Methodology · Model 1.0
How we estimate the rankings
Nobody publishes how many of each graphics card, mouse or laptop gets sold across every shop. So the shares on Ask Bestie Computers are estimates, built each month from the best sales figures we can get in each region. Every ranking says which sources it rests on.
What the number means
"Est. 31.4% of builds" next to a CPU means that, of the CPUs bought for new PC builds in that market that month, we estimate 31.4% were that model. For peripherals and laptops it's the share of purchases in the category. It's a share of what people are buying now, not of every PC in use.
For now: CPUs and graphics cards by what people own
We don't have shop sales data yet. Until we do, CPUs and graphics cards are ranked by the estimated share of PCs in use that have each model, from hardware surveys and benchmark databases: the Steam Hardware Survey, PassMark and Blender Open Data. Those rankings say "est. % of PCs in use" and name their sources. They show what people own, so older cards like the RTX 3060 rank high even though few people buy one new today.
- A ranking needs at least 2 sources. One alone is too skewed: Blender's benchmark, for example, is mostly run by people who render, and Steam only counts gamers.
- Sources are averaged with their weights. Each counts a list's share of its whole population, so unlisted and integrated graphics still count towards the total.
- These rankings are global. The surveys don't split by country in a way we can use.
- The change from last month is only shown when both months use the same sources, so a new source joining doesn't look like a product rising or falling.
- Confidence is low with two sources and medium with three.
Every other category is marked coming soon until we have real purchase data for it. The rest of this page describes the sales-based model those rankings will use.
Where the numbers come from
Two kinds of data go in. Purchase data (shop bestseller lists, unit counts, shopper interest on price comparison sites) shows what people are buying now, and it drives the estimate. Installed-base data (hardware surveys and benchmark databases) shows what people already own. It changes slowly, so we only use it as a check and a light steadying weight.
Purchase data
| Source | Market | What we use | Status |
|---|---|---|---|
| Inet | Nordics | Bestseller order per category | Waiting for access |
| Webhallen | Nordics | Bestseller order per category | Waiting for access |
| Komplett | Nordics | Bestseller order per category (SE, NO, DK) | Waiting for access |
| Proshop | Nordics | Bestseller order per category (SE, NO, DK) | Waiting for access |
| NetOnNet | Nordics | Bestseller order per category | Waiting for access |
| Elgiganten / Elkjøp | Nordics | Bestseller order per category; strongest for laptops and peripherals | Waiting for access |
| Power | Nordics | Bestseller order per category | Waiting for access |
| Prisjakt | Nordics | Popularity across Nordic shops (shopper interest), via the partner API | Waiting for access |
| Mindfactory | Europe | Units sold per product (public counter) | Waiting for access |
| Geizhals | Europe | Popularity across DACH shops (shopper interest) | Waiting for access |
| Galaxus | Europe | Bestseller order per category | Planned |
| Alza | Europe | Bestseller order per category | Planned |
| Amazon.de (via Keepa) | Europe | Best Sellers Rank and units bought in the past month | Planned |
| Tweakers Pricewatch | Europe | Pricewatch popularity (product page views), NL/BE — one-time manual snapshot 2026-11-13, not automated; stopgap until Keepa | In use |
| Amazon.com (via Keepa) | North America | Best Sellers Rank and units bought in the past month | Planned |
| Best Buy | North America | Best-selling rank via the Products API | Planned |
| Newegg | North America | Best-seller order per category | Planned |
| Amazon.co.uk (via Keepa) | United Kingdom | Best Sellers Rank and units bought in the past month | Planned |
| Amazon.com.au (via Keepa) | Australia | Best Sellers Rank and units bought in the past month | Planned |
Installed base (checks)
| Source | Market | What we use | Status |
|---|---|---|---|
| Steam Hardware Survey | Global | Share of Steam users per graphics card (monthly survey) | In use |
| PassMark | Global | Share of CPU benchmark submissions, last 90 days | In use |
| PassMark (memory) | Global | Share of PassMark memory benchmark submissions per module | In use |
| Blender Open Data | Global | CPUs and graphics cards in Blender benchmark runs during the month (CC0) | In use |
We only use data we're allowed to use. We don't scrape sites whose terms forbid it, and we don't use build-sharing sites or forums without an agreement.
How a ranking is made
- Every list becomes a share. Unit counts and percentages are divided by the list's total. A bestseller list only gives an order, so position 1 counts for twice position 2 and three times position 3, and the whole list is taken to cover about 75% of the category's sales. Several versions of one product (an ASUS and an MSI RTX 5070, say) are added together.
- Lists are averaged with weights. Unit counts and shares count in full; a bestseller order counts 0.7 as much, because it says less. Some sources carry more weight than others, for example a shop that publishes real unit counts.
- Each market leans on its own shops. For a market's ranking, its own shops make up 80% of the weight, purchase data from other markets 12% and installed-base data 8%. In the Nordics that means Nordic shops first, with a sprinkle of everything else. If one of the groups has no data, its share goes to the others.
- The global ranking weights markets by size. Purchase data makes up 90%, split between markets by their rough share of PC hardware spending (North America 42%, Europe 30%, United Kingdom 10%, Nordics 8%, Australia 5%). Installed-base data makes up 10%.
- One odd month can't swing it. Each month's figure is 70% the new data and 30% last month's estimate.
- Small shares are dropped. Anything under 0.5% is left out, and the rest is rounded to one decimal.
If a market has no purchase data of its own for a category yet, it shows the global ranking, and the ranking says so.
Confidence
- High: three or more purchase sources including real unit counts or shares, or four or more sources of any kind.
- Medium: two purchase sources, or one with real unit counts.
- Low: a single bestseller list.
Checks
Every month we compare the result with the installed-base data. A product near the top of the sales estimate that nobody seems to own is either brand new or a sign that a product name was matched wrongly, and we look at it by hand before publishing. Product names from shops are matched to our catalogue automatically; anything ambiguous is left out rather than guessed.
What it isn't
It isn't a lab test or a review score. A product can be popular and not right for you, and the ranking only covers the products in our catalogue. Shares are rounded estimates; treat a gap of a point or two as a tie.