You upload a bulk sheet. The parse finishes. The page that opens is called Actions — but most users think of it as the Recommendations page, because that's what it shows: BulkSheet Pro's analysis of your account, laid out as a set of things you could do this cycle.
This is the page you'll spend the most time on. Once a month for established campaigns, twice a month for new ones, it's where the cycle begins. Knowing what's on it — and what each piece is asking of you — is the difference between five minutes of confident review and twenty minutes of staring.
The top of the page: the headline numbers and the flow viz
At the very top, you see four KPIs from your bulk sheet: total spend, total sales, ACoS, and total orders. These are computed from the current bulk sheet you just uploaded. They're the picture of your account right now, before any of the recommendations get applied.
Below the KPIs sits the flow visualization — a Sankey diagram that's worth understanding even if you only glance at it. It shows how every keyword and placement in your account routed through the rec engine: how many entities BulkSheet Pro evaluated, how many were excluded by campaign rules, how many sit on-target, how many had too little data to act on, and how many produced an actionable bid or placement recommendation. From there, the actionable ones flow into routing buckets (bid only, placement only, or both fire) and finally into projection buckets (whether the rec changed after queued negates were applied). Reading the Sankey is the topic of its own article — for this tour, what matters is that the visualization sits at the top of the page as a single picture of where every entity in your account ended up.
The action tiles
Below the flow viz is a grid of tiles. Each tile represents one category of recommendation. The categories typically include:
- Bid adjustments — bid raises and cuts, gated by the diagnostic gate
- Placement adjustments — Top of Search, Product Pages, Rest of Search, and any audience placements that have crossed the validity gates
- Negate terms — per-ad-group search term negations
- Cross-campaign negate — terms appearing in multiple ad groups that should be negated at scale
- Term conflicts — the same term running in multiple ad groups where one is winning and the others are losing
- Harvest terms — proven converters that haven't been promoted to their own exact-match campaigns
- Product pauses — products that have spent past patience without converting
- Search term patterns — n-gram-level waste or harvest opportunities
Each tile shows the high-confidence count for that category. Clicking the tile takes you to the corresponding page, with the Rec Actions filter pre-selected so you see only the high-confidence picks.
The tiles aren't required to be processed in any particular order. A typical cycle is: negate first (cleanest waste removal), then bids and placements (where the diagnostic gate handles routing), then harvest (extending winners). The reason for that order is one of cause and effect — negates clean up the data, which the projection layer then uses to compute more accurate bid and placement recs. We'll come back to that interaction in the projection article.
The three modes
Just above the tiles, there's a mode selector with three options: Balanced, Grow, and Reduce.
Balanced is the default and shows everything. Bid raises and bid cuts, placement increases and decreases, negates, harvest — the full set of recommendations the engine produced.
Grow mode filters to upward moves only. Bid raises, placement increases, and harvest recommendations stay visible. Bid cuts, placement decreases, and negates disappear from the page. Use this when you're in a phase where you want to expand reach, not contract it — for example, when a new product launch needs more impression volume and you don't want the rec engine pulling things back.
Reduce mode is the inverse. Bid cuts, placement decreases, negates, and product pauses stay visible. Bid raises, placement increases, and harvest are hidden. Use this when you're focused on cost control — a month where you need to bring ACoS down quickly, or when a campaign budget overrun has forced a tighter posture.
Most cycles run in Balanced. The other two are tools for specific weeks, not defaults. Switching modes recomputes the counts on each tile so you see what's actionable under the constrained view.
The Rec Actions filter
Every page reachable from a tile (Keywords, Search Terms, Bulk Negate, Harvest Terms, etc.) has a Rec Actions filter tab. When you click a tile, that filter is pre-selected, which narrows the page to only the high-confidence recommendations from the rec engine.
You can switch off the filter at any time and see the full data, but the design encourages you to start with the engine's picks and expand from there. Most of the time, the high-confidence Rec Actions list is exactly what you'd have picked manually if you'd reviewed every row. The filter just saves you the time.
For pages where the engine's picks are conservative — Harvest Terms is the most common example — you may want to widen the filter or scan the broader list to find candidates the engine didn't flag at high confidence. That's a judgment call. The Rec Actions tab is the starting point, not the boundary.
The stale banner
Sometimes the page shows a banner at the very top saying "Settings changed — re-analyze." This appears whenever you change any Advanced Option (target ACoS, sensitivity, any rule). The reason: changing a setting invalidates all the recommendations that were computed under the old settings. The tile counts and the Sankey are still showing the old numbers; the only honest way forward is to regenerate them.
Clicking "Re-analyze" reruns the engine against your current bulk sheet with the new settings. The tile counts update. The Sankey updates. The stale banner disappears. Now you can proceed with confidence that what you're seeing reflects your current configuration.
This is why it's worth setting your rules before doing detailed review work. If you re-analyze halfway through a cycle, you may find that some of the recommendations you were planning to apply have changed. Better to settle the rules first, run the analysis, and then commit to processing the tiles in order.
What's NOT on the Recommendations page
The Recommendations page is the front door, but it doesn't contain every page in BulkSheet Pro. Things that live elsewhere:
- The Campaigns page — the comprehensive view of every campaign with its metrics, bids, and placements. Used for ad-hoc investigation when you want to look at a specific campaign in detail.
- The ASIN Coverage and Target Coverage pages — the matrices that show coverage gaps. These run on the same bulk sheet but aren't part of the recommendation workflow.
- The builders — Auto, Keyword, Product Target, and Harvest. Reachable from the sidebar or from a coverage-page launch link.
- The Renames and Toolmaker pages — bulk operations that don't fit the rec engine's structure.
The Recommendations page handles the cycle work. The other pages handle the investigation, the building, and the bulk edits that fall outside the rec engine's scope.
What BulkSheet Pro won't do
It won't auto-apply recommendations. Every tile is a starting point for review. You click in, you look at the picks, you select what you want to apply, and you stage them as pending changes. Then you export. Nothing gets applied automatically.
It won't keep recommendations consistent across mode switches. The tile counts under Grow are different from under Balanced because the engine is filtering its output. If you switch modes mid-cycle, expect the page to refresh and the numbers to change.
It won't override your judgment on borderline cases. Medium-confidence recommendations exist and are visible if you widen the filter on the underlying page, but they're not in the default Rec Actions view. The engine is conservative about what counts as "high confidence." You decide whether to act on lower-confidence picks.
The first cycle, briefly
On your first BulkSheet Pro cycle, the page can feel busy. Eight or nine tiles, each with counts, plus a Sankey that's animating with flows you haven't yet learned to read. The instinct is to want to act on all of it.
The better approach: start with the negate tiles. Negates have the smallest blast radius (you're blocking specific waste) and the most immediate impact (less spend on the wrong things). Then look at bids — the diagnostic gate has already routed campaigns to the right lever. Then placements, then harvest. By the third cycle, you'll know which tiles tend to have actionable picks for your account and which tend to be empty most months.
The Recommendations page rewards consistency over completeness. You're not trying to clear every tile every cycle; you're trying to ship the high-confidence work and let the rest accumulate signal for next time.