épargne en ligne" actually materialize into rankings is a question for months from now, because the remaining bottleneck is authority, and a domain with a handful of backlinks earns trust slowly.

Part of fixing that bottleneck, in full honesty, is writing posts like this one.

What I would tell you to do differently

1. Keyword volume without SERP verification is a vanity metric. The planner tells you how many people type the phrase. Page 1 tells you what Google decided they meant. Only page 1 is ground truth.
2. Ask what your research tool cannot see. Keyword Planner sees about a quarter of app-discovery demand and says nothing about the rest. Every tool has a blind spot shaped like its data source.
3. If you are a web app, "app" in a query is a wall. Those SERPs belong to the app stores. Do not fight them. Find the queries where spreadsheets, printables, and calculators rank, because you can beat a spreadsheet.
4. Transcreate, do not translate. Each locale has its own winnable keywords, and they are probably not translations of each other.
5. SEO and ASO can diverge completely for the same product. My French web keyword and my French Play Store keyword share zero words. If I had assumed one research pass covered both channels, I would have shipped a store listing optimized for a keyword the store does not use.

The manual verification took an afternoon per language. The strategy it replaced had quietly burned three months. If you take one thing from this: before you rewrite your site around a keyword, open an incognito tab and go look at who already lives on page 1. They are telling you exactly whether you are invited.

Akluma is live at akluma.com if you would like to watch a savings plan hold up under real life. And if you are curious about the engineering side of that promise, I wrote about it separately: the savings plan that rewrites itself.