Writing good prompts
Ask precise market-intelligence questions that map cleanly to WhatWins tools and use fewer credits.
You do not need to name a tool. Describe the evidence you want, the scope and how results should be ranked. The AI client can map that request to the correct WhatWins tool.
A reliable prompt structure
Include four elements when they matter:
- Subject — products, shops, ads or organic posts.
- Scope — tracked shops, niche, platform or country.
- Window — last 30 days, 6 months or another bounded period.
- Ranking — reach, longevity, outlier score, traffic or relevance.
Find 6 active Meta video ads for the tracked skincare shops,
limited to the US, launched in the last 90 days, sorted by longest running.Better prompts by task
Discover competitors
Discover 5 Shopify stores selling red-light therapy masks in the US.
Require at least 100,000 monthly visits and rank by traffic.Analyze ad creative
Find the 8 longest-running active Instagram video ads for these tracked shops.
Summarize the recurring hook, offer and creative format using only returned evidence.Find organic breakouts
Show the top 6 TikTok posts about electrolyte powders from the last 30 days,
ranked by outlier score rather than raw views.Compare shops
Compare 12 months of traffic and Meta live-ad history for Gymshark and Alo Yoga.
Call out direction changes, but do not infer audited revenue or spend.Keep requests bounded
Smaller limits make answers faster and consume fewer data credits. Ask for 5–10 high-signal results first, then refine. Avoid repeatedly requesting the same broad dataset in one conversation.
Separate evidence from interpretation
When you want strategic analysis, say so explicitly while keeping the evidence boundary clear:
First report the verified WhatWins evidence. Then give three hypotheses,
clearly labeled as interpretation, for why these creatives may be durable.Follow-up prompts
Good follow-ups reuse the current scope and change one dimension:
Now show only video creatives.Use the same shops, but rank organic posts by saves.If the client loses context, restate the shops and filters rather than relying on an ambiguous “same as before.”