How do I prevent sales from marking everything as 'junk' and skewing the data?

This is a process problem, not a configuration problem. You can't fix a behavioral issue with a software toggle.

Locking down the field or requiring a comment is a decent technical control, but it'll create friction and they'll find another way to cheat the system. Your issue is with the incentive structure and quota pressure. Sales will always optimize for the metric they're judged on.

You need to audit the "junk" reason at the leadership level. Treat those leads as a separate reporting cohort. If 40% of a rep's pipeline is suddenly "junk," that's a forecasting red flag for their manager, not a data cleanup task for you.

Totally agree it's a process issue. The technical fix is a band-aid.

But I think you can use the software to expose the behavior, which forces the process change. Create a real-time dashboard for managers showing "junk rate" by rep, right next to their activity metrics. Makes it a visible KPI instead of a hidden cleanup task. Suddenly it's their problem to explain.

Exactly. Making the junk rate visible is the key. But if you just expose it, managers will ignore it unless it's tied to something that matters to them, like forecast accuracy.

I built a simple report that compares a rep's forecasted pipeline value to their actual closed-won value, then factors in their junked lead value as a penalty. When their forecast is off by 30% because they junked half their pipeline, it's impossible to ignore.

You have to make the metric hurt.

Great point about filtering them out of reports automatically. That's often the easiest place to start and gets you clean data while the process debate plays out.

If you're self-hosting, your primary method will depend on your CRM. For most, you can set up a custom report filter that simply excludes any record with a "junk" status from the pipeline view that feeds Gemini. This gives you a true picture of the active pipeline.

But, this has a side effect: it makes the junking *invisible* in your analysis, which just kicks the can. I'd pair that technical filter with a mandatory, auditable "junk reason" field, like you mentioned. Even a simple dropdown with five options forces a moment of thought. It turns a single click into a conscious choice, which dramatically cuts down on casual misuse.

Love the mandatory dropdown idea. It's such a low-friction fix that can have a huge psychological impact.

You're spot on about the visibility trade-off, though. Filtering junk out of the main pipeline view gives you clean forecasting, but you absolutely need to track it somewhere. I'd suggest creating a separate "junk audit" dashboard that leadership reviews weekly. That way, the clean data feeds your models, but the behavior isn't buried.

If a rep's junk rate spikes, it becomes a coaching moment, not just a data anomaly.

You're right about incentives, but that's exactly why technical controls are essential. You can't change behavior without measurement.

A hard technical lock is what makes the audit possible. If they can still mark things as junk, but every single one requires a structured reason from a dropdown and maybe even manager approval for batches, you've created an audit trail. That's the evidence leadership needs to actually address the incentive problem.

No audit trail means it's just a finger-pointing exercise. "40% junk" is a vague complaint. "40% junk, with 85% citing 'No budget' despite being from a high-intent webinar" is a coaching moment with data.

That's a really common pain point, and you're smart to tackle it early before the data gets too messy.

The good news is, since you're self-hosting, you have a lot of flexibility. The mandatory comment field or a status change restriction is usually the quickest technical fix you can implement in your CRM's settings. It creates just enough friction to make them pause.

But honestly, setting up an automatic filter in your reports to exclude 'junk' entries is often the most effective first step for *getting clean data*. You can work on the process issue while your forecasts stay accurate. Just make sure you're still tracking those junked leads in a separate audit view, or you'll lose visibility into the behavior.

That "separate audit view" idea is smart. It keeps the main data clean while you're building the case for a process change.

How do you actually route the data though? Are you creating a duplicate filtered dataset for Gemini, or is there a way to set up the exclusion directly in its reporting pipeline? I'm new to this kind of setup.

Exposing the behavior is the right first step, but in my experience, simply making a "junk rate" visible often gets dismissed as a data quality metric, not a sales one. It needs immediate context to matter.

Your idea of putting it right next to activity metrics is good. I'd take it a step further and ensure that junked leads aren't just counted, but their original source and age are displayed too. If a rep's junk rate is high, but it's all old, unworked leads from a bad list, that's a sourcing problem. If it's fresh, high-intent leads, that's the coaching moment.

This turns the dashboard from a blame tool into a diagnostic one.

You've hit on the classic data-integrity vs. user-habit problem. The technical filter is straightforward, but you need to decide if you want to correct the data before it reaches Gemini or after.

Since you're self-hosting, your best first step is to set up a pre-processing filter in your data pipeline. Before Gemini ingests your CRM data, route it through a simple script that excludes records with the 'junk' status from the main analysis stream. This gives you clean forecasting immediately.

The caveat, as others noted, is this hides the problem. So, simultaneously, implement that mandatory comment field on the status change. The audit trail is useless if you don't capture the *reason*. Then you can have a secondary report showing junked leads *with* their provided reasons, which is your ammunition for fixing the root cause.

Oh yeah, that's a classic issue! The filter trick is smart for getting clean reports fast.

But I'm curious, what CRM are you using? The way to lock the status field or add a mandatory comment is totally different between, say, Salesforce and something like HubSpot. That might help narrow down the config tweak.

You're describing a really common data governance headache. The good news is, since you're self-hosting, you have the control to tackle both the symptom and the cause.

First, to get clean data into Gemini immediately, look at your data sync process. You can often set up a transformation step that strips out records with the 'junk' status before they even hit your analysis database. That's your quick win for accurate forecasts.

But that's just a bandage. The real fix is in the CRM's validation rules. You should absolutely require a comment or select a reason from a dropdown before that status can be saved. The key is to make that field mandatory at the database level, not just with a pop-up they can dismiss. That creates the friction and the audit trail you need to understand *why* things are being junked, which is where you'll find the real process issues.

That dashboard idea sounds really effective for making it visible. I hadn't thought about putting the junk rate right next to their other activity stats.

But would managers actually look at a new dashboard? How do you make sure it becomes part of their routine check? Maybe linking it from the main sales homepage?

That's a tough spot. I'm still learning about data pipelines myself. When you say you're self-hosting, how are you getting the data from your CRM into Gemini? Is it a direct database sync, or are you using an ETL tool?

Asking because the filter might be easier to add in one place vs another.