The first time you run a bulk sheet through BulkSheet Pro on a new account, the Recommendations page can feel either oddly empty or oddly full — depending on the state your account was in before BulkSheet Pro saw it.
If the account was running but nobody had been touching it, the page is full. Months of waste have piled up. Negate counts are in the hundreds. Bid and placement recs are everywhere. The Sankey has a thick Actionable band and the tile counts are eye-popping.
If the account was already being managed manually with reasonable diligence, the page is quieter. Most of the obvious waste was already addressed. The negate count is modest, the bid recs are small, the Sankey shows a heavy On Target band.
Either way, the first three cycles follow a recognizable pattern. Knowing what to expect makes the work feel less random.
Cycle one: mostly negate work
The first cycle is dominated by negate recommendations. The reason is simple: negates have the lowest data threshold of any rec type. A search term that has spent past the patience budget without converting is actionable after as little as a few clicks of evidence. Bid and placement recommendations need more data — at least 10 clicks per keyword and 20 clicks plus 2 orders per placement — and many entities in a new-to-BulkSheet-Pro account haven't accumulated enough to qualify yet.
So cycle one tends to look like: a meaningful list of negates (especially in the Negate Terms tile and the cross-campaign negate tile), a smaller list of bid recommendations, even fewer placement recommendations, and possibly some harvest candidates if your auto campaigns have been running long enough to surface converting search terms.
The right move: clear the negate queue first. Apply the high-confidence negate picks across all three negate-related pages (Negate Terms, Cross-Campaign Negate, Bulk Negate Phrase if there are obvious patterns). Then look at bids and placements — but expect the counts to be modest because the validity gates are filtering out a lot of the borderline data.
The first cycle is also when the projection layer pays off most visibly. After you stage 30-50 negates, watch the bid and placement tile counts drop. The Sankey's projection column shifts from "all unchanged" to "many dissolved and softened." Many of the bid recs that looked actionable before the negates dissolve away — they were driven by the waste you're now blocking.
The patience floor matters in the first cycle too. If your account is genuinely new — fewer than 50 lifetime orders — the patience floor of $5 kicks in to prevent the math from negating things on near-zero spend. This is the right behavior for sparse data, but it means you may see fewer negate recommendations than you would on a more mature account. As your data accumulates over the next month, the floor stops binding and the engine's negate recommendations sharpen.
Cycle two: bid and placement work catches up
A month later you download the next bulk sheet, upload it with the cycle-one session JSON, and the picture has shifted.
The Change Impact panel populates. The 30-50 negates from cycle one have done their job — most of those search terms have stopped spending, and account-level ACoS is meaningfully better. The bid and placement changes you applied last cycle now have a month of data attached, and the panel shows how each category performed.
On the new analysis, the negate counts are smaller. The obvious waste was caught last cycle; what's left is the slower drift — newer search terms that accumulated borderline spend, plus a handful of new auto-discovered terms that didn't convert.
What's grown: bid and placement recommendations. Two months of data is enough for the validity gates to qualify many more keywords and placements. Where cycle one had maybe 30 bid recs, cycle two might have 80. Where cycle one had a dozen placement recs, cycle two might have 40.
This is the right phase for the diagnostic gate to start earning its keep. The gate is routing campaigns to either bid or placement based on the bigger gap from target. You'll see the Sankey routing column populate with both blue and amber bands. Most cycles, bid is the more common lever, but you'll see real placement work too.
The harvest tile also starts to fill in cycle two. Search terms that converted twice in cycle one's window — without quite reaching the harvest threshold yet — often hit it in cycle two as the auto campaigns keep generating signal. Harvest counts of 5-20 are common in cycle two.
Cycle three: steady state
By cycle three, the rhythm is established. The Recommendations page looks like a typical month's work — not a backlog being cleared.
You'll start to recognize your account's signature on the Sankey. Maybe your account routes mostly to bid (mispricing is the more common issue). Maybe you have a few campaigns that always show up in the placement column because they consistently have placement-allocation issues. Maybe the dual-fire case never triggers, or always triggers on the same two campaigns.
The Change Impact panel from cycle two shows whether last cycle's calls were right. Bid cuts that were too aggressive show up with ACoS dropping but order volume dropping more than expected. Bid raises that landed correctly show up with both ACoS holding and orders increasing. Placement adjustments that hit the right magnitude show the placement-level ACoS coming into the target range.
This is the cycle where you might start refining your rules. If cycle two's Change Impact showed that branded keyword bid cuts overshot, you might add a keyword rule for branded terms with gentler sensitivity. If certain SKUs are converging too slowly toward target, you might adjust their SKU rule's target ACoS or sensitivity.
The rules layer becomes useful precisely at this point — when you have enough cycles behind you to know which categories of decision are reliably going one way or another. Adding rules before you have that signal is guessing. Adding rules after is configuring BulkSheet Pro to match what you've already learned.
What feels different about new accounts
Three things distinguish the first few cycles from steady state:
The patience floor protects against over-correction. The $5 patience minimum prevents the negate engine from flagging terms on tiny account CPO values. On a new account where your CPO might be artificially low (because you've had a few lucky cheap orders), the floor keeps the engine from being too eager. As your data fills in and account CPO stabilizes, the floor stops mattering.
The validity gates leave more things on the table. The 10-click floor for keyword bids and the 20-click + 2-order floor for placements mean that, early on, many entities will sit in the Insufficient bucket on the Sankey. They're not being ignored — they're being held until they have enough data to evaluate. By cycle three, most of them have crossed the floor and entered the regular rec flow.
The harvest engine is patient. Harvest requires at least two orders on a search term, plus ACoS at half target or better. Most new accounts will not surface harvest candidates in cycle one because the data has not accumulated. Cycle two typically has a handful; cycle three often has many more.
All three of these are protective, not restrictive. They prevent the engine from making confident calls on data that has not earned them. The cost is a slightly emptier-feeling first cycle. The benefit is that the recommendations you do see in cycles one, two, and three are honest about the data behind them.
What to do if the page feels overwhelming in cycle one
If cycle one's tile counts are in the hundreds across the board, that is an account that was running unmanaged for a long time. The page is not broken — there really is that much accumulated waste.
The instinct is to want to clear all of it at once. The better approach: focus on the high-confidence picks from the Rec Actions filter on each page. Do not widen the filter. Do not try to address every medium-confidence rec. Apply the conservative subset, export, upload to Amazon, and then run the next cycle when the data window catches up.
By cycle two, the high-confidence picks from cycle one will have done meaningful work, and the page will look less overwhelming. Each cycle compounds. You do not have to clear the page in cycle one.
What to do if the page feels empty in cycle one
The other failure mode: tile counts in the single digits or zero, across most categories. This usually means one of two things.
Either your account is genuinely well-managed and there is not much for the engine to do, in which case the empty page is a sign of health (and you can shift attention to coverage gaps via the Coverage pages, or to harvest opportunities as they emerge).
Or your bulk sheet is too short. If you uploaded a 7-day sheet, the data window may be below the engine's gate thresholds for many entities. Try uploading a 14-30 day sheet — that is the ideal window. Below 14 days, the engine is conservative; above 45 days, seasonality can blur the picture.
The cycle that doesn't change
By cycle four or five, the workflow is the same one you will be running indefinitely:
- Download the new bulk sheet
- Upload it with the previous cycle's session JSON
- Read the Change Impact panel — confirm last cycle's calls were right (or note which were not, and refine your rules)
- Glance at the Sankey — confirm the account's signature shape looks normal
- Work through the tiles, negates first
- Stage the high-confidence picks; let the projection layer recompute
- Export the three files
- Upload to Amazon
The first three cycles are when this workflow becomes muscle memory. After that, BulkSheet Pro fades into the routine — the goal isn't for the tool to be interesting, it's for the cycle to be predictable. A predictable cycle is one you can trust to do its job, month after month, without taking up the cognitive load you would rather spend on the parts of the account that are not automatable.