Use case: Client wants to sample 1000 cases/week for one team. There are three conditions to consider; NRR macro application, Closed macro application, and Send to Team macro application. They want the distribution of each condition to be based on the distribution of macro usage overall each week. Is this type of dynamic sampling possible? Question to Rovo and response saying it is not attached, along with the static sampling rule I've created.

Question: in sprinklr, i am building a sampling rule for manual quality audits. i want to sample cases from three buckets, and have the count of cases from each bucket be based on the distribution of macros applied on all cases 

Rovo Answer: 

You’re basically trying to do proportional stratified sampling based on macro usage, then split that into 3 audit buckets. 

Let me restate the goal in Sprinklr terms and then walk through how to set it up: 

• You have a population of cases (e.g., all handled cases for a day/week). 

• Each case has a Macro Applied value. 

• You want to: 

• Look at the overall distribution of macros across all cases (e.g., Macro A = 50%, Macro B = 30%, Macro C = 20%). 

• Sample an audit set whose macro mix mirrors that distribution. 

• Then put those sampled cases into 3 buckets, where each bucket still keeps that same proportional macro distribution. 

Because the native “Sampling Rules” UI doesn’t (yet) do full-blown proportional allocation by a dynamic metric directly, you typically get there with a combination of: 

1. A baseline audit volume (total cases to be sampled). 

1. One or more Views / Dashboards / Reports to compute proportions. 

1. Either: 

1. Multiple parallel sampling rules by macro group, or 

1. A single rule that uses a classification field you pre‑compute. 

Below is the cleanest, maintainable approach I’d recommend. 

 

Step 1 – Decide your total sample and bucket design 

First, pick: 

• Total sample size for the period (e.g., N = 300 cases / week). 

• Number of buckets: 3 (e.g., Bucket A, B, C). 

• How you want cases split across buckets: 

• Option 1 – Equal by count: each bucket gets N/3 cases. 

• Option 2 – Equal by weight: each bucket gets ~⅓ of each macro’s sampled cases (better for proportionality). 

I’ll assume: 

• 300 total audits/week. 

• 3 buckets (100 cases each). 

• Within each bucket, keep the macro distribution equal (each bucket gets ⅓ of each macro’s sample). 

 

Step 2 – Get your macro distribution from reporting 

Use Reporting/Insights first to get stable proportions. 

1. Build a Case-level widget / view filtered to your scope: 

1. Channel(s) you care about (FB, X, IG, LinkedIn, etc.). 

1. Time window matching your audit period (e.g., last 7 days). 

1. Only “Completed” or “Closed” cases that are eligible for audit. 

1. Bucket by Macro Applied: 

1. Group by field: Macro Applied (or your org’s equivalent field, like Last Macro Applied). 

1. Measure: Count of Cases. 

1. Export the macro distribution: 

1. For each macro (or macro group), calculate: 

1. Macro Share = Count for Macro / Total Count. 

Example result: 

Macro	Cases	Share
Macro_A – Billing	5,000	50%
Macro_B – Tech	3,000	30%
Macro_C – General	2,000	20%
Total	10,000	100%

 

Step 3 – Convert distribution into per‑macro sample targets 

Now apply those shares to your 300-case sample: 

• Macro_A: 50% of 300 → 150 cases 

• Macro_B: 30% of 300 → 90 cases 

• Macro_C: 20% of 300 → 60 cases 

If you want equal macro distribution per bucket: 

• Each bucket size = 100 

• Per bucket: 

• Macro_A: 50% of 100 → 50 cases 

• Macro_B: 30% of 100 → 30 cases 

• Macro_C: 20% of 100 → 20 cases 

You can keep these numbers in a quick reference sheet; you’ll use them to configure rules and/or quotas. 

 

Step 4 – Translate this into Sprinklr sampling rules 

There are two main patterns that usually work well: 

Pattern A – Multiple sampling rules by macro, then round‑robin buckets 

Use one sampling rule per macro (or macro group), each with its own target count, then assign buckets in a systematic way. 

1. Create 1 Rule per Macro Group 

For each macro group: 

1. Rule name: Audit Sample – Macro_A – Weekly 

1. Filter: 

1. Time window: e.g., Cases created/closed in last 7 days 

1. Macro Applied = Macro_A 

1. Any other constraints (language, queue, region, etc.) 

1. Sample size: 

1. Target count: 150 (per week), or 

1. Percentage such that, given volumes, you land close to 150. 

Repeat for Macro_B and Macro_C using 90 and 60 respectively. 

1. Use a field to assign buckets (1, 2, 3) 

Because Sprinklr doesn’t random-split into buckets natively, you typically: 

1. Add a custom field on Case: Audit Bucket (values: 1, 2, 3). 

1. Use a rule / rule engine / workflow to assign buckets in a round‑robin or hash fashion among sampled cases. 

Common pattern: 

1. When case is flagged for Manual Audit (by your sampling rule), run a Rule: 

1. If Audit Bucket is empty: 

1. Set Audit Bucket = (Total audits selected mod 3) + 1 or equivalent logic using internal counters or routing logic. 

1. Or simpler, if you already have a randomization or hashing field (like Case ID last digit), you can: 

1. Last digit 0–3 → Bucket 1 

1. 4–6 → Bucket 2 

1. 7–9 → Bucket 3 

This gives you three buckets where the macro mix is preserved because: 

1. You’ve already fixed macro proportions at the sampling stage. 

1. Bucket assignment is random or evenly distributed among the sampled set. 

1. Create Views/Queues per Bucket 

1. Bucket 1 view: Manual QA – Bucket 1 → filter Audit Bucket = 1. 

1. Bucket 2 view: Audit Bucket = 2. 

1. Bucket 3 view: Audit Bucket = 3. 

Your auditors (who are already auto-assigned from a work queue) can each be mapped to one or more buckets as needed. 

 

Pattern B – Pre-classify cases and drive sampling via a “Macro Group” field 

If you have many macros and don’t want a separate sampling rule per macro: 

1. Create a “Macro Group” or “Macro Category” field on cases. 

1. Use rule logic to: 

1. When a macro is applied: 

1. Set Macro Group = Billing / Tech / General, etc. 

1. Re-run your distribution analysis on Macro Group instead of individual macros. 

1. Create sampling rules per Macro Group rather than per macro. 

1. E.g., 50% Billing group, 30% Tech group, 20% General group. 

This lets you keep the rule set compact and still approximate macro-based proportionality. 

 

Step 5 – Make it dynamic over time (optional but recommended) 

If your macro usage shifts, you don’t want to manually recalc percentages every week. 

Two lighter-weight options: 

1. Quarterly or monthly re-baseline 

1. Re-run the macro distribution report once a month/quarter. 

1. Adjust the target counts in your sampling rules accordingly. 

1. Use broad macro groups with stable ratios 

1. Group macros into 3–5 big categories whose share doesn’t swing daily. 

1. Sample by those groups, not by each individual macro. 

True fully-dynamic auto-recalculated sampling (e.g., “sample at a rate that keeps macro sample share = macro population share in real‑time”) usually needs a custom backend job or a more complex workflow. If you’re interested, we could sketch how that might look as a rule+classification+dashboard loop. 

 

Step 6 – Validate that your buckets are actually proportional 

After you deploy: 

1. Build a small QA dashboard: 

1. Filter: Flagged for Manual QA = Yes. 

1. Breakdown 1: Macro Applied (or Macro Group). 

1. Breakdown 2: Audit Bucket. 

1. Compare: 

1. Macro distribution across all audited cases. 

1. Macro distribution within each bucket. 

1. They should be roughly aligned with your expected 50/30/20 (or your actual distribution). 

1. Adjust: 

1. If one macro is consistently over/under-sampled: 

1. Tweak its rule’s percentage/target count. 

1. If one bucket is under-filled: 

1. Check your bucket-assignment logic (modulo logic or hash mapping). 

 

Summary 

To get “3 audit buckets whose case counts are aligned to the overall macro distribution,” you’ll typically: 

1. Use Reporting to compute macro (or macro-group) shares. 

1. Convert those shares into per-macro sample targets. 

1. Implement: 

1. Either one sampling rule per macro/macro-group, or 

1. A sampling rule driven by a Macro Group case field. 

1. Assign sampled cases into 3 buckets using a simple, even distribution method (round-robin or hash on case ID). 

1. Verify with a dashboard that your bucket-level macro distributions match expectations.