← All posts

November 6, 2026

Stop Hiring Analysts Before You Hire Engineers

Seth Therrien

Your company hires three analysts. Reporting should improve. Instead, three more people are cleaning the same spreadsheets.

That is the analyst-first trap. A growing company sees a reporting backlog and assumes it needs more people to answer questions. But knowing when to hire a data engineer can prevent that mistake. Analysts cannot produce reliable answers when customer records conflict, source files arrive late, and every department calculates revenue differently.

Consider a team where analysts spend four days each week preparing data and one day explaining results. That is an illustrative 80/20 split, not a universal industry statistic. The problem is not weak analytical talent. It is missing infrastructure that forces skilled employees to repair data before they can study it.

For a company with 50 to 500 employees, the better sequence is usually to establish a dependable data foundation, prove the numbers, and then expand analytical capacity. Hiring can still happen in parallel. What should not happen is expecting analysts to substitute for an engineering function indefinitely.

The Analyst-First Trap

Hiring analysts feels like the fastest response to a business problem. Executives want sales performance, margins, customer retention, and operational reporting. An analyst seems closer to those questions than a technical infrastructure hire, especially when the company already owns several reporting tools.

The first analyst can often make progress with exports and spreadsheets. The second creates reusable queries. By the third hire, the business may have a library of reports, but not a system that reliably produces the information behind them.

Three things usually go wrong:

- Cleanup becomes the job. Each request begins with missing IDs, duplicate records, inconsistent dates, or manual joins between systems.

- Different answers persist. Finance, sales, and operations use separate definitions, so identical questions produce competing numbers.

- The process depends on individuals. A report works because one analyst remembers which file to export and which cells to correct.

The company may mistake the rising workload for proof that it needs another analyst. In reality, every new hire inherits the same unresolved data problem.

What Data Engineers Build That Analysts Should Not Have To Maintain

Analysts and data engineers solve different parts of the same business problem. Analysts interpret patterns and explain decisions. Engineers establish repeatable, controlled data movement and storage so the same inputs produce trustworthy outputs tomorrow.

Reliable Data Pipelines

A pipeline replaces the recurring export-and-clean routine. It collects records from source systems, handles expected changes, and updates shared datasets on a defined schedule. Failures become visible instead of quietly changing a spreadsheet.

An analyst may build a useful prototype. An engineer makes the production process dependable, recoverable, and maintainable as sources and reporting needs change.

A Shared Data Warehouse

A warehouse gives departments a common reporting base. It brings important business records into one governed environment, with clear rules for how customers, invoices, orders, and activity connect.

That does not require moving every record on day one. The first useful foundation might support only revenue, customer activity, and one operational metric.

Quality Checks And Monitoring

Reliable numbers require more than successful file transfers. Data checks can catch duplicate transaction IDs, missing account references, unexpected totals, and failed updates before reports use those records.

Monitoring also identifies who responds when a source changes. Without ownership, a failed refresh can remain unnoticed until leadership challenges the dashboard.

Reusable Definitions And Documentation

Shared definitions prevent each analyst from inventing another calculation. Finance and operations should agree on what revenue, active customers, and churn mean before those rules become reusable reporting models.

Engineers translate agreed rules into maintained processes. Analysts can then test business questions rather than repeatedly rebuilding the source data.

The Cost of Hiring In The Wrong Order

The real cost is not simply the number of salaries. It is the proportion of analytical capacity consumed by repetitive preparation, reconciliation, and investigation. More headcount can increase this cost when every person repeats the same tasks.

Consider three analysts working 40-hour weeks. If each spends 80% of the week cleaning and joining data, the weekly allocation looks like this:

- 96 hours on preparation. Three people spend 32 hours each collecting, correcting, or reconciling records.

- 24 hours on analysis. Only eight hours per person remain to interpret findings and advise the business.

- 120 hours on payroll. All of that work is paid for, but only part directly supports decisions.

This is a hypothetical capacity example, not a measured company result or a promise that automation removes every preparation task. It shows why adding analysts without fixing repeatable data work can disappoint leadership.

The same pattern can reach executives. When finance leaders assemble and reconcile reports themselves, the opportunity cost becomes larger. The article your CFO is your most expensive data engineer examines how to assess that burden without confusing staff time with guaranteed cash savings.

When To Hire A Data Engineer? The Right Hiring Sequence

The order of work matters more than the size of the initial team. A 60-person company with several systems may need engineering sooner than a larger business with dependable reporting. Start with the recurring decisions and their underlying data.

1. Identify the decisions. Choose the reports leadership uses to manage revenue, customer activity, delivery, or financial performance. Assign a business owner to each important metric.

2. Map the sources. Document where each figure originates, how it is currently refreshed, and where manual corrections occur. Start with critical systems rather than every application.

3. Build a limited foundation. Establish a shared data store, a small set of scheduled pipelines, basic access controls, and validation for the highest-value reporting needs.

4. Verify the outputs. Reconcile known-good totals with finance and operations. Confirm that stakeholders agree on definitions before scaling the reporting layer.

5. Add and specialize the team. Hire analysts to investigate and explain the data. Add permanent engineering ownership as source integrations, change frequency, and reliability requirements grow.

Companies already recruiting should not necessarily pause hiring. They can prepare the initial system while interviews continue. For a more detailed preparation plan, see before your first data engineer starts.

Where Fractional Data Engineering Fits

A permanent engineering hire is not always the first practical move. Some businesses need senior technical judgment and a working starting system before they have enough ongoing engineering demand for a full-time role.

A fractional engagement can build the limited foundation, document the decisions, and transfer ownership to an internal hire or existing team. It should leave reusable systems, not a growing dependency on outside consultants.

Start With A Bounded First Release

Choose one useful reporting workflow, not an all-company rebuild. For example, connect core sales and finance data, establish agreed revenue calculations, and test the outputs against a known-good period.

A two-week initial foundation can be a practical target for a tightly scoped effort with access and definitions ready. It is not a promise to complete every integration, dashboard, or edge case in that period.

Make The Handoff Part of The Work

The permanent owner should receive a system they can understand and extend. That means source documentation, approved definitions, code and configuration, test results, access procedures, a recovery process, and a prioritized backlog.

A useful handoff includes a walkthrough with the receiving team. The goal is for the next engineer to improve the system without reconstructing why each decision was made.

Signs You Have An Engineering Problem, Not An Analyst Problem

More analysis will not repair unreliable inputs. These symptoms suggest the business should examine data engineering before adding another analyst. The common thread is recurring preparation work that exists regardless of which report someone requests.

- Reports disagree on core totals. Two teams show different revenue or customer counts for the same period.

- Refreshes depend on manual exports. A recurring dashboard starts with someone downloading, renaming, and uploading files.

- Source changes break reports. A renamed field or altered file format causes recurring repairs across several dashboards.

- No one owns data failures. When a refresh stops or totals change, the team cannot identify an accountable owner.

- Analysts rebuild the same transformations. Several people independently clean identical inputs or maintain competing versions of a metric.

- New hires spend weeks finding data. Onboarding starts with detective work instead of improving reports or investigating decisions.

An analyst may still be the correct next hire if the data is already dependable and the business lacks capacity to interpret it. The distinction depends on the work actually causing the backlog.

What To Measure Before Your Next Hire

Measure the bottleneck before approving another position. A short review of recurring reports and the people maintaining them can distinguish an analysis shortage from a data preparation problem. Record a baseline before changing the process.

- Preparation time. Track weekly hours spent gathering, cleaning, and reconciling data versus hours spent on analysis.

- Reporting reliability. Count missed refreshes, recurring errors, and reports that require manual intervention.

- Metric agreement. Check whether finance and operations can reconcile a small set of important totals.

- Time to an answer. Measure how long a routine business question takes from request to verified response.

Recheck the same measures after improving the first pipelines. If preparation falls and answers become more consistent, the next analyst can contribute to the work leadership actually needs.

Conclusion

Hiring more analysts will not automatically fix unreliable data. The work underneath the reporting layer must be dependable first.

Build the shared sources, scheduled pipelines, and checks that make accurate analysis repeatable. Keep the first scope narrow enough to verify.

A fractional engineer can establish that foundation while permanent hiring continues, then transfer it with clear documentation and ownership.

Once the inputs are trusted, analysts can spend more time interpreting results instead of repairing the same records again.