Working With AI Agent Connected to Voluum

AI assistants are good at producing confident, well-formatted answers, including when the underlying data does not actually support them. The single most effective safeguard is to prepend the three rules below to any analysis prompt. They cost nothing and prevent the most common failure modes. 

1. Every number you report must trace back to a query you actually ran. 

   If you can't point to the specific query that produced a row, what you 

   grouped by, the filters, the date range, the conversion-time mode, don't 

   report that row. Never reverse-engineer rows from totals to fill a table. 

  

2. Don't name a variable, event, or column unless you've verified it exists 

   on this account or campaign. Different traffic sources expose different 

   custom variables (some use a zone id, some only a publisher id, some have 

   no creative dimension at all). Different campaigns track different 

   conversion events. Look up the actual schema before naming things. 

  

3. If the data doesn't support the analysis I asked for, say so. "I can't 

   break this down because the variable isn't there" or "this campaign only 

   tracks Visits to Purchase, no funnel" is the right answer. Don't fabricate 

   rows to fill a table; don't invent insights to fill a section. 

Surface yesterday's problems against your recent baseline, in one pass.

Pull yesterday's data from my Voluum account. Compare it to the prior 7-day 

average. Tell me: 

- which campaigns dropped more than 30% in profit or conversions 

- which campaigns spent more than $X with zero conversions 

- any cost discrepancies (cost shown but suspiciously low/zero visits, or 

  visits without cost) 

- any postback issues (CR dropped to ~0 on previously converting campaigns) 

Filter to workspace(s) [W1, W2, ...] if I have multiple, don't blend across 

workspaces. Don't include campaigns with fewer than 100 visits, not enough 

signal. 

Movement plus a one-line reason, separated from new-launch noise. 

Compare this week (last 7 complete days) to the prior 7 days for my Voluum 

account, scoped to workspace [X] if I have multiple. For each top-50 campaign 

by spend, give me: 

- profit / cost / conversions / CPA week-over-week 

- only flag changes greater than +/-20% 

- one short sentence explaining what likely changed (volume up, CR up, 

  payout change, new geo, etc.) 

- group the output: "Top movers up", "Top movers down", "New campaigns this 

  week" (separate so I can see real movement vs new-launch noise). 
 

Use the same conversion-time mode (VISIT or CONVERSION) for both weeks. 

Don't mix, late-funnel comparisons will show fake drops if you do. 

Catch missing, anomalous, or mis-modeled cost before it distorts decisions. 

Audit cost integrity in my Voluum account for the last 7 days: 

- list campaigns with visits but zero cost 

- list campaigns where the cost-per-visit looks anomalous (3x or 1/3x the 

  campaign's average over the last 30 days) 

- list campaigns where the deal type appears to be CPA but cost is being 

  tracked per-click (or vice versa) 

- group results by traffic source so I know where to focus 

The inverse of "top spenders": efficient performers that could take more budget.

Find sources, zones, or campaigns in my Voluum account that have: 

- high conversion rate or low CPA (top quartile) 

- but low cumulative spend in the last 14 days (bottom quartile) 

- and at least 30 conversions / enough volume to be meaningful 

Tell me which ones could plausibly absorb more budget. Skip anything with 

weak statistical signal. 

A clean, paste-ready snapshot for one workspace, with no cross-workspace blending.

For workspace [X] only, give me: 

- yesterday's KPI snapshot (visits, conversions, cost, revenue, profit, ROI, 

  CPA) vs prior 7-day daily average 

- top 5 campaigns by profit and bottom 5 by profit, in this workspace only 

- any campaigns where the trend over the last 3 days suggests a problem 

  (3-day declining profit, sudden cost spike, cost without conversions) 

Format as plain text I can paste into a team meeting agenda. 

Drill into one campaign by the source's own placement variable, then act on it.

For campaign [Campaign Name], pull the last 14 days grouped by [zone-id / 

sub-id / placement-id, the variable that traffic source uses, look it up from 

the source's custom variable mapping]. 

  

Before pulling the breakdown: 

- Verify which custom-variable slot this source actually populates with 

  meaningful data (it differs per source). 

- Pull a 5-row sample first to confirm the variable has data. If the variable 

  is empty or the source doesn't expose what I asked for, stop and tell me, 

  don't substitute another variable. 

- Filter by campaign at the API level so you're not blending across campaigns 

  that happen to share the same source. 

  

Then give me: 

- top 10 by spend, with conversions, CPA, ROI 

- bottom 10 by spend with no conversions 

- recommend pause for any zone with $X+ spent and zero conversions 

- recommend scale (increase share) for any zone with at least N conversions 

  and CPA below my target of [target value] 

Present as a table I can scan in 30 seconds. 

Find the hours worth running, filtered to windows that actually have signal.

For campaign [Campaign Name], pull last 30 days of performance grouped by 

hour of day. Filter out hours with fewer than 50 visits, not enough signal. 

Identify: 

- top 6 hours by conversion rate, with CPA and confidence interval 

- worst 6 hours by CPA, with confidence interval 

- recommended run window: a contiguous block of hours where CR is at least 

  X% above the campaign average 

Use VISIT-time conversion mode so batched conversions don't distort the 

hourly view. 

Measure each partner against the CPA target you actually care about, not a generic one.

Group last 14 days by affiliate network / partner. For each partner, my 

target CPA is [partner: target] (e.g., Partner A: $40, Partner B: $35, 

Partner C: $50). Tell me: 

- which partners are above target (i.e., losing money against KPI) and by 

  how much 

- which partners are below target (winners) and by how much budget headroom 

  exists before we hit cap 

- ignore partners with fewer than 100 conversions in the period 

Catch traffic that converts on the first event but decays downstream.

For my Voluum account, find campaigns where: 

- the first conversion (Install / Lead / Subscribe) is firing normally 

- but day-1 or day-3 retention, or first-deposit / first-purchase, is 

  significantly below the campaign's 30-day average 

- or below the account's 30-day median for the same offer 

Flag with the dollar gap between expected and actual downstream revenue. 

Sort by absolute dollar impact, not percentage. Skip campaigns with fewer 

than 50 first conversions. 

Find the step where a multi-step funnel is bleeding, across the whole account.

For each top-30 campaign by spend: 

  

1. First, look up which conversion events that specific campaign actually 

   tracks. Don't assume Visits -> LP -> ATC -> Initiate Checkout -> Purchase 

   applies, many campaigns only track Visits -> Purchase. List events per 

   campaign before computing rates. 

  

2. Skip campaigns that only track a single step (Visits -> Purchase), no 

   funnel to leak through. 

  

3. For campaigns with a multi-step funnel, compute step-by-step conversion 

   rates and flag where: 

   - one step has a drop-off significantly worse than the campaign's other 

     steps (i.e., a leak) 

   - one step has a rate that's anomalously high, over 50% on a step that 

     should be 10-20% (likely a misfiring pixel firing on page load) 

  

4. For each flag, suggest the most likely cause (pixel issue / payment-form 

   breakage / tracking lag) but mark it as a hypothesis, not a diagnosis. 

   Multiple hypotheses are better than one if the evidence is ambiguous.

Reallocate budget between geos on volume-qualified evidence, not bid tweaks.

For [campaign or workspace], pull last 30 days grouped by country / region. 

Use a single conversion-time mode for the whole 30-day window, don't mix 

VISIT and CONVERSION mode within the analysis or comparisons will be off. 

  

Filter to geos with at least 200 visits and 5 conversions. Tell me: 

- top 5 geos to scale (low CPA, high CR, room to grow) 

- bottom 5 geos to cut (high CPA, low CR, eating budget) 

- suggest specific dollar reallocations from cut geos to scale geos. 

  

If fewer than 5 geos pass the volume threshold, tell me how many qualified 

and stop, don't pad the list with sub-threshold geos. 

Stop sending traffic to an offer once its daily cap is reached, then resume next day.

Watch offer [Offer Name] in my Voluum account. When daily conversions reach 

[cap value] in [cap timezone, e.g., Europe/Berlin], pause every campaign 

sending traffic to it. At midnight in [cap timezone], resume them all 

automatically. Notify me on [Slack channel / chat] each time you pause or 

resume something, with a summary line: "Paused 5 campaigns at 14:23, cap hit 

at 250/250 conversions." 

Manage one shared budget that spans several offers under the same partner.

Watch the group of offers I share with partner [Partner Name]: [list of offer 

names]. Their combined daily budget cap is [$X]. Alert me on [Slack / chat] 

when: 

- the group is at 75% of cap (so I can decide whether to redistribute) 

- the group hits cap (auto-pause all campaigns sending to these offers) 

At midnight in [cap timezone], unpause everything.

Confirm new campaigns actually start receiving traffic and flag the ones that don't.

I just sent campaigns [list of campaign names] live and they are supposed to 

start running. Check those campaigns daily, and tell me as soon as each one 

starts receiving more than 10 visits. Stop watching a campaign once it crosses 

the threshold. If a campaign hasn't received any traffic after 72 hours, alert 

me, likely a setup problem on the publisher side. 

Go from brief to a live, correctly-linked campaign on both Voluum and the source, in one flow.

I want to launch [a new campaign / N campaign variations] for offer 

[Offer Name] on traffic source [Source Name]. 

  

For each campaign: 

1. Create or pick the affiliate-network offer entry in Voluum. 

2. Generate the Voluum tracking URL with the right macros for [Source Name] 

   (look up the source's macro convention from its docs). 

3. Create the Voluum campaign linking source -> offer, with cost model 

   [CPC/CPA]. 

4. Create the matching campaign on [Source Name] using their API, with the 

   tracking URL from step 2 as the destination. 

5. Set initial bids/budgets to [defaults]. 

  

For bulk launches, the variations are: [list dimensions, e.g., 

country = [US, CA, UK], device = [mobile, desktop]]. Generate names following 

the convention "[source] - [country] - [offer] - [device]". 

  

Confirm with me before executing each pair (Voluum + source). Show me the 

campaign URLs when done so I can verify both sides.

Take a profitable campaign's config and stand it up on additional sources, adjusted per source.

Campaign [Campaign Name] on [Source A] is profitable (CPA $X, ROI Y%). Take 

its config, offer, geo, device, tracking URL pattern, and create the 

equivalent campaign on [Source B, C, D]. 

Adjust for each source's: 

- macro conventions in the tracking URL 

- minimum bid / typical bid floor for this geo 

- creative format requirements (if I provide creative, upload it; otherwise 

  use the same creative as Source A) 

Set starting bids at [policy, e.g., 80% of Source A's CPA target] so we can 

test before scaling. Tell me which campaigns went live successfully and which 

had a setup problem I need to resolve.

When an offer caps, stop accepting traffic in Voluum and stop paying for it at the source.

Watch offer [Offer Name]. When daily conversions reach [cap value] in 

[cap timezone], simultaneously: 

- pause every campaign in Voluum sending traffic to this offer (so the flow 

  stops accepting visits and reporting reflects the pause) 

- pause the corresponding campaigns at the traffic source (so we stop paying 

  for impressions/clicks) 

At midnight in [cap timezone], reverse both. Notify me on [Slack / chat] with 

a summary of what was paused/resumed and any failures. 

Make Voluum's cost match the source-of-truth, with a diff you approve first.

For campaigns on [Source Name] over the last 30 days: 

1. Pull authoritative cost from the source's reporting API (per campaign / 

   per day). 

2. Compare to what Voluum currently shows. 

3. Identify mismatches: missing cost on visited campaigns, wrong cost, 

   partial cost. 

4. Show me a diff before applying. 

5. On confirmation, write the corrected cost back to Voluum so reporting 

   matches the source-of-truth. 

  

Run on demand or on a weekly schedule. 

Verify the full round-trip before you scale spend, instead of finding the leak after.

I just set up campaign [Campaign Name] on [Source Name] in Voluum. Validate 

end-to-end before I scale spend: 

1. Voluum side: tracking URL has the right click-ID and sub-ID macros for this 

   source; postback URL is configured; the offer's payout pixel fires on the 

   destination correctly. 

2. Source side: campaign destination URL matches the Voluum tracking URL 

   exactly; macro tokens match the source's expected format. 

3. Run a small test ($X spend cap) and tell me when the first conversion 

   round-trips: visit recorded in Voluum -> click -> LP -> conversion event -> 

   postback fired to source. Each of the four steps either confirmed or 

   flagged with the specific failure. 

Hold the rest of the budget until I approve. 

Rebalance one offer's budget across the sources running it, without logging into each.

For offer [Offer Name], compare last 7 days performance across all sources 

running it: [list]. For each source, calculate CPA, conversion rate, and 

trend. 

Recommend a budget reallocation: shift X% of budget from underperformers to 

top performers, but keep a minimum floor on each source so we don't lose 

diversification. On confirmation, push the new budget caps to each source's 

API. Show me before/after for each source. 

Get a runway warning before a source runs dry. (Reads balances from each source's API, not from Voluum.)

Once a day, check my account balance on each traffic source where Voluum has 

an API integration: [list]. Compute my 7-day average daily spend on each 

source. If the remaining balance covers fewer than 3 days at the current pace, 

send me a [Slack / chat] alert with the source name, balance, and projected 

days-of-runway, so I can wire funds before the weekend. 

When you run a second tracker after Voluum, find where the two disagree.

For campaigns matching [pattern] over [date range], reconcile Voluum against 

[downstream tracker / partner ledger]: 

1. Pull conversions from Voluum (clickid, payout, conversion type, timestamp). 

2. Pull conversions from [tracker] for the same window, using the shared 

   identifier ([clickid / sub-id / external_id]). 

3. Diff the two sets. Surface: 

   - clickids in Voluum but missing from [tracker] (likely lost postback) 

   - clickids in [tracker] but missing from Voluum (likely lost upstream pixel 

     or different attribution window) 

   - clickids in both but with mismatched payout (currency? partial-credit 

     event? duplicate conversion?) 

   - timestamp drift > [N hours] suggesting attribution-window misalignment 

4. Output a CSV I can paste into my warehouse alongside my own data. 

Run on demand or on a daily/weekly schedule. 

Watch a fresh campaign's first visits and decide whether tracking is sound enough to scale.

For campaign [Campaign Name], I just launched it. Watch the first 500 visits 

and tell me: 

- whether each tracked event (LP, ATC, Initiate Checkout, Purchase) is firing 

  within plausible ranges given the offer type 

- whether any event is firing at an impossibly high rate (likely page-load 

  pixel rather than action pixel) 

- whether the conversion-to-payout cycle looks normal vs broken postback 

Tell me to either go ahead with scale or stop and fix tracking before 

spending more. 

Build the table you'll compare line-by-line against a partner's own report.

For partner [Network Name], pull last month's data: total conversions 

attributed in Voluum, by campaign and country. Compute the sum of "qualified 

FTDs" / "valid leads" / [whatever quality column applies]. Format this as a 

table I can compare line-by-line with the partner's own report so I can see 

where the two systems disagree. 

Use the assistant as an on-demand product guide when you're not sure how to do something in Voluum.

I want to [analyze X by Y in Voluum / set up a rule for Z / track event W]. 

What's the right way to do this in Voluum? Walk me through the exact UI or API 

steps. If there's a feature I'm missing, point me to the docs.