Rules
Part of What analytics tools actually do: recording, modeling and drawing the numbers
Common analytics tools mistakes agencies repeat, from metrics to shared logins
Avoid analytics tools mistakes agencies repeat: undefined metrics, attribution settings and shared logins in influencer campaign reporting.
The costly mistakes with analytics tools are rarely the choice of vendor. They are the quiet ones that make a report wrong without making it look wrong, or that make the tool impossible to leave. Here are the failures that turn up most in agency work, grouped by when they happen.
On influencer campaigns, the reporting chain may include CreatorIQ, GRIN, or Aspire for creator management; Instagram Insights, TikTok Analytics, or YouTube Studio for sponsored-post results; and GA4 or Adobe Analytics for site conversions. Keep platform-reported views and engagement separate from link-attributed site conversions, and record creator and post identifiers, metric definitions, reporting period, and source before comparing. Check whether an export or API supplies the fields the report needs.
What to take away
- Most bad numbers come from a definition nobody wrote down, not from a broken tool.
- A tracking change mid campaign splits the record in two, and the split almost never makes it into the report.
- Anything you cannot export, you do not really have.
Mistakes made before you buy
Judging a tool by the demo. The demo has clean data, a finished setup, and a question chosen by the person selling. Your own setup may have none of those. Ask to run your own awkward sample instead, even if it is small.
Pre-purchase checks
- Run your own awkward sample data
- Check fit at the small end
- Map which pricing line scales
- Ask notice periods and raw export
Buying for your biggest client. The tool then gets rolled out to accounts whose volume, budget, and technical access look nothing like the one that justified the purchase. Check the fit at the small end before you commit, and settle the requirements list against the whole portfolio.
Reading price as a number rather than a shape. Seats, tracked events, data volume, and connected sources can scale differently as campaign activity grows. Compare quotes by billing unit, and ask which unit changes when activity increases; do not treat unlike billing bases as comparable.
Leaving the exit until later. Review the data-processing agreement, the contract's termination notice period, and the available export formats before you commit. Confirm that you can retrieve and use the data without the vendor's help.
Mistakes made during setup
No owner for the tracking plan. Events get named by whoever adds them, and over time the same action can end up with multiple names, some of them wrong. Name one owner, keep one document, and make new events go through it.
Setup mistakes to fix
- Name one owner for the tracking plan
- Date any mid-campaign tracking change
- Replace shared logins with named seats
- Name the attribution model every time
- Check what bot filtering removes
Changing tracking in the middle of a campaign. Changing a creator's campaign tag, link destination, or conversion event changes what the report captures; before-and-after data no longer measure the same thing. If the change cannot wait, record when it happened and exactly what changed, report the periods separately, and do not compare across that line without explaining it.
One login for the whole team. No audit trail, no per-client permissions, and a leaver who takes a shared password with them. Use individual accounts and appropriate permissions rather than treating shared credentials as a cost-saving substitute.
Accepting the default attribution model as a fact. A default is a vendor setting, not an inherent property of the campaign. Record the selected model and attribution window; models such as last-click, linear, data-driven, and position-based may be available depending on the product and configuration.
Trusting bot and spam filtering without checking it. When reconciling GA4 with Adobe Analytics, review traffic filters and exclusions in both, and keep social-platform-reported engagement distinct from site analytics. Find out what each source is removing before building a story on what remains.
Mistakes made in the reporting
Comparing two tools and declaring one wrong. In a GA4-versus-Adobe Analytics review, compare the identity setting, attribution model and window, time zone, reporting currency, and conversion definition used in each for the same campaign and date range. A different attribution window or reporting currency can make totals non-comparable, so record each tool's actual settings before interpreting the gap. Check model names such as last-click, linear, data-driven, or position-based where available.
Reporting metrics with no decision attached. If nobody can say what they would do differently at a higher or lower value, the metric is decoration. For creator campaigns, define views, engagement, link clicks, and conversions separately and identify their sources. Cut a metric, or attach the decision it is supposed to drive.
Letting the client and the agency mean different things by conversion. This survives for months because both sides assume agreement. Put the definition in the report, in the client's own words, and keep it there.
Reporting a number you cannot reproduce. If you cannot get back to the same figure from the same source, do not put it in front of a client. Screenshots are not a method, and the record that lets somebody retrace a figure is an ordinary audit trail rather than a compliance luxury.
| Mistake | What it costs | The fix |
|---|
Mistake, cost, fix
Mistake
- Undefined metric
- Argument about wording
- Tracking changed mid campaign
- Untrustworthy comparison
- Shared login
- No audit trail
What it costs
- Undefined metric
- One written definition
- Tracking changed mid campaign
- Date and annotate chart
- Shared login
- Named seats, permissions
The fix
- Undefined metric
- Tracking changed mid campaign
- Shared login
How to catch these early
- Pick a number in a draft report and ask the writer to name its source. Do this often enough that people expect it.
- Keep a dated change log next to the tracking plan, and put the latest change in the client report.
- Reconcile one metric against a second source regularly. Not to find a winner, but to find out when the gap changes.
- Review permissions when someone joins or leaves, rather than when something goes wrong.
- Before a renewal, run an export and confirm you can read the file without help from the vendor. The same habit, applied across every system an agency runs, is the substance of structured supplier comparison.
A metric that starts as a description and becomes a target can stop describing anything, a concern captured by Campbell's law. Where an agency and a client disagree about what a number means, write the definition down before either side argues about the value.
Common questions
Which of these causes the most damage?
Undefined metrics can cause substantial damage. They are easier to fix before reports rely on them and harder to unpick after those reports have gone out.
We already made the mid campaign change. What now?
Say so in the report. Show the two periods separately, note the date and what changed, and resist the urge to present a single trend line across the break.
Is it worth switching tools to fix these?
Usually not. Almost all of these are process problems that follow you to the next tool. Fix the plan, the owner, and the definitions first, then judge whether the tool is still the constraint. For example, keep sponsored-post engagement from Instagram Insights separate from site conversions recorded in GA4, and label each source and conversion definition. The worked examples can help you apply the same checks to your setup.
Do the same mistakes apply on the campaign side?
Yes. Check that creator and post identifiers survive platform exports, confirm which sponsored-post metrics each platform provides, and reconcile them with site conversions using the same campaign tags. Record what the export or API actually includes rather than assuming every source supplies the same fields.
