Bol.com Keyword & Niche Research (bol-niche-scout)

Identify winnable niches on bol.com before committing to a product. Provide a few seed keywords; the actor expands them into real bol.com search suggestions and scores every keyword on demand versus competition — comparable to established keyword tools for Etsy, applied to the largest marketplace of the Netherlands and Belgium. One run produces one decision document.

Built for bol sellers (print-on-demand, private label, dropshipping) who need to decide which product or keyword to pursue.

What you get

One row per keyword, keyword column first, no arrays, no half-empty cells. Illustrative example:

keyword	opportunity_score	result_count	review_velocity_30d	review_moat_median_top10	bol_self_pct	sponsored_density_pct	price_median
yogamat extra dik	71	1 842	18	24	0	20	29.95
bureau organizer bamboe	63	957	9	41	0	10	24.99
yogamat	38	28 410	6	312	40	30	27.95

Use the Opportunities dataset view for the ranked shortlist, or All columns (ordered) for the full export in a fixed, keyword-first column order (the raw CSV export sorts columns alphabetically; the views preserve the intended order).

All output columns

• keyword — the search term (seeds plus expanded suggestions)
• opportunity_score (0–100) — higher means more demand signal and less competition
• demand_score, competition_score — the two components of the formula (see below)
• result_count — total number of products bol returns for this search
• review_velocity_30d — estimated reviews in the last 30 days across the analyzed top-N products (proxy for sales velocity)
• review_velocity_source — exact (newest-reviews API), estimate (dates from the reviews embedded in the product page; a lower bound) or none
• review_moat_median_top10 — median review count of the top-10 results: how entrenched the incumbents are
• bol_self_pct — share of the top-10 sold by bol itself; competing directly with bol is rarely viable, so this weighs heavily
• sponsored_density_pct — share of the top-10 that is a sponsored listing
• price_min / price_median / price_max — price band of the analyzed top-N
• avg_rating_top_n, fast_delivery_pct — quality and delivery pressure in the top-N
• suggest_rank — position at which this keyword appeared in its parent's autocomplete (1 = top suggestion; 0 for your own seeds)
• depth, parent_keyword — position in the expansion tree
• top1_title, top1_url, serp_url — direct links for manual verification
• domain, language, analyzed_top_n — run parameters for reproducibility

Input

{
    "seeds": ["yogamat", "bureau organizer"],
    "domain": "NL",
    "language": "nl",
    "depth": 2,
    "maxKeywords": 40,
    "analyzeTopN": 5,
    "proxyConfiguration": {
        "useApifyProxy": true,
        "apifyProxyGroups": ["RESIDENTIAL"],
        "apifyProxyCountry": "NL"
    }
}

• domain (required): NL (bol.com Nederland) or BE (bol.com België)
• language: nl (default) or fr; French is only available for BE
• depth: number of suggestion levels to expand (seeds are depth 0)
• maxKeywords: hard cap on the number of analyzed keywords, seeds included
• analyzeTopN: products per keyword that receive review-velocity analysis

How keyword expansion stays relevant

Bol's autocomplete pads short queries with popular but unrelated terms. The actor counters this with two mechanisms, both enabled by default:

• Strict suggestion filter (strictSuggestions): bol's response reports how many characters of the query were actually matched; suggestions that do not match the full keyword (and do not contain it) are discarded.
• A–Z prefix probing (prefixProbes): every seed is additionally queried as "seed a" through "seed z" — a proven technique from Etsy keyword tools. This is the main source of long-tail variants such as "yogamat extra dik" or "yogamat antislip".

A small number of substring near-matches may remain (a seed like "mat" also surfaces "matras"); their own SERP metrics make them easy to dismiss.

How the opportunity score works

The formula is deliberately simple and fully documented:

demand_score      (0–50) = suggest points (0–20) + review-velocity points (0–30)
competition_score (0–50) = review moat (0–15) + result count (0–15)
                           + sold-by-bol (0–10) + sponsored density (0–10)
opportunity_score        = clamp(50 + demand_score − competition_score, 0, 100)

Component	Weight	Calculation
Autocomplete presence	0–20	20 − 2 × (suggest_rank − 1), minimum 2. Seeds score a neutral 10.
Review velocity	0–30	1 point per review per 30 days, summed over the top-N products, capped at 30.
Review moat	0–15	log10(median reviews top-10 + 1) × 5; a median of 1 000 reviews reaches the cap.
Result count	0–15	log10(result_count + 1) × 2.5; one million results reaches the cap.
Sold by bol	0–10	Proportional to the share of the top-10 sold by bol itself.
Sponsored density	0–10	Proportional to the sponsored share of the top-10.

As a rule of thumb: scores above 60 indicate real demand with beatable competition, around 50 is neutral or unknown, and below 40 the ranking is entrenched, ad-saturated, or dominated by bol itself.

Proxy configuration

bol.com's bot protection blocks common datacenter IP ranges, so the actor defaults to Apify Proxy RESIDENTIAL with country NL (suitable for both the NL and BE shop). Measured cost: a 40-keyword run transfers roughly 25 MB of residential data, about $0.20 per run (≈ $0.005–0.007 per keyword) with negligible compute. Datacenter proxies are not supported in practice.

Use as an AI-agent tool (MCP)

The actor works as a tool for AI agents through the Apify MCP server: connect an MCP-capable client (Claude, Claude Code, Cursor, custom agents) to

https://mcp.apify.com?tools=<username>/bol-niche-scout

authenticating with your Apify account (OAuth) or an API token. The agent then calls the actor with the same input as above and reads the scored keywords from the resulting dataset. For Claude Code:

$claude mcp add --transport http apify "https://mcp.apify.com?tools=<username>/bol-niche-scout"

Recommended agent inputs — agents prefer short calls, so trade breadth for speed: depth: 1, maxKeywords: 15, analyzeTopN: 3 finishes in roughly 1–2 minutes and is usually enough to answer "is this niche worth a closer look?". A follow-up full run (depth: 2, maxKeywords: 40, analyzeTopN: 5) can then deepen the shortlist.

Reliability: built-in self-test

Running the actor with { "selfTest": true } analyzes one fixed keyword and fails with a non-zero exit code and a clear error message if any part of bol.com's page structure has changed. Recommended setup: an Apify schedule that runs the self-test daily with a failure notification, so breakage is detected before users encounter it.

Fair use & disclaimer

• Not affiliated with bol.com. "bol" and "bol.com" are trademarks of bol.com b.v.
• Only publicly visible data is used: search suggestions, search result pages and product pages. No login, no buyer accounts, and no personal data of reviewers — only review counts and dates.
• The actor operates at a deliberate, human-like pace and caches within the run (no page is fetched twice). Keep maxKeywords reasonable.
• Review velocity is a proxy for sales velocity, not a sales report. Only a small share of buyers leaves a review; interpret it as a relative signal between keywords.

---

Bol.com keyword research & niche-onderzoek (Nederlands)

De eerste keyword- en niche-researchtool voor bol.com. Voor wie zoekt op bol.com keyword research, bol niche vinden of welk product verkopen op bol: deze actor beantwoordt die vraag met actuele bol-data in plaats van onderbuikgevoel. Een vergelijkbaar hulpmiddel zoals eRank voor Etsy bestond tot nu toe niet voor bol.com.

Wat doet de tool?

1. Zoeksuggesties uitbreiden — de opgegeven seed-keywords worden recursief uitgebreid via de autocomplete van bol.com, aangevuld met A–Z-prefix-probing voor long-tail-varianten. Wat bol suggereert, wordt daadwerkelijk gezocht: dat vormt het vraagsignaal.
2. Concurrentie meten per keyword — aantal zoekresultaten, mediaan aantal reviews in de top-10 (de toetredingsdrempel), het aandeel topposities dat bol zelf bezet (direct concurreren met bol is zelden haalbaar en weegt daarom zwaar mee), advertentiedruk en de prijsband.
3. Verkoopsnelheid schatten — review-velocity: het aantal nieuwe reviews van de topproducten in de afgelopen 30 dagen, als indicator voor omzetsnelheid.
4. Opportunity-score (0–100) — hoge vraag, lage toetredingsdrempel, beperkte aanwezigheid van bol zelf en weinig advertenties leveren een hoge score op. De volledige weging staat hierboven gedocumenteerd.

Voor wie?

Bol-sellers (print-on-demand, private label, dropshipping) die onderbouwd willen beslissen welk product of keyword zij gaan voeren. Eén run levert een gesorteerd overzicht: welke niches bieden ruimte, welke zijn verzadigd.

Voorbeeld

De seeds ["yogamat", "bureau organizer"] op bol.com (NL) leveren binnen enkele minuten 40 gescoorde keywords op, gesorteerd op kans. Bovenaan verschijnen long-tail- keywords zoals "yogamat extra dik" of "bureau organizer bamboe" met aantoonbare vraag en een lage toetredingsdrempel; onderaan de verzadigde hoofdkeywords.

Disclaimer

Niet geaffilieerd met bol.com. De tool gebruikt uitsluitend publieke data (zoeksuggesties, zoekresultaten, productpagina's), zonder login en zonder persoonsgegevens van reviewers — alleen aantallen en datums. De tool werkt in een beheerst, menselijk tempo.