The CDN market has thirty-plus credible vendors and your evaluation has a deadline. This guide is a five-day elimination method: map the tiers, apply your disqualifiers, verify coverage against your actual geography, sanity-check pricing, and finish with three candidates worth benchmarking properly.
Day 1: map the field by tier
Do not evaluate thirty vendors; classify them. The market sorts into five tiers — hyperscaler clouds, platform independents, the value tier, regional and sovereign specialists, and adjacent platforms (media, image, edge-first) — and your workloads usually eliminate whole tiers before any vendor conversation. An estate living in one cloud starts with that cloud's CDN in the list by default; a price-driven static workload starts in the value tier; Gulf or China audiences pull the regional layer in immediately. Spend day one writing one sentence per tier: relevant or not, and why. The comparison section's market map is built to make this pass fast.
Day 2: eliminate on disqualifiers
Take the disqualifier list from your requirements page and apply it ruthlessly to every vendor left standing after the tier pass. This is desk research, not calls: compliance pages, documentation, ToS, status history. Typical eliminations at this stage: no POPs (not "presence" — POPs) in a country that carries 15% of your audience; a token scheme incompatible with your players; contract minimums above your appetite; a sub-processor you cannot accept; no self-serve trial when your process requires hands-on testing before procurement (our trials comparison maps who offers what). Document every elimination in one line — the discipline matters at renewal time, when "why didn't we consider X" returns.
Day 3: verify coverage against your map
Vendors publish coverage maps; day three is checking them against your geography table, city by city for your top countries. Three checks per candidate: does a POP exist in or near each city that matters (published lists, looking-glass tools, community measurement data); is it always-on or overflow (some networks serve secondary cities from hubs at off-peak); and — the question maps never answer — what eyeball networks does it peer with there. You cannot fully resolve the third from a desk, which is precisely why the week ends in a benchmark plan rather than a decision. Candidates that fail the desk check on a top-three country leave the list now.
Day 4: sanity-price the survivors
You are not negotiating yet — you are checking order-of-magnitude fit. Model your traffic profile against published pricing where it exists (the value tier makes this trivial; the majors make it approximate) and against the reference bands in what a CDN should cost where it does not. Two eliminations happen here: vendors whose floor is above your ceiling, and — just as real — vendors so far below the band for your workload that the price implies a coverage or support model that will not survive your requirements. Note the request-pricing and feature line items now; they ambush later otherwise.
Day 5: pick three and write down why
Three is the right number: enough for genuine comparison and negotiating tension, few enough to benchmark properly. Aim for productive diversity — say, an incumbent-tier platform, the strongest value-tier fit, and whichever candidate your workloads made non-obvious — rather than three near-identical options. Write a half page: the three names, the one-line reason each survived, the eliminated names with their one-line reasons, and the open questions the benchmark must answer. Then hand it to the test plan in the benchmark guide. A shortlist produced this way takes a week; defending it takes one paragraph, forever.