Vanity Metrics in Singapore Marketing: Which Numbers Actually Matter
Every month we get sent a marketing report that a business owner describes, carefully, as “quite good?” — with the question mark doing a lot of work. Impressions up 340%. Reach into the hundreds of thousands. Engagement rate above the industry average. Follower count climbing. And then, somewhere near the bottom of the same page, enquiries flat and revenue down.
Nobody in that situation is lying. The numbers are real. They were pulled from the platform, they went up, and the person who assembled them is not trying to deceive anybody. The problem is that most marketing reports are built out of metrics that are extremely easy to move and almost impossible to act on — and once a report is full of those, it stops being a management tool and becomes a monthly reassurance exercise.
This guide is about telling the difference. Not the lazy version of the argument, where “impressions bad, revenue good” and everything in the middle gets deleted, because that throws away the diagnostic numbers you need when something breaks. The useful version is a three-tier test you can apply to any figure in any report in about sixty seconds, plus the specific traps that come from measuring a market the size of Singapore.
What actually makes a metric a vanity metric
A vanity metric is not a “soft” metric, and it is not the same as a top-of-funnel metric. Brand awareness is soft and hard to measure; it is not automatically vanity. Impressions sit at the top of the funnel; in a frequency investigation they are essential.
The distinction that holds up is behavioural, and it comes from the analytics literature rather than from marketing. Eric Ries drew the line between vanity metrics and actionable metrics in The Lean Startup: a metric is actionable if it demonstrates cause and effect clearly enough that you can repeat what worked. Alistair Croll and Benjamin Yoskovitz sharpened it in Lean Analytics into a checklist that has aged extremely well — a good metric is comparative, is understandable, is usually a ratio or a rate rather than a raw total, and, above all, changes how you behave.
That last one is the whole test. If the number went up 40% and the number went down 40%, and in both cases you would do exactly the same thing next month, it is a vanity metric in your report. Not in general — in your report, for your business, at your stage.
Which is why the same figure can be vanity for one company and load-bearing for another. Follower count is decoration for a B2B engineering firm whose deals come from three named accounts and a LinkedIn DM. For a Singapore F&B brand that runs flash promotions to its own audience and can measure the redemption, follower count has a direct line to revenue and belongs on the front page.
In practice we sort every number in a client report into one of three tiers.
Note the wording on the bottom tier: keep them tracked, keep them out of the report. Deleting the data is a mistake. Impressions are how you notice that a campaign stopped delivering. They just should not be the headline, because the headline is what the room argues about.
The deeper problem: one word, six different definitions
Before you can decide whether a metric is useful, you have to know what it counts — and this is where most reports quietly fall apart. The same English word means materially different things on each platform, and dashboards happily sum them into a single row labelled “total views” as though they were the same unit.
A “view” is the worst offender. Meta moved to a unified Views metric across Facebook and Instagram, counting a view essentially as soon as the content begins playing, in line with how Reels were already measured, and replacing watch-time labels with minutes-viewed equivalents. TikTok counts a view almost immediately on playback. YouTube’s paid TrueView view has historically required 30 seconds, the full ad if shorter, or an interaction. LinkedIn counts a sponsored video view at two seconds with at least half the player in view.
The same fragmentation runs through every common term:
| Word in the report | What it can actually mean | Why it matters |
|---|---|---|
| Clicks | Meta separates link clicks from clicks (all), which includes reactions, comments, profile taps and image expansions. Google Ads separates clicks from interactions. | Reporting “clicks (all)” makes cost per click look roughly half what a visitor actually costs. |
| Engagement | Some platforms and tools count likes and comments only; others add shares, saves, and video plays. Denominators vary too — per follower, per reach, or per impression. | Cross-platform engagement-rate comparisons are usually invalid. The gap between two platforms is often a definition, not performance. |
| Sessions vs users | GA4 counts a session per visit; users are deduplicated by device and consent state, not by person. | One customer researching on a phone and buying on a laptop is two users. Small-market reports over-count reach this way constantly. |
| Engaged session | In GA4, a session that lasts longer than 10 seconds, or has a key event, or has two or more page or screen views. | “Engagement rate” in GA4 is not an opinion about content quality; it is that specific three-way test. Bounce rate is simply its inverse. |
| Conversions | Platform-attributed conversions counted at the date of the click, not the sale, on each platform’s own window. | Adding Meta conversions to Google conversions double-counts. See our guides to conversion tracking and attribution models. |
None of this is a scandal. Platforms measure what their own product needs to measure. It only becomes a problem when a monthly report presents the sum as a fact, and a decision gets made on it.
The usual suspects, and what to put in their place
Here are the metrics that most often occupy the front page of a Singapore SME’s report, why each one flatters, and the replacement that answers the same underlying question honestly.
| Vanity metric | Why it flatters | Replace with |
|---|---|---|
| Impressions | Rises with budget and with frequency. In a market of Singapore’s size, the same people absorb most of the increase. | Reach against your addressable audience, plus frequency. Then cost per qualified outcome. |
| Follower count | Only ever goes up, is trivially purchasable, and correlates weakly with reach on every algorithmic feed. | Returning audience, saves and shares, and branded search volume — the signals that someone chose to come back. |
| Total likes | The cheapest possible action, and the one most influenced by posting volume rather than content quality. | Shares, saves and replies. Buffer’s within-account data associates replying to comments with materially higher subsequent engagement, which is a behaviour you control. |
| Email open rate | Since Apple’s Mail Privacy Protection shipped in September 2021, Apple’s proxy servers pre-fetch the tracking pixel, so a large share of “opens” are machine-generated. Vendors put the inflation anywhere from 15% to a doubling, depending on list composition. | Click-to-conversion rate, and revenue per email sent. Both survive proxy pre-fetching. |
| Raw page views | A slow, confusing site with multi-step navigation generates more of them than a fast one. | Engaged sessions to a key page, and the conversion rate of that page. |
| Platform-reported ROAS | Each platform counts conversions it believes it caused, on its own window, often on gross order values that include 9% GST. | Blended return across all spend, checked against the P&L. See ROAS versus ROI. |
| “Leads” | An unqualified form fill, a competitor, and a genuine buyer all count as one. | Qualified leads, defined jointly with whoever answers the phone, then cost per acquisition. |
One consistent theme: the replacement is almost always a rate, a ratio, or a cost rather than a total. Totals go up when you spend more, which is exactly why they feel good and tell you nothing.
The tier people wrongly delete
The overcorrection is as damaging as the original problem. A business owner reads an article like this one, strips the report down to revenue and cost per acquisition, and then has no way to answer the only question that matters when the number goes bad: which part broke?
Diagnostic metrics exist for that moment. Consider a Singapore services business whose cost per enquiry rises from S$62 to S$104 in a month. Revenue-only reporting tells you it happened. The middle tier tells you why, and each candidate points at a different fix:
| What the diagnostic shows | What it means | What you actually change |
|---|---|---|
| Impressions flat, click-through rate down | The audience is seeing you and choosing not to click. | Creative and offer. Nothing in the account settings will fix this. |
| Click-through rate flat, frequency up sharply | You are showing the same small audience the same ad repeatedly — a standing risk in a market this size. | Widen the audience or refresh creative. Raising budget makes it worse. |
| Clicks up, landing page conversion rate down | Traffic quality or page experience changed, not demand. | The page, the form, or the match between ad promise and page. |
| Everything flat, impression share lost to rank | A competitor entered the auction. | A bid, quality, or positioning decision — a commercial call, not a creative one. |
| All platform metrics flat, enquiries down | Tracking broke, or the phone is not being answered. | Check tracking first. It is the most common answer and the least often checked. |
Same headline number, five completely different responses. That is what the middle tier buys you, and it is why “only report revenue” is bad advice dressed as discipline.
Four traps specific to measuring a market this small
1. Reach numbers sound enormous because the country is not. DataReportal’s Digital 2026 Singapore figures put Facebook’s local ad reach at about 3.80 million, Instagram at 3.35 million, and TikTok at 3.80 million adults. A campaign reporting 500,000 reach has touched something like an eighth of the entire addressable audience on that platform. Impressive-sounding, and also the reason frequency climbs so fast here: there is nowhere else for the budget to go. In a larger market you can spend your way out; in Singapore you run out of people and start paying to repeat yourself.
2. Platform “reach” is not monthly active users. DataReportal says so explicitly in its own notes, and the figures get revised. The clearest example is LinkedIn, which reports around 5.10 million registered members in Singapore — a number that implies adult reach above 100% and therefore cannot be compared with Facebook’s or Instagram’s ad-reach figures. Any report that lines those four numbers up in one chart is comparing different units.
3. GST quietly inflates your return. If your purchase pixel passes the gross order value, every revenue figure in your ad accounts includes 9% GST that never belonged to you. That is roughly an 8.3% overstatement of ROAS before anything else goes wrong. The platform side has its own asymmetry: Google bills Singapore advertisers through its local entity and charges 9% GST, reclaimable as input tax if you are registered, while Meta bills under the overseas vendor registration regime and charges none once you supply a GST number. The two platforms only compare on a GST-exclusive basis.
4. Benchmark tables are mostly marketing. The “average CPC by industry in Singapore” tables that circulate every year are compiled from individual agencies’ private client bases, with no stated methodology and no sample size. They are worth a glance as a sanity check and worthless as a target. Your own trailing three months is a better benchmark than any of them, because it is the only dataset that shares your offer, your margins and your sales process.
Rebuilding the report: one page, four questions
A monthly report only needs to answer four questions. Everything else is an appendix.
| Question | The metric that answers it | What a decision looks like |
|---|---|---|
| Did we get more of what we sell? | Qualified leads or orders, versus the prior period and the same month last year | Scale, hold, or investigate |
| What did each one cost? | Fully loaded cost per acquisition, including fees, tools and staff time | Shift budget between channels |
| Was it worth it? | Gross profit per customer against that cost, and the payback period in months | Change the offer, the price, or the target |
| Is it durable? | Repeat rate or retention, and branded search volume | Invest in retention rather than acquisition |
A worked version for a Singapore SME spending S$8,000 a month across Google and Meta: 61 qualified enquiries, fully loaded cost per enquiry S$187, close rate 22%, so 13 customers at a fully loaded acquisition cost of about S$877. Gross profit per customer S$1,450, giving a payback inside the first sale and a contribution of roughly S$7,400 for the month. Repeat rate 31%, branded search up 12% quarter on quarter.
Nine numbers. Every one of them is a rate, a cost, or a count of something the business actually sells — and each maps to a decision. The impressions figure for that account was just over 2.1 million, which is a fine thing to know and a terrible thing to lead with. If you want the numbers laid out so they stay visible without being rebuilt by hand every month, that is what a marketing dashboard is for, and our guide to reading a marketing report covers how to interrogate one you did not build.
The sixty-second test
Take any number in your current report and ask, in order:
- Is it a total, or a rate? Totals rise with budget. Be suspicious.
- Do I know exactly what it counts? If you cannot state the threshold or the denominator, you cannot interpret the change.
- Could it double without the business improving? If yes, it is not a decision metric.
- If it halved, what would I do on Monday? If the honest answer is “nothing”, move it to the appendix.
Most reports lose half their front page to this test, and become considerably more useful in the process. The half that survives tends to be uncomfortable, which is the point — a report that never delivers bad news is not measuring anything.
Frequently asked questions
Are impressions always a vanity metric?
No. Impressions are a vanity metric as a headline and a diagnostic metric in context. Paired with reach they give you frequency, which is one of the most actionable numbers in a small market. Paired with clicks they give you click-through rate. On their own, as a monthly total that goes up whenever budget goes up, they tell you nothing you did not already know from the invoice.
Is engagement rate a vanity metric?
Usually, in the way it is normally reported. The problem is comparability: platforms and tools use different numerators and different denominators, so a cross-platform engagement chart is often comparing definitions rather than performance. Engagement rate is genuinely useful in one narrow way — tracked against your own account over time, on a single platform, with the definition held constant. Our guide to measuring social media ROI goes into which social numbers survive contact with a P&L.
What about follower count for a Singapore B2B brand?
For most B2B firms here it is a tier 3 metric, because the buying committee is small, known, and unlikely to be reached by follower growth. What matters far more is whether the right ten to thirty people at your target accounts see the content at all, which follower count does not measure. That said, follower count has one legitimate use: as a credibility check for a prospect who looks you up mid-deal. Treat it as a hygiene threshold to clear, not a growth metric to optimise.
Should I report ROAS to my board?
Report it with the P&L number beside it, or not at all. Platform ROAS is a revenue-to-media-spend ratio that ignores cost of goods, fees, staff and GST, which is why a 4:1 ROAS routinely coexists with a loss. If the board only sees one figure, make it gross profit generated against fully loaded marketing cost.
How many metrics should a monthly report contain?
On the front page, four to six. In the appendix, as many as you like. The constraint is attention rather than data: a page with thirty numbers gets skimmed for the green ones, which is exactly the failure mode this whole article is about. Depth belongs behind the summary, not in it.
Can brand awareness be measured without a big research budget?
Partly, and the proxies are better than most people assume. Branded search volume over time is the strongest cheap signal, because typing your name is a deliberate act. Direct traffic, returning visitor share, and the rate at which enquiries mention you unprompted all point the same way. None of them is a brand-lift study, and you should not present them as one — but tracked consistently over quarters they will tell you whether awareness is moving.
What to do with this
Open last month’s report and run the four-question test on every figure on the front page. Move everything that fails into an appendix — do not delete it, because you will want it the next time something breaks. Then check that the numbers left standing are defined the same way in every source they come from, which is usually where the second round of surprises lives.
The report that survives will be shorter, less flattering, and considerably more useful in a meeting. It will also make the following month’s decisions faster, because there is nothing to argue about except the things you can change.
If you would rather not do that surgery yourself, our performance marketing team rebuilds client reporting around decision metrics as the first step of any engagement — tracking verified, definitions documented, numbers reconciled against actual sales. The wider framework sits in our performance marketing guide, the profitability arithmetic in customer lifetime value, and the reports we build for clients are visible in our case studies. If your GA4 is the source of the confusion, start with the GA4 reports worth checking.
Last updated 2 August 2026. Written by Adrian Tan and the SDM team. Sources: Google Analytics 4 help documentation on engaged sessions, engagement rate and bounce rate; platform help documentation on video view counting for Meta, TikTok, YouTube and LinkedIn; Apple Mail Privacy Protection documentation and vendor impact analyses; DataReportal Digital 2026 Singapore; Buffer’s State of Social Media Engagement; IRAS GST guidance. Eric Ries, The Lean Startup, and Alistair Croll and Benjamin Yoskovitz, Lean Analytics, for the actionable-metric framework.
