Why footfall breaks the event-led plan
The generic plan proves each goal with an event that a tool records for the person who did it. A store visit has no such event. The website sees someone search for the nearest branch and then loses them; the ad platform may later infer, from location data on some phones, that a person who saw an ad was near the store; the till records a sale with no link to either. Each of those is a fragment of the same journey, and none of them can be stitched to the others reliably at the level of a person.
That pushes the plan in two directions at once. Downwards, towards signals the website and the business profile can capture, which stand in for the visit. And upwards, towards aggregate comparisons of sales by region and week, which are the only place the actual outcome is counted. The local marketing hub describes the local system the plan has to measure: presence, reputation, local advertising and the multi-location layer.
Decisions that change for store visits
| Plan component | In the generic plan | In a store-visit plan | Decision it forces |
|---|---|---|---|
| Goal evidence | A tracked conversion | Website intent proxies and business-profile actions | Which proxies are trusted enough to carry a target |
| Platform conversions | Observed and attributed | Store visits modelled from location data, available above eligibility thresholds | Whether to use them for bidding, reporting or neither |
| Outcome data | Analytics or CRM, per user | Till data by store and week | How stores map to media regions |
| Proof of effect | Attribution report | Exposed against holdout regions | Which regions are held out, chosen before launch |
| Segments | Channel, device, audience | Store, catchment, region | Which store attributes explain differences |
| Owners | Marketing and analytics | Marketing, store operations and finance | Who supplies sales data, and how fast |
Choose the website proxies
Store locator searches, requests for directions, clicks on a store's phone number and stock or availability checks are the closest things to a visit that the site records. Each is a KPIA metric tied directly to an objective. candidate, but they differ in how near they sit to the trip: a directions request usually precedes a visit, a page view of opening hours may only be curiosity. Specify each as an event with a store identifier attached, so it can be read per location. The trade-off is between many weak proxies and a few strong ones; choose two or three and check, using a past period, that they rose and fell with regional sales before trusting them.
Treat platform store visits as modelled estimates
Some ad platforms report store visits for advertisers who meet their eligibility thresholds, inferred from location data on a sample of phones and scaled up. The figures are useful and they are models: they credit visits under the platform's own AttributionAssigning conversion credit to touchpoints. rules and cannot tell a visit caused by the ad from one that would have happened anyway. The plan should record which platforms report them, whether the account is eligible, and what they are used for. Choose to use them for comparing campaigns inside one platform, never for adding across platforms or proving the total effect. The geo-targeted campaigns guide covers setting up store visit and call goals in the campaigns themselves.
Match till data to media by region and week
The outcome the business cares about, sales in stores, is counted by the till system, and it cannot be joined to individual ad exposures. The plan therefore defines a geography that both sides share, such as groups of postcode districts around each store, and a weekly calendar both sides use. Media delivery and spend are reported into the same regions and weeks. The cost is precision: regional data hides who responded. The gain is that the measure is the real outcome. Choose regions large enough for weekly sales to be stable, and get store operations to commit to a delivery date for the data, since a late feed stalls every reading.
Design holdout regions before the campaign starts
Regional matching shows that sales rose where media ran; only a comparison with similar regions where it did not run shows the rise was caused by the media. Pick holdout regions using past sales so they track the exposed regions closely, and fix them before launch. This is IncrementalityConversions caused by marketing. testing applied to footfall; the incrementality testing guide covers geo designs and how to read them. Holding regions back costs sales if the campaign works; choose a holdout share the business can afford and say so in the plan, so nobody switches the holdouts on mid-flight.
Read business-profile actions beside media data, per location
Calls, direction requests and website clicks from each location's business profile are among the few signals tied to a specific store. Read them weekly per location beside the media delivered into that store's catchment. The Google Business Profile guide covers keeping each profile accurate, and the guide to location pages and listings at scale covers the single source of location data that makes per-store reporting possible. Choose to add store attributes, such as size and parking, as segments, because they often explain more than the media does. When one location's calls and direction requests fall while its profile still appears, the diagnosis for Business Profile calls and directions dropping checks the profile's details, reviews, numbers and duplicates before the media is blamed.
Decision points in a worked store-visit plan
A footfall measurement plan for an invented garden-centre group, Fernhollow
Fernhollow runs 18 garden centres. Its spring campaign runs for five weeks from mid-March, with £150,000 across local search, social and digital outdoor near 14 of the centres. The other 4, matched on the previous two springs' weekly sales, are held out. The plan, condensed:
| Goal | KPI | Source | Grain | Target | Read |
|---|---|---|---|---|---|
| People plan a visit | Directions requests and store-page stock checks | Website events with store_id, business profiles | Store, week | Up 25% on last spring in exposed stores | Weekly |
| Platforms see visits | Modelled store visits per campaign | Search platform, account eligible | Campaign, week | Used only to compare campaigns | Weekly |
| Visits turn into sales | Till transactions and revenue | Point-of-sale feed, delivered Tuesdays | Store catchment, week | n/a | Weekly |
| Media causes the uplift | Sales growth in exposed against holdout stores | Point-of-sale feed | Group, campaign period | Exposed stores 6 points ahead of holdouts | Two weeks after the flight |
| Locations stay accurate | Profile accuracy audit pass rate | Listings tool | Store | 100% before launch | Once, then monthly |
In the first spring the exposed stores grew sales 11% and the holdouts 4%, a gap of 7 points. Platform store visits suggested a much larger effect, because they also counted regular weekend visitors.
The plan's value lay in the decision taken in January, not in March: choosing the four holdout centres while everyone still agreed they were fair. Without them, the spring rise would have been credited to the campaign in full, since garden centres are busy every spring.
Measuring the plan: KPIs and cadence
The KPIs for a footfall plan sit at three levels. Intent: directions requests, calls, store-locator searches and stock checks per store. Modelled visits: platform store-visit counts per campaign, read within each platform. Outcome: store transactions and revenue per region and week, and the gap between exposed and holdout groups. Read intent and media delivery weekly per store. Read sales weekly but judge the campaign only after the flight, on the full period, since weather and bank holidays swamp single weeks in retail. The plan also tracks its own inputs: the date the till feed arrived, the share of events carrying a store identifier, and the profile audit result.
Where a store-visit plan misleads
Modelled visits are built on samples of phones whose owners share location, which may not resemble the customer base. Regional comparisons absorb local events, a competitor opening, roadworks or a local news story, that land in one group and not the other. Website proxies favour customers who research online and miss the ones who simply drive past. And a holdout shows the effect of the campaign as a whole, not which channel within it did the work; splitting that needs separate tests.
The store visits objective hub sets out the KPI set this plan collects and the channels whose effect it has to measure.
Common mistakes
- Using modelled visits as the outcome. They steer campaigns within a platform; they do not prove footfall.
- Choosing holdouts after the campaign. Regions picked later tend to be the ones that make the result look good.
- Events without a store identifier. A proxy that cannot be read per location loses most of its value.
- Reading single weeks. Weather and holidays move footfall more than media does in any one week.
- Letting the till feed slip. If sales data arrives late, the plan reports only proxies.
- Ignoring listing accuracy. Wrong hours or locations on profiles undo the campaign at the last step.
