Performance marketing of a brand in social networks: how to calculate real ROI
Why social networks, Google Analytics, and CRM show different numbers
Facebook and Instagram record conversions if they occur within the platform's attribution window — usually 7 days after a click or 1 day after a view. GA4 works on a different logic: event-based data model instead of sessions allows seeing the user journey across different channels and devices, but the attribution window is different there and does not match what "Meta sees". CRM does not know about clicks at all — it records a deal with the date when the manager closed it, which may be a month after the first touch in social networks.
As a result, the same sale can 'belong' to three channels simultaneously, or it may not belong to anyone at all — if the attribution is set up incorrectly. The only way to get a true picture is to link three data sources through cross-channel analytics: user ID, UTM tags, and CRM deal must be stitched together into one chain, preferably at the CRM or BI system level, rather than manually cross-referenced in three Excel sheets.
ROI and ROMI are different metrics, and they must not be confused.
ROI (Return on Investment) calculates the return on all investments in the project — including salaries, tools, and overhead costs. ROMI (Return on Marketing Investment) accounts for only marketing expenses and almost always looks more impressive, which is why agencies like to show clients this metric.
Example of ROMI calculation: revenue 200 000 rubles, advertising costs 50 000 rubles. (200 000 − 50 000) / 50 000 × 100% = 300%. A beautiful figure. But if you need to add the salary of a targetologist, a designer who created the creatives, an analytics service subscription, and part of a content manager's work to this campaign — the actual ROI will be considerably lower.
| Cost item | Included in ROMI | Included in ROI |
|---|---|---|
| Advertising budget | Yes | Yes |
| Targetologist/marketer salary | No | Yes |
| Analytics and automation tools | No | Yes |
| Content and creative production | No | Yes |
| Agency commission | Sometimes | Yes |
If you report to management that ROMI is 300%, but in reality, ROI is negative when all costs are taken into account — this will eventually surface during a budget audit, and trust in the marketing department will drop more than it would with an honest report from the start.
Attribution resolves the dispute over whether "the ad worked"
An attribution model is a rule by which the value of a conversion is distributed among touches. For branded campaigns in social networks, where the customer journey rarely fits into a single click, the choice of model changes the result by dozens of percentage points.
- Last Click — the entire conversion value goes to the last touch. A simple model, but it systematically undervalues the top of the funnel: reach and engaging social media posts look useless, although it is precisely they that form demand.
- Linear — the value is divided equally among all touches in the chain. Fairer to the top stages of the funnel, but crude: a touch one month before purchase and a touch one hour before it receive equal weight.
- Time Decay — weight grows as you approach conversion. Works well for the middle and lower stages of the funnel, where it handles 'closing the deal' rather than the first brand introduction.
- Data-Driven Attribution (DDA) — a model based on real data about conversion and non-conversion paths, available in GA4. Does not assign weight by a rigid rule, but calculates each channel's contribution statistically. This is closer to reality, but requires a sufficient data volume — it does not work stably on small accounts.
Practical tip: for evaluating the top stage of the funnel (reach, brand awareness) use Linear or DDA, for the bottom (retargeting, abandoned cart) — Time Decay or Last Click. Comparing ROI of campaigns calculated in different models is a methodological error that even experienced teams make.
How to set up UTM to avoid losing touches
- Record not only the source and channel, but also the creative, audience and funnel stage in the utm_content parameter — this will allow you to break down ROI by segments, not only by campaign as a whole.
- Use a single UTM template for the entire team — discrepancies in case or parameter order silently break cross-analysis.
- Pass UTM to CRM as a deal field on first touch and preserve the history of all touches, not just the last one — otherwise Last Click will become your default attribution model, even if you didn't plan it.
- Set up a GA4 to CRM linkage via API or connector so that the "lead" event in GA4 maps to the actual deal amount, not just form submission.
Micro- and macro-conversions: where is the real impact lost
The problem that is rarely examined: click and purchase are not the only events that matter. Subscription to an account, saving a post, navigating to Direct, adding to favorites — these are micro-conversions, and they don't fit well into the ROI formula because they don't have direct monetary value at the moment.
But if you consider LTV of a subscriber — the revenue that he will bring during the entire period of interaction with the brand — the picture changes. A user who did not buy after the first click, but subscribed and became a regular customer after six months, is completely 'lost' in the classic ROI by last-click model. In the LTV model, he makes a major contribution to the revenue. Therefore, for branded performance campaigns in social networks, it is reasonable to consider three metrics in parallel: CPL (cost per lead), CPA (cost per action), and LTV — not only the final ROI from direct sales.
What real cases show
The New Zealand bank ASB's campaign "Like a loan" is a good example of what a funnel looks like in reality, not in theory: 21 496 unique entries turned into 17 778 leads, of which 229 became new clients, bringing the bank several million dollars in revenue. The conversion from lead to client here is around 13%, and this is a normal indicator for a financial product, if you count honestly, not by last-click with a one-day window.
The IKEA case with geo-targeting shows another side of the problem: the A/B test yielded 31% increase in store visits for the 22–25 age group and 11% increase in the 26–35 age group from the same creative concept, only with different targeting. The formal ROI of the campaign without geo and age breakdown would have hidden this effect — the averaged figure would have looked modest, although one segment performed multiple times better than the other.
What ROI is considered normal
There is no universal benchmark, and articles that claim "a good ROI is 200%" without industry context are misleading. Below are working ranges to use as a reference when evaluating paid advertising in social networks.
| Industry | Typical ROI range | Comment |
|---|---|---|
| E-commerce (mass market) | 150–400% | Short deal cycle, easy to track by last click |
| B2B / SaaS | 100–250% | Long cycle, requires Time Decay or DDA model |
| Financial services | 200–500% | High client LTV compensates for expensive leads |
| Local business / services | 100–300% | Heavily depends on geo-targeting, see the IKEA case |
| Luxury and branded goods | Often below 100% in direct sales | The main effect is in LTV and brand recognition, not immediate conversion |
If your ROI for direct sales in the luxury segment shows 60%, this is not a campaign failure, but a normal consequence of the fact that the brand is working to build demand, which is then realized later and through other channels. Judging such a campaign by the formula for e-commerce is a methodological error, not an analytical insight.
Why social networks, Google Analytics, and CRM show different numbers
Facebook and Instagram record conversions if they occur within the platform's attribution window — usually 7 days after a click or 1 day after a view. GA4 works on a different logic: event-based data model instead of sessions allows seeing the user journey across different channels and devices, but the attribution window is different there and does not match what "Meta sees". CRM does not know about clicks at all — it records a deal with the date when the manager closed it, which may be a month after the first touch in social networks.
As a result, the same sale can 'belong' to three channels simultaneously, or it may not belong to anyone at all — if the attribution is set up incorrectly. The only way to get a true picture is to link three data sources through cross-channel analytics: user ID, UTM tags, and CRM deal must be stitched together into one chain, preferably at the CRM or BI system level, rather than manually cross-referenced in three Excel sheets.
ROI and ROMI are different metrics, and they must not be confused.
ROI (Return on Investment) calculates the return on all investments in the project — including salaries, tools, and overhead costs. ROMI (Return on Marketing Investment) accounts for only marketing expenses and almost always looks more impressive, which is why agencies like to show clients this metric.
Example of ROMI calculation: revenue 200 000 rubles, advertising costs 50 000 rubles. (200 000 − 50 000) / 50 000 × 100% = 300%. A beautiful figure. But if you need to add the salary of a targetologist, a designer who created the creatives, an analytics service subscription, and part of a content manager's work to this campaign — the actual ROI will be considerably lower.
| Cost item | Included in ROMI | Included in ROI |
|---|---|---|
| Advertising budget | Yes | Yes |
| Targetologist/marketer salary | No | Yes |
| Analytics and automation tools | No | Yes |
| Content and creative production | No | Yes |
| Agency commission | Sometimes | Yes |
If you report to management that ROMI is 300%, but in reality, ROI is negative when all costs are taken into account — this will eventually surface during a budget audit, and trust in the marketing department will drop more than it would with an honest report from the start.
Attribution resolves the dispute over whether "the ad worked"
An attribution model is a rule by which the value of a conversion is distributed among touches. For branded campaigns in social networks, where the customer journey rarely fits into a single click, the choice of model changes the result by dozens of percentage points.
- Last Click — the entire conversion value goes to the last touch. A simple model, but it systematically undervalues the top of the funnel: reach and engaging social media posts look useless, although it is precisely they that form demand.
- Linear — the value is divided equally among all touches in the chain. Fairer to the top stages of the funnel, but crude: a touch one month before purchase and a touch one hour before it receive equal weight.
- Time Decay — weight grows as you approach conversion. Works well for the middle and lower stages of the funnel, where it handles 'closing the deal' rather than the first brand introduction.
- Data-Driven Attribution (DDA) — a model based on real data about conversion and non-conversion paths, available in GA4. Does not assign weight by a rigid rule, but calculates each channel's contribution statistically. This is closer to reality, but requires a sufficient data volume — it does not work stably on small accounts.
Practical tip: for evaluating the top stage of the funnel (reach, brand awareness) use Linear or DDA, for the bottom (retargeting, abandoned cart) — Time Decay or Last Click. Comparing ROI of campaigns calculated in different models is a methodological error that even experienced teams make.
How to set up UTM to avoid losing touches
- Record not only the source and channel, but also the creative, audience and funnel stage in the utm_content parameter — this will allow you to break down ROI by segments, not only by campaign as a whole.
- Use a single UTM template for the entire team — discrepancies in case or parameter order silently break cross-analysis.
- Pass UTM to CRM as a deal field on first touch and preserve the history of all touches, not just the last one — otherwise Last Click will become your default attribution model, even if you didn't plan it.
- Set up a GA4 to CRM linkage via API or connector so that the "lead" event in GA4 maps to the actual deal amount, not just form submission.
Micro- and macro-conversions: where is the real impact lost
The problem that is rarely examined: click and purchase are not the only events that matter. Subscription to an account, saving a post, navigating to Direct, adding to favorites — these are micro-conversions, and they don't fit well into the ROI formula because they don't have direct monetary value at the moment.
But if you consider LTV of a subscriber — the revenue that he will bring during the entire period of interaction with the brand — the picture changes. A user who did not buy after the first click, but subscribed and became a regular customer after six months, is completely 'lost' in the classic ROI by last-click model. In the LTV model, he makes a major contribution to the revenue. Therefore, for branded performance campaigns in social networks, it is reasonable to consider three metrics in parallel: CPL (cost per lead), CPA (cost per action), and LTV — not only the final ROI from direct sales.
What real cases show
The New Zealand bank ASB's campaign "Like a loan" is a good example of what a funnel looks like in reality, not in theory: 21 496 unique entries turned into 17 778 leads, of which 229 became new clients, bringing the bank several million dollars in revenue. The conversion from lead to client here is around 13%, and this is a normal indicator for a financial product, if you count honestly, not by last-click with a one-day window.
The IKEA case with geo-targeting shows another side of the problem: the A/B test yielded 31% increase in store visits for the 22–25 age group and 11% increase in the 26–35 age group from the same creative concept, only with different targeting. The formal ROI of the campaign without geo and age breakdown would have hidden this effect — the averaged figure would have looked modest, although one segment performed multiple times better than the other.
What ROI is considered normal
There is no universal benchmark, and articles that claim "a good ROI is 200%" without industry context are misleading. Below are working ranges to use as a reference when evaluating paid advertising in social networks.
| Industry | Typical ROI range | Comment |
|---|---|---|
| E-commerce (mass market) | 150–400% | Short deal cycle, easy to track by last click |
| B2B / SaaS | 100–250% | Long cycle, requires Time Decay or DDA model |
| Financial services | 200–500% | High client LTV compensates for expensive leads |
| Local business / services | 100–300% | Heavily depends on geo-targeting, see the IKEA case |
| Luxury and branded goods | Often below 100% in direct sales | The main effect is in LTV and brand recognition, not immediate conversion |
If your ROI for direct sales in the luxury segment shows 60%, this is not a campaign failure, but a normal consequence of the fact that the brand is working to build demand, which is then realized later and through other channels. Judging such a campaign by the formula for e-commerce is a methodological error, not an analytical insight.
Recent articles
MegaFon, BMW, VTB and more 240+ companies. See how it looks in the showcase reel:
Our branding agency is approached to order:
Branding agency services
