Smarter recruitment reporting: dashboards over your own data

Most recruitment agencies steer on gut feeling. They know some recruiters perform better than others. They suspect certain types of conversations are more effective. They have a sense for which vacancies are easy or hard to fill. But hard data? That’s usually missing.

The problem isn’t that agencies don’t want to measure. It’s that the numbers live in three places at once. Some sit in the system, some in a spreadsheet one person maintains, and the rest in the heads of the people doing the work. Getting a straight answer to “where is the pipeline stuck” costs an evening and a CSV export.

That is what reporting in Simply is for. And it’s worth being precise about what it does, because this article used to promise something rather different.

What happened to Simply Insights

Simply used to ship a set of named analytics modules that scored individual conversations, profiled individual candidates, and benchmarked individual recruiters against each other. Those modules are gone. What replaced them is narrower, and the narrowing was the point.

Reporting is now dashboards you build yourself over the objects in your own data model. Jobs, applications, placements, companies, and the object you added last month because your desk needed it. Because the data model is metadata rather than code, the reporting over it is too: a field you create on Tuesday is a dimension you can group by on Tuesday.

The aggregation engine does two things and does them properly. It buckets a measure over time, and it breaks a measure down by a dimension. Between them those cover most of what an agency actually asks on a Monday morning. Conversion between stages comes along with it, because stages in Simply are data rather than labels on a board.

Then there is the constraint that defines the whole thing: every number is an aggregate, and there is no way to click through one to the rows underneath.

What the dashboards answer

Six questions cover most of what gets asked, and not one of them requires knowing anything about a specific person.

• Where the pipeline is stuck. Applications by stage, and how long they have been sitting in the one they’re in.
• What converts. Stage-to-stage conversion per job, per client or per type of work, counted from the applications themselves rather than from a status somebody remembered to update.
• Where placements come from. Source, client and job type behind the hires you actually made, taken from the placement records.
• Workload per desk. Open jobs, live applications and outstanding tasks per recruiter, without opening anybody’s records.
• What the AI is costing. Spend per model and per action, in euros.
• Anything you added yourself. A custom field is a dimension the day you define it.

Notice what isn’t on that list. No personality profile, no call-quality score, no ranking of your recruiters by talk ratio. Those were interesting numbers. They were also the kind of number that quietly turns a dashboard into a surveillance tool, and they are not in the product.

Practical applications

Coaching, at the level where it works

Coaching from aggregates works better than it sounds. If intakes for one client consistently produce applications that die at the same stage, that’s a coachable pattern, and the dashboard surfaces it without anybody’s conversation being scored.

What it doesn’t give you is a leaderboard. If you want to coach a specific recruiter on a specific conversation, you do it the way you always did: you sit down together and open it. The transcript is anchored to the audio, so you can click a line and hear it said instead of arguing about what was meant. That’s a record, and opening a record needs permission to open that record. The dashboard deliberately isn’t a shortcut around it.

Process improvement

Dashboards are at their best on the boring questions. Where does the time go between intake and first presentation? Which stage holds applications longest? Is the bottleneck the summary, the hiring manager’s calendar, or an intake that was too thin to work from?

One honest caveat. Metrics built on stage transitions are computed from transitions that actually happened in Simply, so if you migrated historic data, the first months of a trend line are only as detailed as your old export was. Worth knowing before you put that chart in front of a board.

Quality assurance

As a team grows you want output to stay consistent rather than depend on who happened to take the call. Reporting gives you the baseline: how many applications reach each stage, per desk, per period. Consistency in the work itself comes from the summary formats rather than from a chart, but the chart is where you notice it slipping.

Aggregate only, and what that costs you

Most reporting layers let you click a bar and get the list behind it. Simply doesn’t, and it’s worth saying why, because a buyer who doesn’t know the reason files it as a missing feature.

Drill-through is the back door every reporting layer eventually becomes. The aggregate is harmless. The list underneath it is a copy of your database, reachable by anyone who was ever handed a dashboard. Once that path exists, your permission model has a hole in it that nobody can see from the outside.

So the check happens per request, across four layers: workspace, object, record and field. Authority is re-derived every time rather than frozen into a saved report, which means a permission change takes effect on the next load instead of the next login. And the check fails closed. Where authority for a slice of data can’t be established, that slice is missing from the total rather than quietly included on the assumption that it’s probably fine.

The visible consequence is that two people can open the same dashboard and legitimately see different totals. A recruiter with team-level access and a partner with workspace-level access are counting different sets. The dashboard says what it’s counting, so nobody has to guess where the difference came from.

This costs you something real. You cannot click a number and get the names. If you need the names you go where the records are, and the permissions apply in full. We think that trade is the right way round for candidate data. Not everyone will agree, and it’s better to disagree about it now than in month four.

Data, privacy and trust

Reporting on recruitment data touches privacy from two directions at once, and the two are worth separating rather than blurring into one reassuring sentence.

The platform and your candidate data are hosted in the Netherlands, on infrastructure Simply runs, and Simply is ISO 27001 certified. AI processing runs in European regions, or on your own key with your own provider. The models themselves come from OpenAI and Anthropic. That last part gets said plainly here, because it’s the first thing a security review checks and the answer doesn’t improve by being vague about it.

The record-level counterpart to a dashboard is the audit log: who did what to which record and when, including the attempts that were denied. That’s where per-record questions get answered, under per-record permissions. Two surfaces, two purposes, and the dashboard isn’t a way into the other one.

From data to action

A dashboard nobody acts on is a slower spreadsheet. The numbers worth building are the ones with a decision attached to them.

• Applications pile up at the same stage three weeks running? That stage needs a person, or needs removing. Not another report.
• Conversion from presentation to interview drops for one client? The briefing was probably wrong, and the intake is where to look.
• One desk carries twice the open jobs of the others? That’s a capacity question you can finally put a number on.

The AI cost ledger belongs in the same list. You connect your own key and your provider bills you directly, which is honest and completely useless if you can’t see where the money went. The ledger records what each model and each action cost in euros rather than in tokens or credits, broken down per model, so moving a routine job from the intelligent tier to the fast one is a decision you can price before you take it. No credits, and no markup on what your provider charges. Pricing sets out why.

How reporting fits the rest of Simply

Reporting has no data of its own. Everything it counts was put there by the rest of the product.

Your data model decides what there is to report on in the first place.

Jobs and pipelines make stages into data, which is what conversion between them is counted from.

Data extraction and field proposals fill the fields, with a confidence score per row and a human approving what lands.

Transparency handles the record-level questions a dashboard isn’t allowed to answer.

The public API serves the same objects under the same permission rules, so a warehouse or BI tool reads exactly what the account it authenticates as may read. There’s no export path that skips the check, which is rather the point of having one.

Team level versus individual level

This is the distinction the whole design turns on, so it’s worth ending on.

At team level the questions are about the set. How many, how fast, how often, in which direction. Those are the questions a partner asks on Monday and the ones a board asks in April, and a dashboard answers them without needing to know anything about a specific person.

At individual level the questions are about a person. What did this candidate say about their notice period. Why did that application stall. Those are real questions with real answers, and the answer lives in the record, under the permissions that apply to the record, with an audit entry for having looked. Not behind a chart.

The version of this product that existed two years ago blurred that line, and the dashboard sat on the wrong side of it. Scoring individual recruiters against each other is easy to build, easy to demo, and hard to defend to the person being scored. Aggregates plus a properly permissioned record is the harder version, and the one that survives a works council, a DPIA and an uncomfortable conversation.

If you want to see it against your own numbers, bring the report your board currently asks for. The one that takes a CSV and an evening. We’ll build it live, and show you exactly what it does and doesn’t let the person reading it click into.