Proxies for Michaels meet Akamai at the door and a coupon-driven price behind it. On 24 August 2026 we sent michaels.com a plain request from a home connection and from a datacenter server, and both returned Akamai's Access Denied. The wall is Akamai, judged on the request from either address, so a real browser is required. Behind it sits a craft retailer where the shelf price is almost beside the point, because coupons do the real work.
Michaels is the largest US arts-and-crafts retailer, and like Kohl's, its data is complicated by a promotion engine: frequent percent-off coupons, category sales and weekly-ad deals make the effective price a moving, coupon-dependent number. That makes the proxy job an Akamai-defended, promotion-aware read.
What proxies work best for Michaels?
US rotating residential proxies behind a real browser, region-matched to the stores you read. A static ISP proxy for a logged-in Michaels Rewards account. Datacenter proxies get Access Denied on contact.
Akamai judged the request
Both test requests got Access Denied from Akamai with no product. Akamai decided on the request, the TLS handshake and the header order a real browser sends with client hints (what is JA3/JA4 fingerprinting, how to scrape past Akamai). Past the door, Akamai's sensor and the _abck cookie keep scoring. The proxy gives the browser a US residential address; the browser satisfies the sensor.
The coupon is the price
Michaels runs on coupons, and this is the central complication for a price dataset. A percent-off coupon, a category sale and a weekly-ad deal can stack, and they change weekly, so a snapshot of the shelf price on a random day misrepresents what a Michaels shopper pays. Comparing Michaels to a flat-price competitor without the active coupon is comparing two different things. The dataset has to read the active promotion and coupon alongside the price, capturing the banner and terms, not just the price field. This is the Kohl's Cash problem in a different category.
Stock is the second axis: availability keys to a selected store, inferred from location, so pin the store, match the exit to its region, and record it.
Shelf price
The listed number
rarely what is paid
Weekly ad
Category sales
changes weekly
Percent-off coupon
Stacks on top
40% off is common
Effective price
What a shopper pays
well below shelf
Which proxy type fits which job
| Michaels job | Proxy type | Why |
|---|---|---|
| Price and promotion monitoring | US rotating residential, region-matched | Akamai denies datacenter; coupons drive the price |
| Weekly-ad and coupon tracking | Rotating residential | Reads banners and terms alongside price |
| Store-level stock | Sticky residential per region | Availability keys to the store |
| Category and catalog collection | Rotating residential | Volume across a pool |
| Michaels Rewards account, cart | Static ISP | The session must keep one address |
| Parser tests against saved HTML | Datacenter or free | The live site denies bare clients |
Rotating residential IPs pass Akamai's reputation check, and ISP proxies hold one address for a Rewards account. The two are compared in rotating vs static residential proxies.
Setup
Drive a real browser so the sensor sets _abck, and keep the cookies (proxies for Playwright). Read the active coupon and weekly ad, not just the price, and stamp the week. Pin the store, match the exit to its region, and record it. Pace like a shopper.
Sizing
Size by request rate: find how fast one residential IP reads behind a browser before Akamai challenges it, and add exits. Budget browser-level bandwidth (headless browser proxy bandwidth cost). Accounts are one static ISP address each, and our pricing is pay-as-you-go with a balance that does not expire.
Free versus paid for Michaels
The test settles it: a datacenter client with no browser got Access Denied on both IPs. Free proxies confirm the pipeline forwards traffic, which our free proxy list and proxy checker cover. Reading a price needs paid residential, from $0.44/GB pay-as-you-go with no KYC, behind a real browser. The line is in datacenter vs residential proxies.
Staying unblocked
- Do not retry Access Denied without a browser. Akamai will keep refusing; the fix is the sensor.
- Capture the coupon and weekly ad with the price. A shelf price without them is a wrong number.
- Keep the sensor cookies. Discarding
_abckrestarts the sensor. - Pin the store and match the exit. A blended-store price is misleading.
- Watch the challenge rate. A rising 403 share means slow down or widen the pool. The full list is in avoiding IP bans while scraping.
What a proxy does not do here
A proxy gives a browser a region-correct residential address and keeps accounts apart. It does not run Akamai's sensor, does not set the store, and does not compute the effective coupon price for you. Scraping michaels.com runs against its terms of use, a risk that stays with you regardless of the IPs. What a residential exit does is let a real browser look like a US shopper so Akamai admits it, after which reading the coupon correctly is the real work. Test with the free proxy list, then read price and promotion with residential from $0.44/GB.
Sources
- Akamai, Bot Manager: the vendor behind our Access Denied page.
- Cloudflare, JA3/JA4 TLS fingerprinting: fingerprinting before the first HTTP request.
- HProxy test on 24 August 2026: plain GET requests to michaels.com from a residential connection and a datacenter server, both returning an Akamai Access Denied page.