Työkaluvertailu.fi is a Finnish-language software comparison site for small businesses. I built it as my own business in April 2026, and in this series I report openly what the site contains, where its traffic comes from and what it has earned. Every figure has a date.
In short (as of November 6, 2026):
- What I built: 142 tools in 19 categories, 68 comparison articles and 76 pages that put two products head to head.
- What happened: clicks from Google grew from 44 in April to 586 in October, and the average position improved from 28.1 to 16.9.
- What it earned: €689.57 (excluding VAT) from four payments. That is revenue, not profit: the cost table shows only the direct costs of hosting and search data.
- What did not work: the site does not rise in generic buying searches, and in October the largest affiliate network passed on 742 clicks without a single paying customer.
Why I built a comparison site
When a small business switches accounting software or looks for a better invoicing tool, it compares prices, terms and other users’ experiences. The searcher does not want an advertisement. They want to know whether the product suits their own company and what it really costs.
I wanted to find out two things: can one person build a comparison site that readers trust, and can it make money without giving up that trust. Many small businesses face the same question in their own field: how does a website earn a searcher’s trust and turn it into revenue.
What the site contains
As of November 6, 2026:
| Content | Count | What it means |
|---|---|---|
| Tools | 142 | Each has its own page: who it suits, who it does not, the price in euros with a source, pros, cons and user experiences |
| Categories | 19 | For example accounting, invoicing, CRM, web hosting, e-commerce and time tracking |
| Comparison articles | 68 | The products of a category side by side, for example free accounting software |
| Head-to-head pages | 76 | Two products against each other, for a searcher who has already narrowed the options |
The site follows three rules:
- Every price has a source and a check date. A new version of the site is not published if the source of a price or a claim has expired.
- Researched is not tested. A product is marked tested only if it has been used. The others are researched, and the page says so.
- No “best” labels without grounds. In October 2026 I removed every tool-level “best” label (291 of them) and replaced it with a “who it suits” description. Every tool has the same button whether or not the product pays a commission.
The technology in brief
The site is a static Astro site on Netlify, and the tool data lives as files in the same code. On November 6, 2026 PageSpeed Insights gave one comparison page 100/100 on both mobile and desktop (mobile LCP 1.2 s). That is a single run on one page, not a site-wide average.
The same code also builds a Swedish sister site, programjamforelse.se, whose sitemap was submitted to Google on October 20, 2026.
Traffic from Google by month
Clicks, impressions and positions are from Google Search Console (retrieved November 6, 2026), and revenue counts only payments received. April starts on the 8th, the day the site was published.
| Month | Clicks from Google | Impressions | Average position | Revenue in the month |
|---|---|---|---|---|
| April 2026 | 44 | 13,189 | 28.1 | €0 |
| May 2026 | 201 | 46,629 | 24.9 | €0 |
| June 2026 | 293 | 54,656 | 24.1 | €0 |
| July 2026 | 290 | 77,811 | 22.0 | €0 |
| August 2026 | 431 | 96,603 | 20.6 | €0 |
| September 2026 | 475 | 72,884 | 17.2 | €400 |
| October 2026 | 586 | 93,853 | 16.9 | €139.57 |
| November 2026 | – | – | – | €150 |
Clicks grew every month except July, and the average position improved every month. The best single day so far was November 2, 2026, with 49 clicks.
In September impressions fell even though clicks grew. About a third of the drop came from searches for Isolta: they had 9,567 impressions in August and 1,665 in September, while the page stayed in positions 7–9. The ranking did not weaken; fewer people searched. Most of the remaining lost impressions were in positions 21 and below, which bring almost no clicks. Part of September’s improvement in average position (from 20.6 to 17.2) is therefore arithmetic, because there were fewer impressions far down the results. The growth in clicks is the more reliable sign of progress.
Where the clicks come from
Most clicks come from searches where the reader has already shortlisted a product. The most common form is “[product name] kokemuksia” (reviews), followed by the product name and “[product name] hinta” (price). For review searches the site often ranks 1–4, with a click-through rate often between 5 and 24 % (August 1 – November 3, 2026).
Generic buying searches such as “kirjanpito-ohjelma” (accounting software) are a different story. In October the site’s position for that search fell from 32.8 to 47.3. The most likely cause is authority: on October 2, 2026 ten domains linked to the site, and none of the links was followed (dofollow). That is not a certain cause: Google’s AI answers and search intent may also explain the pattern.
Revenue: €689.57 in seven months
Revenue is shown excluding VAT and counted from payments received. I do not name the partners.
| Type | Amount | Paid |
|---|---|---|
| Commission from an affiliate program, August (1 new account) | €50.00 | September 3, 2026 |
| Commercial partnership, labelled on the site (6-month pilot) | €350.00 | September 30, 2026 |
| Payment from an affiliate network: a content reward and one commission | €139.57 | October 16, 2026 |
| Commission from the same program as in August, September (3 new accounts) | €150.00 | November 2, 2026 |
| Total | €689.57 | as of November 6, 2026 |
The commercial partnership is a six-month pilot that is clearly labelled on the site. The money does not affect the review or the order, and no position is promised.
Note the split: more than half of the revenue came from one partnership and one content reward, not from click commissions. With commissions alone the result so far would be about €210.
What the site cost
| Cost | Amount |
|---|---|
| Domain tyokaluvertailu.fi (first year) | €12.90 + VAT |
| Hosting (Netlify free tier) | €0 |
| Server-side tracking (free tier) | €0 |
| Search data (DataForSEO, May 27 – November 6, 2026) | about $10 |
The free tier has its limits. Netlify’s 300 monthly build minutes are shared by five of my sites. By October 24, 290 of them had been used in October, 156 by this site. I moved the rest of October’s releases to my own machine, but the quota was exceeded anyway: on October 26 usage was at 125%.
What did not work
- Four SEO experiments in September proved nothing. I tested the email marketing, invoicing, CRM and web hosting pages and ended the experiments early on October 2, 2026. Clicks on the invoicing pages fell from 18 to 6, and the CRM page did not rise.
- Editing titles and descriptions was not the answer. When a page did not rise, the cause was search intent and the layout of the results page, not the title.
- Clicks are not sales. In October the largest affiliate network passed on 742 clicks, which led to 2 sign-ups and 0 paying customers. The 3 new accounts from another program in September did not repeat in October.
- Vendors do not offer payment on their own. In October four software vendors got in touch, and none of them offered a paid partnership.
- Not every product has an affiliate program. On November 6, 2026 there were 43 active affiliate programs (on 45 tools). The other 97 tools have no active affiliate link, and 6 of them are waiting for an application to be processed. Some of the products that get the most clicks are among them.
What you can learn from this for your own business
- Specific searches bring buyers before generic ones. The “reviews” and “price” searches brought clicks first, because the searcher is already close to a decision. For a service business, a search like “[service] price [city]” can reach the top long before “[service]” alone.
- Traffic is not the result. 742 clicks and 0 sales show that the metric worth choosing is revenue, not clicks.
- Trust has to be built into rules. A price source, a check date and a “researched is not tested” label work because they are part of the site’s structure, not something the writer has to remember.
- Links are the most likely next bottleneck. In generic buying searches the most likely limit is authority, and new pages alone do not fix that.
In the next part
In the next part I will report how the sections built around search intent performed, what the Swedish sister site shows with 56 days of data (December 16, 2026) and how the six-month commercial pilot is going.
If you want the same method for your own website, see SEO consulting.